Making the World a better place with Information Systems
Page 1
Australasian Conference on Information SystemsFilippou, Cheong & Cheong2019, Perth Western AustraliaPersuasion in a DLE516Evaluating Persuasion in a Digital Learning EnvironmentFull PaperJustin Filippou1School of Computing and Information SystemsUniversity of MelbourneMelbourne, AustraliaEmail: justin.filippou@unimelb.edu.auChristopher CheongSchool of Business IT & LogisticsRMIT UniversityMelbourne, AustraliaEmail: christopher.cheong@rmit.edu.auFrance CheongSchool of Business IT & LogisticsRMIT UniversityMelbourne, AustraliaEmail: france.cheong@rmit.edu.auAbstractThe massification of higher education has produced cohorts of students with varying motivation andability to meet their academic potential. Providing individualised support is not always feasible forinstructors as class sizes continue to grow, so this research evaluates the persuasive design of a digitallearning environment (DLE) to address the aforementioned issue. A system with persuasive featurescalled Task-Test-Monitor (TTM) was used by students for a semester at an Australian university. At theconclusion of the semester, students were surveyed on their experience of using the system. Resultsshowed students were strongly in favour of using such a system to help them study, with a significantportion of respondents reporting that the system influenced how they studied. Educators and systemdesigners can benefit from these findings by applying persuasive design principles used in this researchin their own pedagogy or system designs.Keywords: persuasive systems, higher education, digital learning environments, technology enhancedlearning, behaviour change1 Justin Filippou was employed at RMIT University while conducting this research.
Page 2
Australasian Conference on Information SystemsFilippou, Cheong & Cheong2019, Perth Western AustraliaPersuasion in a DLE5171 INTRODUCTIONMore students than ever are attending university to complete undergraduate studies and there is abroader range of these students’ academic ability. This has been described as the “massification” ofeducation (Guri-Rosenblit et al. 2007; Norton et al. 2013) and has impacted universities across the world(Chan and Huang 2018; Mohamedbhai 2014; Yang 2004). Problematically, the learning environmentsused to teach students have not adapted to meet the needs of the more diverse student cohort (QUT2014). One way to address this is for modern universities to take a more student-centred approach indesigning learning systems by moving from a “one size fits all” style to a more tailored learningenvironment (Sledge and Fishman 2014). Students are also diversifying their interests away from formalstudy and are increasingly working part-time while they study (ABS 2013), dividing their attention andcreating larger amounts of time away from academic responsibilities.As more students enter higher education with varied interest, motivation and ability to perform at thehighest level, some form of intervention will be required to better engage students. However, withenrolment numbers continuing to increase, it will become decreasingly feasible for teaching staff toprovide the tailored learning experience expected by students. Personalising learning for individualstudents may take too much time away from teaching responsibilities to the cohort of students in generaland so a more feasible solution is to leverage an information system to assist both teaching staff andstudents. One solution that could be used is a persuasive system, which is one that is designed andimplemented in such a way as to encourage deliberate behavioural responses from end users (Fogg2002; Oinas-Kukkonen and Harjumaa 2009). Such systems have been shown to be effective in healthand exercise (Karppinen et al. 2016; Langrial et al. 2012), environmental sustainability (De Kort et al.2008) and computer security (Forget et al. 2008). The aim of this research is to evaluate a persuasivesystem for the learning environment (or Persuasive Learning System) and the ways that it can be usedin encouraging students to engage with coursework.2 BACKGROUNDThe research presented in this paper forms one segment of a larger research project on persuasivesystems for learning. Overall there are three phases to the project, of which this research marks thebeginning of the third phase. Phase 1 surveyed students to understand their study behaviour anduncover barriers to successful study (Filippou et al. 2016). Phase 2 enhanced a learning system usingpersuasive principles and incorporated the study behaviour findings from the first phase. The aim ofphase 3 is to evaluate the persuasiveness of the target learning system in order to inform thedevelopment of design principles that other educators and learning designers will be able to implementto address issues of engagement. This section provides context for the targeted study behaviours andexplains how behaviour can be triggered, the process in which a persuasive system can be designed andevaluated and finally how persuasive systems have been a useful tool in identifying issues in otherdisciplines.2.1 Study Behaviour ModelsTo carry out phase 1 (Filippou et al. 2016), an existing study behaviour survey instrument called theMotivated Strategies for Learning Questionnaire (MSLQ) was used. The instrument features 81 itemsthat cover a range of different study behaviours across several categories of behaviour and motivation(Pintrich 1991). Automatic linear modelling (ALM) was performed to determine the most impactful ofthese items on two measures of academic performance: a self-perceived performance measure and aresults-based measure. These two measures were used to provide a balance of subjective and objectiveperspective from respondents. The results of the ALM were used to perform multiple linear regression(MLR) in order to identify three to five of the most impactful behaviours on each of the performancecriteria. The process was then repeated on two subsets of the data, current students and alumni, toanalyse any differences between the immediacy of studying against the reflective responses of graduates.The models informed the enhanced persuasive design of the digital learning environment used for thisresearch. The system and its persuasive features are further explained in section 3.2.2 Behaviour ChangeThere is an unstated assumption that behaviour simply occurs, but it can be argued that behaviour isdriven by two aspects: motivation and ability. The theory of planned behaviour, for example, representsthese constructs as behavioural intention and perceived behavioural control, respectively (Ajzen 1985).However, it does not specify any trigger mechanism to begin the process of behaviour change. One modelthat does this is the Fogg Behavioural Model (FBM). The FBM helps improve the understanding of
Page 3
Australasian Conference on Information SystemsFilippou, Cheong & Cheong2019, Perth Western AustraliaPersuasion in a DLE518behavioural processes by mapping the relationship between motivation and ability with the pointbehavioural triggers can be effective (Fogg 2009). Triggers can only be effective when an individual hasboth sufficient motivation and ability to perform the action. Unless an individual has the necessary skillsto do something, they will be unable to do so no matter their level of motivation when prompted. A lackof motivation to do something despite having the capability will also fail to encourage the person to actwhen triggered. The FBM is a useful tool to diagnose issues with behaviour or at least to betterunderstand how behaviour functions generally. However, the model does not provide guidance on howto systematically design an intervention to achieve desired behaviour change in people. A frameworkthat can be used for such guidance is Persuasive Systems Design (PSD).2.3 Persuasive Systems DesignThe PSD model merges various work on psychology, influence, behaviour and system design into threekey phases: (1) understanding key issues behind persuasive systems, (2) analysis of the persuasioncontext and (3) the design of the system qualities (Oinas-Kukkonen and Harjumaa 2009). The finalphase of the PSD model is analysing the design of the system features that implement various persuasiveelements. These can be categorised into four principles of support: primary task, dialogue, credibilityand social. Primary task support focuses on helping users complete their main objective. Dialoguesupport is concerned with improving the human-computer interaction. Credibility support is designedto improve users’ perception of a system and to ensure that they do not feel manipulated or coerced.Finally, social support leverages the human relationships that compel people to behave in certain ways.Each of these principles can be broken down into specific components of persuasion. Such componentsused in this research will be explained in later sections.2.4 Related WorkOne of the strengths of the PSD is that it can be used for both designing and evaluating persuasivesystems. In a study that used the PSD as an evaluation tool to evaluate the persuasiveness of variousweight loss websites, the authors were able to find common strengths and weaknesses and the overallpersuasiveness of those types of systems (Lehto and Oinas-Kukkonen 2010). Underutilised systemcharacteristics were found, including tailoring and rewards, and social support in general. The studyalso highlighted the weak use of dialogue support and the effective use of expert moderated socialfeatures. The PSD has also been used to assess software designed to assist with medication management(Win et al. 2017). In that study, it was found that most systems provided primary task and dialoguesupport, while social support was generally not strongly implemented. Collectively, these studiesdemonstrate the ability of the PSD to diagnose issues with and provide insight into the levels ofpersuasion in a system, which is the primary goal of this research. In the following section, the systemthat the PSD will be applied against for evaluation will be described.3 SYSTEM OVERVIEWThe target system used in this research is called Task-Test-Monitor (TTM). TTM is a Next-generationDigital Learning Environment that supports students in completing weekly tutorial tasks. TTM has asimple interface that is designed to allow students to find what they require quickly. At a high level,students have access to three areas: course content, performance tracking and analytics. The coursecontent area is where students spend most of their time (see Figure 1) since it is where the weeks of thesemester are listed, and tasks and tests are accessed. Tasks are broken down into bite-sized pieces (seesection 3.1) to make them easier to complete. On the right-hand side of the screen there are theautomated to-do list (see section 3.2). The performance area of TTM tells students how they areprogressing through their course (see section 3.3) by allowing them to see which tasks and tests havebeen completed, the average score for tests, as well as overall percentage progression. All of theseindividual components combine to create the intended persuasive effect of the DLE. The following sub-sections detail the design choices for each persuasive feature and how they align with the existingbehaviours identified in phase 1 of the broader research project.
Page 4
Australasian Conference on Information SystemsFilippou, Cheong & Cheong2019, Perth Western AustraliaPersuasion in a DLE519Figure 1: Overview of the TTM structure3.1 Bite-sized tasks and testsTo varying degrees, students struggle with completing coursework. The reasons why include feelingunmotivated, lacking persistence when the material is difficult to understand, fearing the results of testsand struggling to record and use notes (Filippou et al. 2016). One explanation is that course materialcan be quite long since it is designed to be initiated in-class and then continued at home. A micro-learning approach (Hug and Friesen 2007) can address these common student issues by breaking downcontent into smaller manageable pieces, targeting the following five study behaviours and strategiesidentified in phase 1 of the research:1. I often get so lazy or bored when I study for a class that I quit before I finish what I planned todo (negative impact on academic performance).2. When a subject’s work is difficult, I either give up or only study the easy parts (negative impacton academic performance).3. I’m certain I can understand the most difficult material presented in the readings for a subject.4. When I take tests, I think of the consequences of failing.5. I rarely find time to review my notes or readings before an exam.In line with the micro-learning pedagogy, TTM breaks down traditionally lengthy tutorial tasks intosmaller 15 to 20 minute “bite-sized” components. This technique heavily leverages the concept of“reduction” from the PSD model, which involves reducing complex behaviour into simple tasks toincrease the benefit/cost ratio of performing behaviour.Another problem to overcome when attempting to encourage students to work is procrastination.Students tend to put off completing work if the task appears too difficult, particularly if the content isnot interesting to them (Blunt and Pychyl 2000; Harrington 2005; Milgram et al. 1988). Therefore, bydividing tutorial tasks into smaller pieces, students would be required to complete a single small task,which would appear far more manageable and improve the likeliness of completion. This design patternaddresses the issue in behaviour 2 (listed above), which indicates that when work is difficult, studentslose motivation to complete it, which has a negative impact on self-perceived academic performance.In TTM, each task outlines prerequisites before commencing. After the task is completed, students arepresented with the next steps (contained in the task instructions), thereby providing a basic level of“tunnelling”, which involves guiding users through a process or experience. Tasks are also labelled byweek and then task number, indicating order. Tunnelling is particularly useful here to help reduceprocrastination by providing direction, indirectly reducing the effort required of students to deduce whatthey should be doing next. Most tasks have an accompanying test consisting of five multiple-choice
Page 5
Australasian Conference on Information SystemsFilippou, Cheong & Cheong2019, Perth Western AustraliaPersuasion in a DLE520questions that provide immediate feedback to students about their selected answers, supporting themin understanding the material (addressing behaviour 3) without the consequences of failing (addressingbehaviour 4). Students can attempt the tests an unlimited amount of times, enabling “rehearsal” (simplybeing able to rehearse the desired behaviour), which can be valuable in preparation for assessments andexaminations (encouraging behaviour 5). The tasks and tests are organised by week and listed innumerical order, reinforcing tunnelling albeit at a higher level than the “next steps” mentionedpreviously. The combination of tunnelling and rehearsal could address the issue of feeling lazy or bored(behaviour 1) as it could minimise the chances of students getting distracted from what they need to doin TTM by making it easy to follow and understand.3.2 Automatic to-do listWhile breaking down tutorial tasks into smaller pieces assists with reducing the perceived effortrequired in completing course content, it does not guarantee students will continue to work throughcontent. That is, the bite-sized pieces make it easier to begin an interaction, but do not solve the longer-term issues of completing all the work. The problem of long-term engagement is evident in a behaviouridentified in phase 1: “I find it hard to stick to a study schedule.” An undesirable consequence of breakingdown weekly tutorial tasks into bite-sized pieces is that it results in a higher number of tasks and teststhat students need to complete. Students may have difficulty keeping track of what they have and havenot done and without support this could reintroduce procrastination. To alleviate this problem, built-into-do lists were designed into TTM to automatically track what has been completed, including both tasksand tests. When manual management of a to-do list is required, it can be neglected after a period of timeas people often have difficulty managing it (Bellotti et al. 2004), which may be why students “find it hardto stick to a study schedule”. To make sticking to a schedule easier, the TTM automated to-do list updatesas soon as a student submits a task or test in the system. To-do items are timed to only appear in theirrelevant week, which avoids overwhelming students by listing every task and test to complete for thewhole semester. The intended effect of the to-do list is that students will be more likely to seek out thetasks and tests they are yet to complete and then do so.3.3 VisualisationCapturing students attention quickly has been shown to be an effective persuasive strategy (Orji et al.2018). One way to achieve that is through visualisation. The most prominent use of visualisation in TTMare the use of bars measuring progression of course completion, the task and test performance chartsand the recent test attempts chart. These are displayed at various locations and provide self-monitoringby supporting the primary task in the system. All three implementations visually represent differentlevels of abstraction for learning performance. Being able to measure learning performance can helpsupport one of the behaviours identified in the first phase of the broader research project: “When I study,I set goals for myself in order to direct my activities in each study period”.Students can quickly digest high-level information about their overall performance using the progressbars, which can then be progressively broken down into the low-level details of attempts made for aspecific test. This scale of visual representation of progress provides the PSD elements of“personalisation” and a light implementation of “reduction”, as students do not have to expend greatefforts to quickly understand how they are performing at various levels. Students can also easily set goalsfor themselves using these data, thereby addressing the aforementioned study behaviour. This is avaluable feature considering that students are likely to use the system in bursts, particularly if they haveexternal work commitments. Quickly making students aware of what their next task or test is and thenreminding them of their overall progression in the course is expected to more effectively trigger studentsto re-engage with coursework. As a result, less time and effort are expended on re-acquainting studentswith that they should be doing, and instead more time and effort can be dedicated to completingcoursework.The most basic view of student progress in TTM is found in the performance section, where studentscan see how much they have completed for each course using TTM. To register as completed, studentsmust score 80% or higher. The figure of 80% is set in accordance with the mastery learning theory (Blockand Burns 1976), but it also acts as a persuasive feature by indicating a suitable goal for students toachieve. In addition, the progress bars leverage the idea of goal gradients, where people are inclined tosee something reach completion and often work harder to achieve this the closer they get to the target(Kivetz et al. 2006). Although the 80% result to register as complete and the nature of progress bars areindividually subtle in design, they should have a combined positive persuasive effect of motivatingstudents to continue working throughout a semester and help them identify problem areas for tests.
Page 6
Australasian Conference on Information SystemsFilippou, Cheong & Cheong2019, Perth Western AustraliaPersuasion in a DLE5214 METHODOLOGYFor persuasion to be measured, students need to use the persuasive learning system for a considerableamount of time. In this research, students use TTM in class for an entire semester of 12 weeks. Studentsare shown all of the features of the system and how to use them in the first week of classes. TTM is usedin several business information systems courses, including systems development, e-business and mobilesystems, as well as in research methods in social science courses. Students are given generalrecommendations to first complete tasks and then complete tests, in that order. Instructors are asked toprovide regular consultations every two to three weeks to monitor whether students complete work andaddress any concerns that arise. At the conclusion of the semester, students were invited to complete asurvey questionnaire about their experience of using TTM.4.1 Survey instrument designA survey instrument was constructed to collect data about student demographic details, system usageand persuasive impact. The demographic details include gender, age, student type and study load. Thesystem usage section enquires about aspects of the system, such as how often the student completestasks, and tests and how often students use the system. Answers are given using a Likert scale. Thepersuasive impact questions are adapted from a related study that surveyed users about a health-basedbehaviour change support system (Lehto et al. 2012). That study’s questionnaire survey consisted of 21questions covering six constructs: primary task support, dialogue support, perceived credibility,design aesthetics, perceived persuasiveness and unobtrusiveness. These constructs align with theconstructs to be measured in this study and given its demonstrated reliability, is a suitable candidate touse as a survey instrument for this research. The questions were modified to suit the learningenvironment of the present study, with some rewording made in some cases. For example, “…providesme with a means to lose weight” is changed to “…provides me with a means to study”. All questions inthe PSD evaluation section use a five-point Likert scale ranging from “very much agree” to “very muchdisagree”.5 RESULTS AND DISCUSSIONThe survey was distributed at the end of the semester. There were 69 total respondents across all thecourses using TTM. Eight of the 69 responses were not sufficiently completed and were excluded, leaving61 usable samples. The data collected skewed heavily towards males and students aged between 18 and25 and were predominantly full-time local students, which is representative of the general population ofstudents in Information Technology related programs (DET 2016). The survey was also open to studentsin Social Science programs, however the majority of respondents were from the information systemsprograms and so it is not expected that the demographic representation will substantially be impacted.Table 1 summarises the demographic (demo.) and percentage of respondents (resp.) of the surveyrespondents.Demo.Resp.Demo.Resp.Demo.Resp.GenderProgramStudy LoadMale67%Bachelor of Business/IT69%Full-time85%Female33%Other (Social Sciences)31%Part-time15%Student typeCourseAgeLocal73%Systems Development25%18–2145%International27%E-Business40%22–2535%Mobile Systems21%26–296%Research Methods in SocialSciences14%30–398%40+6%Table 1: Demographics of survey respondents5.1 Use of Task-Test-MonitorIt was found that the system was regularly used with 16% of students reporting that they used it morethan five times per week, 30% using it three to four times per week and the majority (46%) using it atleast one to two times per week. Without further examination, this at least suggests that TTM was usefuland easy to use for students, given that 92% of respondents used it regularly throughout semester.
Page 7
Australasian Conference on Information SystemsFilippou, Cheong & Cheong2019, Perth Western AustraliaPersuasion in a DLE522Further, 72% of respondents stated that they would like to see TTM used in their other courses. Intentionto use a system in the future is a factor in the technology acceptance model (Davis 1989), so this providesbasic evidence for general acceptance of the system. Respondents were also asked whether TTM assistedthem with managing their time for the course, with 48% of respondents stating that it did help themmanage their time. This is an encouraging result since one negative study behaviour identified in theearlier phases of the broader research project involved students struggling to stick to study schedules.This result indicates that TTM may have had some level of influence in helping students manage thisprocess. However, 36% of respondents were unsure of whether this help occurred, so furtherinvestigation is required. Table 2 presents the results of the three questions related to student use ofTTM.How many times have youused TTM per week onaverage?Would you like to see TTMused in your other courses?Has TTM Helped you inmanaging your time in thiscourse?ScaleRespondentsScaleRespondentsScaleRespondents08%No15%No16%1-246%Unsure13%Unsure36%3-430%Yes72%Yes48%5+16%Table 2: TTM usage survey results5.2 Task completionAnalysing lower-level system usage, the survey results showed that 66% of students regularly completedthe tasks (either most of the time or all of the time), which represents a positive result for TTM’s abilityto encourage task completion. Task completion is optional in TTM as it forms part of the foundationcoursework and therefore is not assessed. These students were likely either not satisfied with the contentor had other issues regarding their study that may have adversely affected their ability to complete tasks.Unsurprisingly, given the active use of TTM, 95% of respondents found the tasks to be at least“somewhat useful”. While 5% did not find them useful at all, it should be noted that only 3% neversubmitted the tasks. This implies that some of the respondents did not find the tasks useful but stillcompleted them. This could be a very minor indication of TTM’s persuasive ability in encouragingstudents to at least attempt some coursework, even if it appears to not be beneficial in the minds of thesestudents. Table 3 shows the respondents views regarding task completion in TTM.How frequently did you complete the tasks?How useful did you find the tasks?ScaleRespondentsScaleRespondentsNever3%Not useful at all5%Some of the time31%Somewhat useful45%Most of the time45%Very useful50%All of the time21%Table 3: Task completion survey results5.3 Test completionFollowing the high levels of completion of tasks, many respondents (97%) also completed the tests, albeitfor varying reasons. The multiple-choice tests were arguably the most flexible component in TTM asthey could be completed either before or after a task, lecture or tutorial class as well as for revision. Themost popular method of completing tests by respondents was after the tutorial class, which suggests thatthey would use the time in class to complete the task, and then test their knowledge afterwards. Twointeresting results from the data are the number of respondents who stated they took the tests earlier inthe semester and then stopped, and the number of respondents who used the tests to catch up on theirstudies. This indicates that time management is an issue for students, as 22% stopped completing testsafter some time. This could be a result of the optional nature of the tests, coupled with an increasinglyheavier workload being placed on students as the semester progresses. Students may be opting to skipthe tests to regain control of their time and may also be a symptom of surface learning as students tendto perceive MCQ-style work as less demonstrative of skills and therefore less important to complete(Scouller 1998). It is unlikely that the TTM system or the content of the tests caused these respondentsto stop since 54% stated that the tests were “somewhat useful” and 43% found them very useful.
Page 8
Australasian Conference on Information SystemsFilippou, Cheong & Cheong2019, Perth Western AustraliaPersuasion in a DLE523Therefore, it is reasonable to conclude that external factors likely led to students either stoppingcompleting tests or needing to catch up on their studies by going back to tests later.In terms of academic performance, the majority of students (75%) were aiming for a score of 100% whencompleting tests. The tests were not assessed and students had unlimited attempts available to them, soa target of 100% for most students suggests that students are intrinsically motivated to perform at thehighest level. Only 5% of students “didn’t care” what results they achieved. It could be argued that withno risk, there is no reward, and so the number of students not caring was expected to be higher. Despite5% not caring about results, 97% of respondents found the tests to be either “somewhat useful” or “veryuseful”. While there was nothing to be gained in terms of tangible outcomes such as grades, it is evidentthat students could make the connection to the knowledge gains they would acquire from completingthe tests, which they could then apply to assessments at a later point in time.When assessed using the Fogg Behavioural Model (Fogg 2009), it is clear in this instance that studentswere highly motivated, and the bite-sized structure of the tests and consequence-free nature of attemptsmeant their ability to carry out this action was high. This indicates that the recent history charts placeddirectly above the buttons to access a test may have been an effective trigger in compelling students toattempt the tests and continue doing so until they reached their target of 100%. This can be evidencedby 79% of respondents also stating that they attempted the tests at least two to three times, with 6%attempting them four or more times on average. This result gives an early indication that students,contrary to popular belief, are not unmotivated to study; rather, it may be more likely that restrictionson their ability to study hampers their academic performance. Table 4 contains the results in whichparticipants reported their interaction with the test functionality of TTM.What score were youaiming to achieve on thetests?How useful did you find thetests?Generally, how many attemptsdid it take you to achieve yourdesired score?ScaleRespondentsScaleRespondentsScaleRespondentsDidn’t care5%Not useful at all3%115%Pass3%Somewhat useful54%2-379%60%2%Very useful43%4 or more6%80%15%100%75%Which statement best reflects how you completed the tests?ScaleRespondentsI didn’t take the tests3%I did the tests when I was catching up on my studies27%I took the tests regularly at first but then stopped22%I regularly took the tests after my tutorial31%I regularly took the tests before my tutorial17%Table 4: Test completion survey results5.4 Impact on study behaviourThe results for TTM’s ability to persuade study behaviour were positive. Of respondents, 46% agreedthat TTM made them reconsider the way they studied, and 61% stated that it helped change theirapproach to study. Furthermore, respondents reported that TTM was able to encourage them (47%agreed), instil confidence (49% agreed) and have an influence on them (41% agreed). Very few disagreedcompletely, however, there were considerable numbers of respondents who neither agreed nordisagreed that TTM influenced their study process or behaviour. One explanation for this could berelated to the focus of the subject in the sentence. That is, the items about persuading the study processimply that respondents have control over their behaviour and TTM simply helped them achieve it.Misremembering one’s attitude pre- and post-intervention is possible (Bem and McConnell 1970),which may be what has occurred in this instance. Alternatively, the questions that lead the respondentto consider how it made them “reconsider” the way they study and how it changed their “approach” tostudying had lower levels of neutrality. Observing the items related to impact on behaviour, thosequestions imply TTM directly altered their behaviour regardless of whether it was desired. Respondentsmay have had a natural hesitation to admit that a system could have that level of control or influence onbehaviour and may be more comfortable with persuasion occurring when they are in control of theirlearning process with TTM simply assisting them. This is an important finding for persuasive DLE
Page 9
Australasian Conference on Information SystemsFilippou, Cheong & Cheong2019, Perth Western AustraliaPersuasion in a DLE524because it indicates that students need to believe they retain control even if they are being influenced,even when there is clearly no coercion taking place in a system. Table 5 contains reported results on howTTM impacted student study behaviour.TTM makes me reconsider the way I study.TTM helps me change my approach to studyingScaleRespondentsScaleRespondentsStrongly Disagree5%Strongly Disagree3%Disagree21%Disagree10%Neutral28%Neutral26%Agree30%Agree43%Strongly Agree16%Strongly Agree18%TTM encourages me.TTM instils confidence.TTM has an influence on me.ScaleValueScaleValueScaleValueStrongly Disagree 3%Strongly Disagree 3%Strongly Disagree3%Disagree7%Disagree7%Disagree23%Neutral43%Neutral41%Neutral33%Agree34%Agree42%Agree33%Strongly Agree13%Strongly Agree7%Strongly Agree8%Table 5: Impact on study behaviour survey results6 IMPLICATIONS AND RECOMMENDATIONSThe results in the previous section clearly indicate that the design of the TTM Digital LearningEnvironment was conducive to persuasion and that students demonstrated a preference for this kind ofintervention. The results also provide evidence of microlearning being well-suited to persuasive designas the nature of making tasks more manageable fits very well with a number of the persuasive featuresdesigners can implement. This result does not preclude other learning pedagogies from working withpersuasive design, rather that in this case it appears to have been a good match.A recommendation for educators resulting from this research is to assess their current digital learningenvironment against the Persuasive Systems Design framework and identify areas that could beimproved. Feature implementation should be carried out with a student-centred focus by firstperforming an appropriate study of current behavioural issues exhibited by students. The design of apersuasive system intervention will also need to be tailored to particular cases for issues of eithermotivation or ability. More specifically, educators and system designers should prioritise the concept ofreduction in order to address issues of procrastination deriving from a lack of motivation or ability.7 LIMITATIONS AND FUTURE WORKThis research is exploratory in nature and so the results are not generalisable in their current form. Moredata will need to be collected across various study disciplines to validate the results. Lack ofgeneralisability was in part due to the available population of students using TTM. Only students fromseveral business IT and social science courses were actively using the system at the time of this research.It also restricted the types of students that could participate in the study to IT and social sciences, andso the findings were specific to the context of those courses.The results of this research are encouraging as there is evidence to suggest that learning systems can bepersuasive in influencing student study behaviour. This research is a foundational step towards betterunderstanding how that persuasion occurs in a DLE, and what factors contribute to it being effective.To achieve this, future work may focus on conducting a factor analysis to reveal a better understandingof how and why students felt TTM had an influence on their learning experience. This would be achievedby examining the impact certain features have on learning behaviour. Such an analysis would bebeneficial to researchers as the key aspects of persuasion can be isolated and then replicated in otherlearning systems. Longer term goals would be to enhance the PSD by providing more guidance todesigners on how to design for persuasion, particularly in a learning system.
Page 10
Australasian Conference on Information SystemsFilippou, Cheong & Cheong2019, Perth Western AustraliaPersuasion in a DLE5258 SUMMARYThis research has presented an evaluation on the potential for a learning system to persuade students tobetter engage with their coursework. There is a growing need for Digital Learning Environments to bemore persuasive, as universities are facing a wider range of motivated students. To address this problem,this research selected a Next-Generation Digital Learning Environment called Task-Test-Monitor(TTM) and assessed its persuasive ability. This research found that the design of a learning system canhave an impact on student behaviour and that students are aware that such a system can be influential.In particular, TTM provided high-levels of primary task support and credibility, which made it aneffective tool for persuasion. Students self-reported that it encouraged and influenced them to changetheir approach to study, which provides an opportunity for future research to be conducted in this areato identify factors that contributed to these outcomes.9 REFERENCESABS. 2013. "Hitting the Books: Characteristics of Higher Education Students."Ajzen, I. 1985. From Intentions to Actions: A Theory of Planned Behavior. Springer.Bellotti, V., Dalal, B., Good, N., Flynn, P., Bobrow, D. G., and Ducheneaut, N. 2004. "What a to-Do:Studies of Task Management Towards the Design of a Personal Task List Manager," Proceedingsof the SIGCHI conference on Human factors in computing systems: ACM, pp. 735-742.Bem, D. J., and McConnell, H. K. 1970. "Testing the Self-Perception Explanation of DissonancePhenomena: On the Salience of Premanipulation Attitudes," Journal of personality and socialpsychology (14:1), p. 23.Block, J. H., and Burns, R. B. 1976. "Mastery Learning," Review of research in education (4), pp. 3-49.Blunt, A. K., and Pychyl, T. A. 2000. "Task Aversiveness and Procrastination: A Multi-DimensionalApproach to Task Aversiveness across Stages of Personal Projects," Personality and IndividualDifferences (28:1), pp. 153-167.Chan, S.-J., and Huang, T.-M. 2018. "The Development and Progress of Higher Education Research inTaiwan: Massification Matters," in Researching Higher Education in Asia: History,Development and Future, J. Jung, H. Horta and A. Yonezawa (eds.). Singapore: SpringerSingapore, pp. 195-211.Davis, F. D. 1989. "Perceived Usefulness, Perceived Ease of Use, and User Acceptance of InformationTechnology," MIS quarterly), pp. 319-340.De Kort, Y. A., McCalley, L. T., and Midden, C. J. 2008. "Persuasive Trash Cans: Activation of LitteringNorms by Design," Environment and Behavior (40:6).DET. 2016. "Ucube." Australian Government Department of Education and Training.Filippou, J., Cheong, C., and Cheong, F. 2016. "Modelling the Impact of Study Behaviours on AcademicPerformance to Inform the Design of a Persuasive System," Information & Management (53:7),pp. 892-903.Fogg, B. J. 2002. "Persuasive Technology: Using Computers to Change What We Think and Do," in:Ubiquity. p. 5.Fogg, B. J. 2009. "A Behavior Model for Persuasive Design," Proceedings of the 4th internationalconference on persuasive technology: ACM, p. 40.Forget, A., Chiasson, S., van Oorschot, P. C., and Biddle, R. 2008. "Persuasion for Stronger Passwords:Motivation and Pilot Study," in Persuasive Technology. Springer, pp. 140-150.Guri-Rosenblit, S., Šebková, H., and Teichler, U. 2007. "Massification and Diversity of Higher EducationSystems: Interplay of Complex Dimensions," Higher Education Policy (20:4), pp. 373-389.Harrington, N. 2005. "It’s Too Difficult! Frustration Intolerance Beliefs and Procrastination,"Personality and Individual Differences (39:5), pp. 873-883.Hug, T., and Friesen, N. 2007. "Outline of a Microlearning Agenda," Didactics of Microlearning.Concepts, Discourses and Examples), pp. 15-31.
Page 11
Australasian Conference on Information SystemsFilippou, Cheong & Cheong2019, Perth Western AustraliaPersuasion in a DLE526Karppinen, P., Oinas-Kukkonen, H., Alahäivälä, T., Jokelainen, T., Keränen, A.-M., Salonurmi, T., andSavolainen, M. 2016. "Persuasive User Experiences of a Health Behavior Change Support System:A 12-Month Study for Prevention of Metabolic Syndrome," International journal of medicalinformatics (96), pp. 51-61.Kivetz, R., Urminsky, O., and Zheng, Y. 2006. "The Goal-Gradient Hypothesis Resurrected: PurchaseAcceleration, Illusionary Goal Progress, and Customer Retention," Journal of MarketingResearch (43:1), pp. 39-58.Langrial, S., Lehto, T., Oinas-Kukkonen, H., Harjumaa, M., and Karppinen, P. 2012. "Native MobileApplications for Personal Well-Being: A Persuasive Systems Design Evaluation," Pacific AsiaConference on Information Systems, p. 93.Lehto, T., and Oinas-Kukkonen, H. 2010. "Persuasive Features in Six Weight Loss Websites: AQualitative Evaluation," in Persuasive Technology. Springer, pp. 162-173.Lehto, T., Oinas-Kukkonen, H., and Drozd, F. 2012. "Factors Affecting Perceived Persuasiveness of aBehavior Change Support System," in: Thirty Third International Conference on InformationSystems. Orlando, Florida.Milgram, N. A., Sroloff, B., and Rosenbaum, M. 1988. "The Procrastination of Everyday Life," Journalof Research in Personality (22:2), pp. 197-212.Mohamedbhai, G. 2014. "Massification in Higher Education Institutions in Africa: Causes,Consequences and Responses," International Journal of African Higher Education (1:1).Norton, A., Sonnemann, J., and Cherastidtham, I. 2013. "Taking University Teaching Seriously," p. 6.Oinas-Kukkonen, H., and Harjumaa, M. 2009. "Persuasive Systems Design: Key Issues, Process Model,and System Features," Communications of the Association for Information Systems (24:1), p. 28.Orji, R., Reilly, D., Oyibo, K., and Orji, F. A. 2018. "Deconstructing Persuasiveness of Strategies inBehaviour Change Systems Using the Arcs Model of Motivation," Behaviour & InformationTechnology), pp. 1-17.Pintrich, P. R. 1991. "A Manual for the Use of the Motivated Strategies for Learning Questionnaire(Mslq)."QUT. 2014. "Review of Qut’s Virtual Learning Environment (Vle)." Queensland University ofTechnology.Scouller, K. 1998. "The Influence of Assessment Method on Students' Learning Approaches: MultipleChoice Question Examination Versus Assignment Essay," Higher Education (35:4), pp. 453-472.Sledge, L., and Fishman, T. D. 2014. "Reimagining Higher Education." Deloitte University Press.Win, K. T., Mullan, J., Howard, S., and Oinas-Kukkonen, H. 2017. "Persuasive Systems Design Featuresin Promoting Medication Management for Consumers," Proceedings of the 50th HawaiiInternational Conference on System Sciences.Yang, R. 2004. "Toward Massification: Higher Education Development in the People’s Republic of Chinasince 1949," in Higher Education: Handbook of Theory and Research. Springer, pp. 311-374.AcknowledgementsThis research was supported by an Australian Government Research Training Program Scholarship.This research project was approved by the RMIT University Business College Human Ethics AdvisoryNetwork (BCHEAN) under project numbers 18732 and 19308.Copyright: © 2019 Filippou, Cheong & Cheong. This is an open-access article distributed under theterms of the Creative Commons Attribution-NonCommercial 3.0 Australia License, which permits non-commercial use, distribution, and reproduction in any medium, provided the original author and ACISare credited.
Page 1
Australasian Conference on Information SystemsToms & Lee2019, Perth Western AustraliaReconfiguration of IoT Nodes384Towards Trusted Seamless Reconfiguration of IoT NodesFull PaperTony TomsSchool of Information TechnologyDeakin UniversityGeelong, AustraliaEmail: tonytoms01@gmail.comKevin LeeSchool of Information TechnologyDeakin UniversityGeelong, AustraliaEmail: kevin.lee@deakin.edu.auAbstractIoT networks are growing rapidly with the addition of new sensors, nodes and devices to existing IoTnetworks. Due to the ever-increasing demand for IoT nodes to adapt to changing environmentconditions and application requirements, the need for reconfiguring these already existing IoT nodes isincreasing rapidly. A reconfiguration of an IoT network includes alterations to the devices connected,changing the behavioural patterns of the devices and modifying the software modules that control theIoT network and devices. Reconfiguring an already existing IoT network is a challenge due to the amountof data loss and network downtime faced when carrying out a reconfiguration procedure in a limitedpower supply environment. This paper proposes an architecture for trusted dynamic reconfiguration ofIoT nodes with the least amount of data loss and downtime. The proposed approach uses multiple IoTnodes to facilitate dynamic reconfiguration.Keywords: IOT, reconfiguration, middleware, seamless data collection
Page 2
Australasian Conference on Information SystemsToms & Lee2019, Perth Western AustraliaReconfiguration of IoT Nodes3851 INTRODUCTIONThe Internet of Things (IoT) is becoming a major trend in computing and is having a high impact bysolving real-world problems such as improving the energy efficiency of IT rooms, enabling smart citiesand supporting logistics. The core idea behind IoT is embedding tiny, networked, electronic devices intothe surrounding physical objects (Iwanicki 2018), thus enabling the everyday real-world objects tointeract with each other. As per recent studies, by the year 2020, there would be around 50 billionsensors and devices connected to the internet (Zhou 2016). As IoT networks become larger and larger,an increasing concern is the degree of heterogeneity and need to evolve IoT deployments.In an IoT network, nodes and sensors are commonly not from a single vendor. A typical IoT network isequipped with multiple types of devices and sensors that are communicating directly or via the Internet.This increased degree of heterogeneity has encouraged the development of flexible IoT networks. Aflexible IoT network would discover these heterogeneous devices dynamically, update the softwarerequired for the devices and provide an updated end user application (Heo 2015).Flexible IoT enables the addition of a nodes and sensors after deployment whilst the network isoperating. They provide agents that enable the networks to handle various communication protocolsand data formats used by different devices (Pazos 2015). However, changes happening to an IoTnetworks won’t be limited to the communication requirements for just a family of devices. An IoTnetwork would eventually face situations where a change in the behavioral patterns of the devices are tobe done or a newly introduced device must be incorporated. Further changes to the network wouldinclude updating the core software modules in different nodes, changing node responsibilities orupdating the node behavior.A flexible IoT network would be capable of handling the any deployment reconfiguration. To do this,requires that IoT nodes and IoT networks are dynamic reconfigurable. A reconfigurable IoT networkwould enable the network to handle all these changing network requirements. It would enable thenetworks to accept new requirements by allowing computation and communication services to evolveover time with minimal disruption (Raagaard 2017). The main advantage of making an IoTreconfigurable is that it allows functionality to be altered without much network downtime or data loss(Pena 2017).For IoT networks to adapt to changing conditions, a reconfigurable IoT node is needed. The mainadvantage of making an IoT node reconfigurable is that it allows functionality to be altered without anyphysical change for the systems (Pena 2017). Thus, making a reconfigurable IoT node would enable theentire IoT systems to be dynamic and easy to maintain.This paper proposes a new method to enable trusted dynamic reconfiguration of IoT networks withoutcompromising the network downtime and providing seamless data collection from the nodes throughoutthe reconfiguration. Reconfiguration is carried out through a middleware that updates nodes and thendynamically switches the running application programs from the old versions to the newly updatedversions with least data loss. To evaluate this approach, a prototype was built using Raspberry Pi’s to actas the IoT nodes and a middleware module is written in Python.The remainder of the paper is as follows. Section II provides a background on reconfiguration in theInternet of Things (IoT). Section III proposes a new architecture for seamless IoT reconfiguration.Section IV describes the implementation of the architecture. Section V presents an evaluation of theproposed architecture. Section VI presents some conclusions and discusses future work.2 BACKGROUND2.1 IoT Network OverviewAn IoT network can be represented as layered architecture, as illustrated in Figure 1. The top most layer,which is the application layer consist of application software and middleware software. The nodes anddevices are connected to the application layer via the physical layer objects like WIFI, Bluetooth, LTE,Internet etc. The system layer consists of system kernel and UDP packets belong to the transport layer.The network protocols and the gateways belong the network layer.
Page 3
Australasian Conference on Information SystemsToms & Lee2019, Perth Western AustraliaReconfiguration of IoT Nodes386Figure 1. IoT Network Layered Structure2.2IoT Structural and Behavioural ReconfigurationReconfiguring a network can be broadly classified as either i) structural reconfiguration or ii)behavioural reconfiguration. The structural reconfiguration deals with replacing a section or a functionof the IoT network where as a behavioural reconfiguration deals with changing the way sensors behavein different scenarios by changing the properties or parameters.One of the most common approach in behavioural reconfiguration is the use a monolithic image whichis a collection of all the drivers, middleware and the core software modules. Any changes performed tothe network is by the use of monolithic image where the middleware would completely change all thedata and replace it with the new image. This approach is flexible but consumes a lot of energy due to thelarge data transfer. Due to the large quantity of data consumption, a better approach to the behaviouralreconfiguration is adopted where the system loads only the updated software modules and link them tothe devices during the runtime.In a behavioural reconfiguration, the modules modify the properties and change the connectionsbetween the components during the network runtime. This approach provides an efficient mechanismfor reconfiguration as it causes the least amount of resource utilization and data transfer. However, thistype of reconfiguration is extremely narrow in terms of the functionality it provides.The key aspect in providing an effective reconfiguration strategy is to adopt the positive sides from bothruntime reconfiguration and behavioural reconfiguration.2.3 Recent approaches for Dynamic reconfigurationAn approach to dynamically add sensors to IoT network is proposed by (Boman 2014) by using GSN,Firebase and IoT data interpreter. By using GSN and virtual sensors the authors where able to add anynumber of sensors as they want into their network. The data interpreter uses an XML file as an input tosee the status of all the devices connected. Firebase, which is a real-time NoSQL database managementsystem owned by google is used as the data storage to enable interfacing with 3rd party application (Prado2017). Even though the system handles the addition and removal of sensors, they won’t deal with thedowntime faced when the core application modules needs an update.(Myint 2017) proposes a wireless sensor network-based system for monitoring water quality. In thesystem architecture, sensors like Ph sensor and ultrasonic sensors are connected to a FPGA system(Hauck 1994). The software applications for the sensors are written in C and is run using an Eclipse IDE.
Page 4
Australasian Conference on Information SystemsToms & Lee2019, Perth Western AustraliaReconfiguration of IoT Nodes387Here the system is said to be reconfigurable using a “. sopcinfo” file, which saves all the configurationinformation. Here the system can only alter parameters like time interval of monitoring, temperaturereading formats etc. That is, this architecture deals with the behavioural change (Sarray 2015) anddoesn’t deal with software/hardware reconfiguration.(Lu 2016) proposes an IoT inventor which is a web-based composer for reconfigurable agentizedservices. Here the authors talk about temporarily reconfiguring the services based on user interest. Thatis instead of tightly coupling sensors, they use sensors to infer context of interest. Here the paper dealswith changing the context in which the sensors work and is not dealing with modifying the IoT networkas such.More work in this area is proposed by (Chi 2014). Here the authors use CPLD designed based onIEEE1451. Since the components belong to IEEE1451 family they are interoperable (Becari 2016). Therethe proposed system can identify newly connected devices intelligently. Here all the drivers of thesensors which are to be used must be written into the CPLD. In case of a new sensor, the system mustgo down and an update is required.There have been substantial efforts in virtualization and cloud computing in the reconfiguration ofvirtual machines and cloud instances (Anton 2010, Ali 2010). There is also considerable effort inadaptively scheduling jobs on Cloud resources (Lee 2011, Lee 2009). This is supported by dynamicallyreconfigurable networking infrastructure (lee 2006). These approaches are in situations with fullycontrolled environments and non-resource constrained. This is opposed to IoT deployments, whichhave difficult deployment scenarios and are resource constrained. Thus, there is still a need for domain-specific reconfiguration solutions for the Internet of Things.3 A TRUSTED ARCHITECTURE FOR SEAMLESS IoTRECONFIGURATION3.1 Design OverviewThe main idea in this proposal is to use multiple nodes to facilitate IoT dynamic reconfiguration withseamless data collection from the sensors, as illustrated in Figure 2. A controller node is used to controlthe activities of these nodes.The controller manages the activities of the connected nodes. At any point in time one of the nodes isactive and collecting the data from all sensors. During a reconfiguration request from the maincontroller, the control is changed to the backup node. The standby node starts up and collect data fromthe sensor and is send to the cloud storage. The main node would stop execution for undergoing theupdates. Once the software deployment activities and done, the main node programs are fired up andthe backup node goes to idle state.All the executables which are fired up at the node side would be running on top of the proposedmiddleware. The middleware also handles all the main controller activities. Different communicationprotocols are used by the middleware to send files between the controller and the nodes.Figure 2. The Node Arrangement for the Architecture
Page 5
Australasian Conference on Information SystemsToms & Lee2019, Perth Western AustraliaReconfiguration of IoT Nodes3883.2 ArchitectureFigure 3 illustrates a middleware software architecture for enabling the seamless reconfiguration of IOTnodes. All the programs which are to be executed, the data to be transferred and all the communicationsbetween the nodes are handled by the middleware. All the programs and software run on top of thismiddleware. The middleware also provides a platform for the dynamic deployment of software modulesinto these nodes. A Cloud-based MQTT, which is a light weight message queuing protocol (Kodali 2016),is used as the Cloud broker to obtain data.Figure 3. The Middleware ArchitectureThis architecture can be broadly classified into 3 versions namely, the controller version, the main nodeversion and the backup node version. The controller node version is implemented at the main controllernode, the node version at the main node and the backup node version at the back up node respectively.The modules inside each of these versions can be categorized into three levels, as detailed in Figure 4.Figure 4. Module Levels in each Version of the Middleware
Page 6
Australasian Conference on Information SystemsToms & Lee2019, Perth Western AustraliaReconfiguration of IoT Nodes389Users directly interact with the API interfaces. All the reconfiguration needs like addition of nodes,deletion of nodes, file transfer, node-side method invocation callers and status check can be achievedusing the API interfaces. All these facilities are exposed to the world as API interfaces.The second level contains all the modules inside the middleware. The file handler module handles allthe API requests related to file management. The API caller module is used to remotely call the APIs ofnode and backup node from the controller node. The configuration handler module handles all theconfiguration needs of the middleware.The third level contains all the libraries and protocol handlers. Different protocols are used for remotemethod invocations, remote file distribution and control transfer.3.3 List of APIsAt the controller version, there are mainly 3 packages namely the config package, the controller packageand the masterCopy package. At the node version there is only 1 main package called the stub package.At the backup node version, the main package name is backupstub package.Table 1. List of all the APIs in the middlewareIn total there are 28 APIs, a listed in Table 1, out of which 10 are API handler. An API handler is an APIthat calls other APIs remotely.The APIs in config package is used to alter the details on configuration xml file. The ‘CreateConfigNode’API is used to add new nodes details into the configuration xml file. The ‘UpdateConfigNode’ API is usedto update the details of the nodes and the ‘DeleteConfigNode’ is used to delete the node details from theconfiguration fileThe ‘masterCopy’ package contains a single API called the ‘copyAllToMaster’ API. This API is used tocopy the updated software to the controller for copying it to corresponding nodes.The controller package contains 2 kinds of APIs. The normal APIs and handler APIs. The ‘testNode’ APIand ‘testBackupNode’ API would check if the node and backup nodes are up and running.‘sendDataToNode’ would send the data (the updated software) to the node and the‘sendDataToBackupNode’ would send the updated software to backup node.The rest of the APIs in controller package are handler APIs which are used to call other APIs from thenode and backup node. The ‘getStatusNodeHandler’ would call the ‘getNodeStatus’ from the main node.
Page 7
Australasian Conference on Information SystemsToms & Lee2019, Perth Western AustraliaReconfiguration of IoT Nodes390This API is used to get the current node status. Similarly, ‘getStatusBackupNodeHandler’ would call the‘getBackupNodeStatus’ which would give the status of the backup node to the controller.The ‘loaderNodeHandler’ API and ‘loaderBackupNodeHandler’ would invoke the ‘loaderNode’ from thenode and ‘loaderBackupNode’ from the backup node respectively. These APIs would load the receivedupdated software from the received folder to corresponding execution folder.The ‘startNodeExecutionHandler’ would call the ‘startNodeExecution’ from the node. This API wouldexecute files from the execution folder whose names are listed in the configuration file’s executables list.Similarly, ‘startBackupNodeExecutionHandler’ would call the ‘startBackupNodeExecution’ from thebackup up node, which would start executing the files at the backup node.‘stopNodeExecutionHandler’ would call the ‘stopNodeExecution’ which in turn would stop the nodeexecution. ‘stopBackupNodeExecutionHandler’ would call the ‘stopBackupNodeExecution’ which inturn would stop the backup node execution.4 IMPLEMENTATION4.1 OverviewTo test the architecture, an implementation was built using Raspberry Pi nodes and Grove PI+ sensors.A Raspberry Pi acts as the controller and two other Raspberry Pi’s act as the node and backup node.Sensors are connected to the Node and backup node Raspberry Pi. The idea is to check if a continuousand seamless data collection from these sensors can be achieved during a reconfiguration process.To test this, the Raspberry Pi’s are first implemented with sensors and other required software alongwith the proposed middleware. The setup would be working, and the data collected would be stored onto the cloud data storage. Then a reconfiguration request is initiated. During the process, the middlewarewould switch the control from the main node to the backup node and check if backup node starts fetchingthe data from the sensors by checking the cloud storage. Once the backup node started successfully, theexecution is stopped at the main node. The updated software and programs are then sent to the mainnode. Once the files are received, the loaderNode API is used to load the files and then the execution ofthe updated software module would be started. Once the main node starts successfully with the updatedsoftware, the backup node is stopped. The data collected at the cloud data store can be analyzed to checkif data reading is seamless.4.2 RequirementsA temperature and humidity sensor is used in this implementation. The data from the temperature andhumidity sensor is being collected from the sensor. The temperature is read in degree Celsius. Thereconfiguration request is be change it from degree Celsius to Fahrenheit reading. The updated softwaremodule should be deployed on to the devices dynamically with no network downtime.During this reconfiguration change, following requirements should be strictly followed to check if thisarchitecture achieved its goals.Requirement 1: Data should be continuously collected to the cloud storage throughout thereconfiguration without any delay (Seamless Data Collection).Requirement 2: There should be no network downtime when carrying out the reconfiguration. (NoNetwork Downtime).Requirement 3: The files should be sent remotely without any physical contact with the nodes. (RemoteReconfiguration)Requirement 4: The node after reconfiguration should successfully run the updated software module.(software Deployment)4.3 Physical SetupTo test the working of the algorithm, 3 Raspberry Pi’s are used. All the Raspberry PI nodes are connectedto the same network using a router. One of the Raspberry Pi nodes act as the controller node and wouldcontrol other Raspberry Pi nodes. The other two Raspberry Pi nodes are installed with node version andbackup node versions. The picture in Figure 5 illustrates the experimental setup.
Page 8
Australasian Conference on Information SystemsToms & Lee2019, Perth Western AustraliaReconfiguration of IoT Nodes391Figure 5. The Experiment setupThe diagram for the setup is given in Figure 6. The controller node is connected to the network via WIFI.The main node and the backup node are connected via RJ45 cables. The sensor is connected to the portD4 and the LCD display is connected to port 12C.Figure 6. Overview of the IoT network.4.4 Initial SetupThe application for reading the temperature and humidity is up and running on to the main node. TheApplication is executed on top of the middleware software module at the main node side. Theapplication reads the values at an interval of 2 seconds each. The values are displayed on the LCD screenconnected along with publishing the collected values into the MQTT broker (Singh 2015) each time.The MQTT broker used here is having the host name as “test.mosquitto.org”. The values are publishedto "seamlessIOTReconfig/data". The value is having the following format:“System Time, IP Address, Temp= Value, Humidity = Value”A client program on a PC connected to the network subscribes to the node and data is logged to log.txt.This is done to check the data is collection frequency and to test if there is any network downtime.4.5 Reconfiguration StepsThe general flow of reconfiguration process is listed below.step 1: Update the main configuration file. Called the API UpdateConfigNode()step 2: Copy the updated program to controller node. Called the API CopyAllToMaster()step 3: Start the backup node. Called the API startBackupNode ExecutionHandler()step 4: Stop the main node. Called the API stopNodeExecution()step 5: Send data to main node. Called the API SendDataToNode()step 6: Load the updated program onto main node. Called the API loaderNode()step 7: Start execution of main node. Called the API startNodeExecution()step 8: Stop the backup node. Called the API stopBackupNodeExecution ()step 9: Send data to the backup node. Called the API sendDataToBackupNode()
Page 9
Australasian Conference on Information SystemsToms & Lee2019, Perth Western AustraliaReconfiguration of IoT Nodes392step 10: Load data to the backup node. Called the API loaderBackupNode()4.6 ResultThe reconfiguration was successful, The LCD starts showing the temperature in Fahrenheit value. Figure7 shows the LCD output. The LCD on top shows the reading in Celsius, before the reconfiguration. TheLCD on the bottom shows the reading in Fahrenheit, which is after the reconfiguration.Figure 7. The LCD Output5 EVALUATIONTo achieve the goal of reconfigurable IoT, all the 4 requirements namely, seamless data collection., nonetwork downtime and remote reconfiguration, software deployment should be achieved. The proposedarchitecture is implemented as described in Section 4.3. Primary evaluation is performed on the datacollected by the device, which is collected into a log.txt file on a client computer.Figure 8. log.txt file. Before and after reconfigurationThese records show each sensor reading with the corresponding reading time (see Figure 8). The log fileon top shows data shows the reading before reconfiguration whereas, the log file on bottom shows thereading after reconfiguration – demonstrating the continuity of service from the IoT node.Figure 9. The configuration file, before and after reconfiguration
Page 10
Australasian Conference on Information SystemsToms & Lee2019, Perth Western AustraliaReconfiguration of IoT Nodes393The data collection time shown in the log file shows a continuous seamless collection of data, whichstates that Requirement 1 is satisfied. A closer look onto the log file will reveal that the time gapbetween each consecutive record are almost constant. The minor difference can be ignored as it iscreated due to network lag during data subscription. Neglecting the minor difference, the resultconcludes that there was no network downtime during reconfiguration stating that Requirement 2 issatisfied. All the remote API calls are made using the node IP addresses from the configuration file. Fromthis, remote reconfiguration requirement, which is Requirement 3 is obeyed.Figure 9 shows the configuration of the node before and after reconfiguration. A closer look into the filereveal that executable program changed its version from v1 to v2. This concludes that softwaredeployment successfully happened during the reconfiguration, thus satisfying Requirement 4.6 CONCLUSIONSeamless reconfiguration of IOT networks without any network downtime and data loss is one of themajor problems faced by engineers for reconfiguring the widely distributed Internet of connected things(IOT) (Xiaohui 2013). The architecture proposed in this paper solves this problem using multiple nodesand the proposed middleware software. The proposed architecture enables seamless collection of datawith no network downtime throughout the reconfiguration procedure. All the reconfiguration processesin this network is handled by the proposed middleware software module which is written in pythonprogramming language. This middleware module exposes many API interfaces that can be used forreconfiguration of the IOT network.To check if the architecture has achieved all its goals, a real implementation of this architecture is alsocreated as part of this paper. The results show that the proposed architecture along with the middlewaresuccessfully achieved the goal of seamless IOT reconfiguration.The proposed architecture successfully demonstrated its ability to do seamless reconfiguration of IOTnetwork without any data loss of network downtime. However, this architecture has some limitationsand should be improved over time.This architecture is particularly designed to run the programs which are coded in python programminglanguage. As of now, the developed middleware won’t be able to run software written in any otherprogramming language. Overtime, the middleware should automatically identify the software languagein which a program is written and should run the program.The proposed architecture does almost everything without any human interventions by the help of APIsand configuration files. However, the nodes connected to the network is identified using their IP address.As of now, the user must scan the network manually to identify the node IP address and this addressmust be updated into the configuration files. In future, an automated identification of IP addresses ofthe nodes can be added into the middleware.The proposed architecture and middleware system only handle the reconfigurations if the programs arestateless. The proposed architecture would not save the state of the device at any time of thereconfiguration. This is the last limitation of the architecture since the program would lose any runningstate after the reconfiguration happens. In future, a mechanism to save the program running state andvariables must be introduced into the system.7 REFERENCESBecari, W. and Ramirez-Fernandez, F.J., 2016, September. Electrogoniometer sensor with USBconnectivity based on the IEEE1451 standard. In 2016 IEEE International Symposium onConsumer Electronics (ISCE) (pp. 41-42). IEEE.Beloglazov, A. and Buyya, R., 2010, May. Energy efficient resource management in virtualized clouddata centers. In Proceedings of the 2010 10th IEEE/ACM international conference on cluster,cloud and grid computing (pp. 826-831). IEEE Computer Society.Boman, J., Taylor, J. and Ngu, A.H., 2014, October. Flexible IoT middleware for integration of thingsand applications. In 10th IEEE International Conference on Collaborative Computing:Networking, Applications and Worksharing (pp. 481-488). IEEE.Chi, Q., Yan, H., Zhang, C., Pang, Z. and Da Xu, L., 2014. A reconfigurable smart sensor interface forindustrial WSN in IoT environment. IEEE transactions on industrial informatics, 10(2), pp.1417-1425.
Page 11
Australasian Conference on Information SystemsToms & Lee2019, Perth Western AustraliaReconfiguration of IoT Nodes394De Prado, A.G., Ortiz, G. and Boubeta-Puig, J., 2017. CARED-SOA: A context-aware event-drivenservice-oriented Architecture. IEEE Access, 5, pp.4646-4663.Hauck, S. and Borriello, G., 1997. Pin assignment for multi-FPGA systems. IEEE transactions oncomputer-aided design of integrated circuits and systems, 16(9), pp.956-964.Heo, S., Woo, S., Im, J. and Kim, D., 2015, October. IoT-MAP: IoT mashup application platform for theflexible IoT ecosystem. In 2015 5th International Conference on the Internet of Things (IOT) (pp.163-170). IEEE.Iwanicki, K., 2018, July. A distributed systems perspective on industrial IoT. In 2018 IEEE 38thInternational Conference on Distributed Computing Systems (ICDCS) (pp. 1164-1170). IEEE.Khajeh-Hosseini, A., Greenwood, D. and Sommerville, I., 2010, July. Cloud migration: A case study ofmigrating an enterprise it system to iaas. In 2010 IEEE 3rd International Conference on cloudcomputing (pp. 450-457). IEEE.Kodali, R.K. and Soratkal, S., 2016, December. MQTT based home automation system using ESP8266.In 2016 IEEE Region 10 Humanitarian Technology Conference (R10-HTC)(pp. 1-5). IEEE.Lee, K. and Coulson, G., 2006. Supporting runtime reconfiguration on network processors. Journal ofInterconnection Networks, 7(04), pp.475-492.Lee, K., Paton, N.W., Sakellariou, R., Deelman, E., Fernandes, A.A. and Mehta, G., 2009. Adaptiveworkflow processing and execution in pegasus. Concurrency and Computation: Practice andExperience, 21(16), pp.1965-1981.Lee, K., Paton, N.W., Sakellariou, R. and Fernandes, A.A., 2011. Utility functions for adaptively executingconcurrent workflows. Concurrency and Computation: Practice and Experience, 23(6), pp.646-666.Lu, C.H., Hwang, T. and Hwang, I.S., 2016, May. IoT Inventor: A web-enabled composer for buildingIoT-enabled reconfigurable agentized services. In 2016 IEEE International Conference onConsumer Electronics-Taiwan (ICCE-TW) (pp. 1-2). IEEE.Myint, C.Z., Gopal, L. and Aung, Y.L., 2017, May. Reconfigurable smart water quality monitoring systemin IoT environment. In 2017 IEEE/ACIS 16th International Conference on Computer andInformation Science (ICIS) (pp. 435-440). IEEE.Pena, M.D.V., Rodriguez-Andina, J.J. and Manic, M., 2017. The internet of things: The role ofreconfigurable platforms. IEEE Industrial Electronics Magazine, 11(3), pp.6-19.Pazos, N., Müller, M., Aeberli, M. and Ouerhani, N., 2015, December. ConnectOpen-automaticintegration of IoT devices. In 2015 IEEE 2nd World Forum on Internet of Things (WF-IoT)(pp.640-644). IEEE.Raagaard, M.L., Pop, P., Gutiérrez, M. and Steiner, W., 2017, October. Runtime reconfiguration of time-sensitive networking (TSN) schedules for fog computing. In 2017 IEEE Fog World Congress(FWC) (pp. 1-6). IEEE.Sarray, I., Ressouche, A., Gaffé, D., Tigli, J.Y. and Lavirotte, S., 2015, December. Safe composition inmiddleware for the internet of things. In Proceedings of the 2nd Workshop on Middleware forContext-Aware Applications in the IoT (pp. 7-12). ACM.Singh, M., Rajan, M.A., Shivraj, V.L. and Balamuralidhar, P., 2015, April. Secure mqtt for internet ofthings (iot). In 2015 Fifth International Conference on Communication Systems and NetworkTechnologies (pp. 746-751). IEEE.Xiaohui, X., 2013, June. Study on security problems and key technologies of the internet of things.In 2013 International conference on computational and information sciences (pp. 407-410).Zhou, Z., Du, C., Shu, L., Hancke, G., Niu, J. and Ning, H., 2015. An energy-balanced heuristic for mobilesink scheduling in hybrid WSNs. IEEE Transactions on Industrial Informatics, 12(1), pp.28-40.Copyright: © 2019 authors. This is an open-access article distributed under the terms of the CreativeCommons Attribution-NonCommercial 3.0 Australia License, which permits non-commercial use,distribution, and reproduction in any medium, provided the original author and ACIS are credited.
Page 1
Australasian Conference on Information SystemsRendell, Adam & Eidels2019, Perth Western AustraliaSetting the e-Commerce Scene478Setting the e-Commerce Scene: A Qualitative Investigationof the Use of Nature Imagery in User Interface DesignResearch in ProgressAshlea RendellSchool of Electrical Engineering and ComputingThe University of NewcastleAustraliaEmail: Ashlea.Rendell@uon.edu.auDr Marc T. P. AdamSchool of Electrical Engineering and ComputingThe University of NewcastleAustraliaEmail: Marc.Adam@newcastle.edu.auDr Ami EidelsSchool of PsychologyThe University of NewcastleAustraliaEmail: Ami.Eidels@newcastle.edu.auAbstractUser experience designers often employ natural landscapes as background imagery in e-commerce userinterfaces (UI). However, at this stage, there is only limited work on how nature imagery in UI designaffects user perception and behaviour. In this paper, we present a qualitative study involving semi-structured interviews into the use of nature imagery in UI design. Our study builds on theories inenvironmental psychology and seeks to develop a theoretical framework for the role of nature imageryin user perception and behaviour. Further, building on the expertise of user experience practitionersand end users, we seek to develop theoretically-grounded design guidelines for the capture, selection,and integration of nature imagery in e-commerce UI designs.Keywords: e-commerce, exploratory study, interviews, nature imagery, user interface design
Page 2
Australasian Conference on Information SystemsRendell, Adam & Eidels2019, Perth Western AustraliaSetting the e-Commerce Scene4791 INTRODUCTIONUser experience designers often employ natural landscapes as background imagery in e-commerce userinterfaces (UIs). For instance, a survey of 500 international university websites, a service with noinherent connection with nature (as opposed to, for example, a landscaping business), found thatapproximately 40% depicted garden imagery on their homepage (Vilnai-Yavetz and Tifferet 2013).However, despite its abundant use in UI design in practice, a recent literature review by Rendell et al.(2019) found that (1) overall there is limited research investigating the influence of nature imagery onuser perception and behaviour, and (2) there has been no exploratory qualitative investigation thatconsiders the expertise of user experience (UX) practitioners and end users themselves. This gap isconcerning because it limits the knowledge base of how nature imagery can be created, selected, andintegrated into UI designs for optimal user experience in different application areas. In this paper wepresent an in-progress qualitative investigation of the knowledge held be expert stakeholders and end-users in relation to the use of nature imagery in e-commerce UI designs.Overall, there is ample evidence of the positive impact of nature exposure on human physiology,perception, and behaviour. For instance, potted plants in retail environments positively affect customersatisfaction and service quality perceptions (Tifferet and Vilnai-Yavetz 2017). Similarly, using natureimagery in print advertising can improve consumer attitudes towards brands (Hartmann et al. 2016;Schmuck et al. 2018). In a recent meta-analysis, Twohig-Bennett and Jones (2018) reported positiveeffects of nature exposure on 11 health outcomes, including improved heart rate variability, salivarycortisol levels, and cardiovascular mortality rates. Perhaps of most interest to human-computerinteraction (HCI) researchers are nature’s beneficial impacts on attention restoration and informationprocessing (see Berto (2014) for a review). Recently, Gerber et al. (2017) showed that natural elementscan mitigate physiological stress (e.g., blood pressure, heart rate) as well as information-processingfluency (as evidenced by eye fixations), even in virtual reality (VR) environments. Further, recent workhas demonstrated that specific animal imagery can elicit phobic affective responses associated withincreased information recall (Riaz et al. 2018). Despite this evidence, most existing research primarilyfocuses on nature exposure beyond digital interfaces. As such, this research does not take into accountconsiderations unique to the digital or e-commerce environment (e.g., the alignment of nature imagerywith personalisation principles in e-commerce UI design), nor does it provide design knowledge for howto successfully integrate nature imagery into e-commerce UI designs.The present paper sets out to address this gap by conducting a qualitative investigation of nature imageryuse in e-commerce UI designs. Our study is rooted in environmental psychology and theories emergingfrom the Biophilia Hypothesis (Wilson 1984). In particular, we consider the impact of nature imageryon human affect and cognition through the lenses of Stress Reduction Theory (Ulrich 1993) andAttention Restoration Theory (Kaplan and Kaplan 1989), as well as UI aesthetic preferences via theLandscape Preference Matrix (Kaplan 1987). In this vein, we build on existing propositions from theliterature and investigate these qualitatively with design practitioners and end-users. The overall goal ofthe study is to develop a set of practical guidelines that can support researchers and practitioners insuccessfully employing nature imagery in UI designs, linking these guidelines to established theories inenvironmental psychology.2 THEORETICAL BACKGROUND ON NATURE IMAGERY ANDBIOPHILIAThe Biophilia Hypothesis by Wilson (1984) posits that humans have an innate biological connectionwith nature, and seek affinity with living things for health and vitality. Wilson argued that replacingforests and woodlands with urban jungles and concrete cities poses a threat to human survival due tothe disconnection with our natural living ecosystem. Three related theories are critical to investigatingwhy, and how nature imagery influences users in HCI. Herein we discuss these theories.2.1 Stress Reduction TheoryStress Reduction Theory (SRT) is an evolutionary psychology approach that focuses on the affectiveinfluences of nature (Ulrich 1993). According to Ulrich (1993), humans who were better able to recogniseand process specific natural cues of a healthy environment were more likely to survive during thePleistocene Epoch. The recognition of these cues triggered an automatic reduction in sympatheticnervous system activity, which is subjectively experienced as a decrease in arousal, and correspondingincrease in positive affect. Due to the survival-enhancing nature of being able to automatically recognisepositive environmental cues, Ulrich suggested that humans developed domain-specific brain modulesresponsible for processing specific natural elements without conscious thought. The elements identified
Page 3
Australasian Conference on Information SystemsRendell, Adam & Eidels2019, Perth Western AustraliaSetting the e-Commerce Scene480by Ulrich which trigger this process are (1) clear calm water (as it provides a hydration source), (2)healthy verdant vegetation (as it provides sustenance, both directly and indirectly), as well as (3)landscape openness, herein referred to as vantage (it affords visibility of approaching threats; Ulrich1993). Research in lab and field environments has provided strong support for SRT (see a recent meta-analysis by McMahan and Estes 2015). In the lab, Ulrich (1981) found improvements in self-reportedpositive affect, as well as water specifically, and nature generally, being more beneficial in increasingfeelings of wakeful relaxation, as measured by alpha amplitudes, in comparison to urban environments.Similarly, Friedman and colleagues (2008) found that individuals who looked at a live-feed of a localpark displayed on a screen in their office reported feeling refreshed and refocused. Improvements inheart rate variability have also been seen in participants viewing videos of natural environments incomparison to urban environments (Laumann et al. 2003). Such varied evidence of affectiveimprovements triggered by nature exposure suggest possible benefits to user engagement via theinclusion of nature imagery within e-commerce UI designs.2.2 Attention Restoration TheoryIn contrast to SRT, Attention Restoration Theory (ART) is a cognition focused interpretation ofBiophilia (Kaplan and Kaplan 1989). It builds on the notion that the finite capacity of humaninformation processing can cause fatigued executive functioning when exposed to complexenvironments (Kaplan and Kaplan 1989). Symptoms of this fatigue can be seen in a range of behaviours(e.g., impatience, reduced performance; Berto 2014). However, according to ART, as a result of theinnate “soft fascination” humans have for nature, our attention systems are effortlessly engaged whenwe are exposed to it, thus allowing the recovery of depleted attention resources (Kaplan and Kaplan1989). It is this recovery of cognitive functioning that ART focuses on in explaining the signs and effectsof Biophilic responding. There is strong support of ART in the literature, such as improved proof-readingafter nature exposure (Hartig et al. 1991; Laumann et al. 2003), increased search-task accuracy innature-inspired virtual environments (Juliani et al. 2016), and improved information processing asevidenced by eye tracking (compared to urban imagery; Dupont et al. 2014; Wang and Sparks 2016).Further, the inclusion of nature imagery in tourism advertising has also been demonstrated to increaseviewer recall (Sparks and Wang 2014). Similar to SRT, evidence such as this suggests positive effects ofthe inclusion of nature imagery in e-commerce UI design.2.3 The Landscape Preference MatrixThe Landscape Preference Matrix is an alternative theory to explain the aesthetic preference for naturalenvironments. Unlike SRT and ART, it does not depend on the natural elements themselves (such aswater or vegetation). Kaplan (1987) hypothesised that rather than specific types of natural elements, itis the macro-level environmental characteristics (e.g., form and structure) that are important foraesthetic preference. Within the matrix, Kaplan identifies visual coherence, legibility, complexity andmystery as critical indicators of an environments’ visual aesthetic potential (Kaplan 1987). In additionto evidence from other domains, the Landscape Preference Matrix has been perhaps the most commonof the Biophilia theories translated to the IS context. The factors have been assessed in relation tonumerous contexts and outcomes, including webstore trust, satisfaction, and purchase intention (Leeand Kozar 2009; Yeh and Li 2014), the usage intention of blogs (Liao et al. 2011), and the aestheticpreference of website design more generally (Rosen and Purinton 2004). While each of these studiesfound evidence that websites scoring high in the Landscape Preference Matrix factors are likely to triggermore positive behavioural outcomes than those which do not, the analyses did not consider naturecontent specifically, nor was there consideration of stakeholder expertise.3 THEORETICAL MODEL AND RESEARCH QUESTIONDEVELOPMENTAs detailed above, there is ample evidence indicating potential benefits of including nature imagery inUI designs. While a recent theoretical framework captures the potential pathways between the use ofnature imagery in UI design (e.g., the presence of water and vegetation), user perceptions (e.g., perceivednature presence, perceived visual aesthetics), and UI outcome variables that are of central interest touser experience designers (e.g., brand attitudes, purchase intentions, see Figure 1; Rendell et al. 2019),the framework was derived from a review of existing research and hence is limited to the results, andcontexts, of those studies. As the review found, no study so far has jointly considered the pathways ofperceived nature presence and perceived visual aesthetics on UI outcome variables, and more generallythere is limited IS literature on this topic. As a result, there is no acknowledgement or investigation ofUI specific constraints and knowledge that is held by experts in this context. For example, should natureimagery be the environment a product is displayed within, or should the nature image be displayed on
Page 4
Australasian Conference on Information SystemsRendell, Adam & Eidels2019, Perth Western AustraliaSetting the e-Commerce Scene481the product being sold (e.g., on the screen of a television on a product page)? Further to this, shouldpractitioners be making an effort to personalise the nature imagery used, to reflect the users’ own nativeenvironment, or does a generic stock nature image satisfy requirements? Hence, the goal of this researchis (1) to validate and extend the pathways in the framework and (2) to develop a set of practical guidelinesthat can support researchers and practitioners in successfully employing nature imagery in UI design.Figure 1. Theoretical framework on the influence of nature imagery on UI outcomevariables, adapted from Rendell et al. (2019)3.1 Validation and Extension of the FrameworkWhile Figure 1 (Rendell et al. 2019) provides an important conceptualization of what is already knownabout the influence of nature imagery on user perception and behaviour, the review also revealed thatthere is a range of factors that have not been considered in the literature so far. First, in the UI contextthere has been no research so far on whether the native landscape of an individual may affect theirresponses to particular nature imagery. For instance, a user coming from a background rich in mountainand lake scenery may respond differently to imagery of grasslands than a user growing up in a coastalenvironment. Second, nature imagery is linked to weather conditions and time of day. Prospect-RefugeTheory (Appleton 1975) suggests that a nature scene photographed on a sunny afternoon may triggerdifferent approach-avoidance responses than the same scene photographed on a rainy morning. Third,as with every background image, UX designers have a range of options for how to embed the imageryinto the site (e.g., as the scene in which a car is driving through, or on the screen of a computer beingsold) and how to link it to the products and services on the site. All of these aspects could potentiallyplay an important role in the impact of nature imagery in different contexts and hence require furtherinvestigation. Against this backdrop, our first research question states:RQ1: How can the influence of nature imagery in e-commerce UI design be conceptualised into atheoretical framework?3.2 The Development of Design GuidelinesNature imagery is widely used by user experience designers in practice (Vilnai-Yavetz and Tifferet 2013).Though there are established guidelines to identify the psychological restoration potential of natureimagery (Thake et al. 2017), and there has been research on the application of the Landscape PreferenceMatrix to UI design, to the best of our knowledge there are currently no guidelines available to assistdesigners in capturing, selecting, and integrating nature imagery into e-commerce UI designs. This isconcerning because (1) the wide proliferation of nature imagery in practice builds on a limitedknowledge base, hence, (2) it is difficult for system designers to decide and justify how to integratenature imagery into their UIs, and (3) it is thus also difficult to discern what effect this may have onusers. Hence, our second research question centres on the development of practical, yet theoretically-grounded design guidelines for employing nature imagery in UI design.RQ2: How can system designers create, select, and integrate nature imagery in e-commerce UIdesign?4 RESEARCH METHODIn light of the scarcity of stakeholder expertise consideration in the extant literature we are conductinga qualitative study engaging with a range of stakeholders, including professional UI designers, imageryand nature experts, and private users to address these research questions. In doing so we aim to validateand extend the theoretical propositions identified in the extant literature (RQ1) and derive a set ofPsychological ConstructsUI Outcome VariablesNature ImageryWater(e.g., lake, river, ocean)Vegetation(e.g., plants, flowers, trees)Vantage / Scope(e.g., visual depth, openness)Information RecallPurchase IntentionsUser SatisfactionPerceived NaturePresencePerceived VisualAestheticsCognitive Responses(e.g., attention,cognitive load)Affective Responses(e.g., arousal, stress)Website Loyalty
Page 5
Australasian Conference on Information SystemsRendell, Adam & Eidels2019, Perth Western AustraliaSetting the e-Commerce Scene482practical guidelines for the capture, selection, and integration of nature imagery in e-commerceinterfaces (RQ2).4.1 ParticipantsBased on our review of the literature, we identified three groups of stakeholders, namely (1) nature anddesign experts (e.g. academics, book authors, consultants), (2) professional staff in organisations (e.g.user experience designers, marketing designers, developers), and (3) private users (e.g. home users,students). This enables us to capture a range of different expertise and experiences and to triangulateresponse themes (Myers and Newman 2007). As varying ‘levels’ will be recruited for groups one and two(e.g. design management as well as ‘on-the-ground’ designers), we reduce the risk of collecting a biasedsample of experiences and opinions. As data collection and analysis in qualitative research runsconcurrently (Braun and Clarke 2006), the final number of included interviews will be confirmed oncethematic saturation can be confirmed.4.2 Interview StructureBased on the theoretical background summarised in Section 2 and Section 3, we follow a theoreticallydriven deductive analysis, while also allowing for inductive themes to be identified within the specificcontext of user perceptions and behaviours toward e-commerce user interfaces. This is seen in the semi-structured nature of the interviews being conducted. The structured portion of each interview focusesfirst on the validation and possible extension of the theoretical framework shown in Figure 1. The secondpart of the interviews then focuses on best practices for the capture, selection, and integration of natureimagery in UI interface design. For designers and practitioners, this includes questions on participants’current and ideal image selection process, the purpose of using nature imagery in their UI designs, andthe potential influences of nature imagery on user perceptions and behaviours. The interviews willconclude with participants offering their suggestions on guidelines for image selection processes andpractises. With regard to the private user group, interviews will begin with general questions regardingthe participants’ experience with e-commerce user interfaces, and their beliefs regarding the influenceof nature imagery within e-commerce user interfaces. This will be followed by questions regarding theusers’ own preferences and responses toward e-commerce interfaces, including their opinions andresponses toward organisations utilising nature imagery in their UI designs.4.3 ProcedurePotential participants will be given an information statement and consent form prior to participation inthe study, with informed written consent being given prior to the commencement of each interview.Participants will be able to withdraw their data from the study at any time, even if prior consent hasbeen given. During the interview participants are allowed to ask for their recording to be stopped andedited or erased. Participants will also have the opportunity to review the transcript of their interviewand edit their contribution if they wish. Interviews shall last no longer than one hour, with the possibilityof revisiting to discuss issues or matters that require clarification during analysis. This study wasapproved by the ethics committee at the University of Newcastle, Australia.4.4 Modes of AnalysisFor this study we are using thematic analysis to consider the interview material of participants. Thematicanalysis, as described by Braun and Clarke (2006), provides theoretical flexibility and allows us to makea direct interpretation of participant motivations, meanings, and experiences. The interview recordingswill be orthographically transcribed, creating a verbatim account of each interview. Following thethematic analysis phases suggested by Braun and Clarke (2006), once each interview is conducted andtranscribed (Phase 1) we will analyse the transcripts by generating initial codes (Phase 2), search forthemes (Phase 3), review themes (Phase 4), define and name themes (Phase 5), and conclude with theproduction of a final manuscript (Phase 6). As per our research questions, the themes relate to thepathways in the theoretical framework (RQ1) and the development of practical guidelines (RQ2). Weacknowledge phases 2 through 5 are iterative, and as such the identified themes and sub-themes maydevelop or become more refined as more interviews are conducted. It is possible that identified themesmay include specific natural elements, compositions or styles, weather or contexts of imagery thatshould be included when selecting nature images for UI designs, or specific responses users experiencewhen interacting with nature-filled UI designs. The final thematic map resulting from analysis will beused as the foundation for a set of design guidelines for the identification and selection of nature imageryto be used in e-commerce UI designs.
Page 6
Australasian Conference on Information SystemsRendell, Adam & Eidels2019, Perth Western AustraliaSetting the e-Commerce Scene4835 EXPECTED CONTRIBUTIONS AND CONCLUSIONModern consumers are inundated with product and organisational choice via the Internet, making e-commerce UI designs an increasingly critical factor in organisational success. Evidence from a range ofsettings on the beneficial influence of nature and nature imagery on human perception and behavioursuggests the use of nature imagery in e-commerce UI designs could trigger positive user responses suchas reduced arousal/stress responses and improved information processing (e.g., Friedman et al. 2008;Yeh and Li 2014). Indeed, recent work has demonstrated that the inclusion of animal imagery canenhance user recall of online information (Riaz et al. 2018). However, there is a present juxtapositionbetween the abundant use of nature imagery in e-commerce designs and a lack of understanding for thecontext-specific constraints around the integration of such imagery. Design guidelines have been theoutput of research in a range of human-computer interaction contexts (e.g. mHealth applications(Noorbergen et al. 2019); gamified information systems (Liu et al. 2017); biosensor-enabled decisionsupport systems (Astor et al. 2013)), and provide a useful starting point for systems designers byfacilitating empirically-driven design decisions. Further, to the best of our knowledge this is the first ISresearch attempting to uncover the expertise of stakeholders working within this space to contribute tothe integration of domain knowledge from both academic literature and real-world experiences.The present study aims to provide the first qualitative exploration of the use of nature imagery in e-commerce designs. The inclusion of multiple stakeholder perspectives will enhance theoreticalunderstanding of the role of nature imagery in e-commerce UI designs, and allow the refinement of theexisting theoretical model on the influences of nature imagery on user perception and behaviour(Rendell et al. 2019). Knowledge obtained from stakeholder interviews will guide the development ofdesign guidelines for use in the process of selecting nature imagery for e-commerce UI designs. Theseguidelines will offer a reference point for practitioners to select nature imagery that is most suitable totheir context of use and offers a mechanism for triggering positive user responses targeted whenattempting to increase engagement.6 REFERENCESAppleton, J. 1975. The Experience of Landscape. London: Wiley.Astor, P. J., Adam, M. T. P., Jerčić, P., Schaaff, K., and Weinhardt, C. 2013. "Integrating Biosignals intoInformation Systems: A Neurois Tool for Improving Emotion Regulation.," Journal ofManagement Information Systems (30:3), pp. 247-278.Berto, R. 2014. "The Role of Nature in Coping with Psychy-Physiological Stress: A Literature Review onRestorativeness," Behavioral Sciences (4:4), pp. 394-409.Braun, V., and Clarke, V. 2006. "Using Thematic Analysis in Psychology," Qualitative Research inPsychology (3:2), pp. 77-101.Dupont, L., Antrop, M., and Van Eetvelde, V. 2014. "Eye-Tracking Analysis in Landscape PerceptionResearch: Influence of Photograph Properties and Landscape Characteristics," LandscapeResearch (39:4), pp. 417-432.Friedman, B., Freier, N. G., Kahn Jr, P. H., Lin, P., and Sodeman, R. 2008. "Office Window of theFuture?-Field-Based Analyses of a New Use of a Large Display," International Journal of HumanComputer Studies (66:6), pp. 452-465.Gerber, S. M., Jeitziner, M. M., Wyss, P., Chesham, A., Urwyler, P., Müri, R. M., Jakob, S. M., and Nef,T. 2017. "Visuo-Acoustic Stimulation That Helps You to Relax: A Virtual Reality Setup for Patientsin the Intensive Care Unit," Scientific Reports (7:1).Hartig, T., Mang, M., and Evans, G. W. 1991. "Restorative Effects of Natural Environment Experiences,"Environment and Behavior (23:1), pp. 3-26.Hartmann, P., Apaolaza, V., and Eisend, M. 2016. "Nature Imagery in Non-Green Advertising: TheEffects of Emotion, Autobiographical Memory, and Consumer’s Green Traits," Journal ofAdvertising (45:4), pp. 427-440.Juliani, A. W., Bies, A. J., Boydston, C. R., Taylor, R. P., and Sereno, M. E. 2016. "NavigationPerformance in Virtual Environments Varies with Fractal Dimension of Landscape," Journal ofEnvironmental Psychology (47), pp. 155-165.Kaplan, R., and Kaplan, S. 1989. The Experience of Nature: A Psychological Perspective. Cambridge:Cambridge University Press.
Page 7
Australasian Conference jours feries france on Information SystemsRendell, Adam & Eidels2019, Perth Western AustraliaSetting the e-Commerce Scene484Kaplan, S. 1987. "Aesthetics, Affect, and Cognition: Environmental Preference from an EvolutionaryPerspective," Environment and Behavior (19:1), pp. 3-32.Laumann, K., Gärling, T., and Stormak, K. M. 2003. "Selective Attention and Heart Rate Responses toNatural and Urban Environments," Journal of Environmental Psychology (23:2), pp. 125-134.Lee, Y., and Kozar, K. A. 2009. "Designing Usable Online Stores: A Landscape Preference Perspective,"Information and Management (46:1), pp. 31-41.Liao, Y. W., Wang, Y. S., Tang, T. I., and Tian, Y. W. 2011. "Investigating the Influence of the LandscapePreference of Blogs, User Satisfactory and Behavioral Intention," 2011 Eighth InternationalConference on Information Technology: New Generations, pp. 833-838.Liu, D., Radhika, S., and Webster, J. 2017. "Toward Meaningful Engagement: A Framework for Designand Research of Gamified Information Systems," MIS Quarterly (41:4), pp. 1011-1034.McMahan, E. A., and Estes, D. 2015. "The Effect of Contact with Natural Environments on Positive andNegative Affect: A Meta-Analysis," The Journal of Positive Psychology (10:6), pp. 507-519.Myers, M. D., and Newman, M. 2007. "The Qualitative Interview in Is Research: Examining the Craft,"Information and Organization (17), pp. 2-26.Noorbergen, T. J., Adam, M. T. P., Attia, J. R., Cornforth, D. J., and Minichiello, M. 2019. "Exploringthe Design of Mhealth Systems for Health Behaviour Change Using Mobile Biosensors,"Communications of the Association for Information Systems (44), pp. 944-981.Rendell, A., Adam, M. T. P., and Eidels, A. 2019. "Towards Understanding the Influence of NatureImagery in User Interface Design: A Review of the Literature," Hawaii International Conferenceon System Sciences, Maui, Hawaii, pp. 4795-4804.Riaz, A., Gregor, S., and Lin, A. 2018. "Biophilia and Biophobia in Website Design: Improving InternetInformation Dissemination," Information and Management (55:2), pp. 199-214.Rosen, D. E., and Purinton, E. 2004. "Website Design: Viewing the Web as a Cognitive Landscape,"Journal of Business Research (57:7), pp. 787-794.Schmuck, D., Matthes, J., Naderer, B., and Beaufort, M. 2018. "The Effects of Environmental Brand Attributesand Nature Imagery in Green Advertising," Environmental Communication (12:3), pp. 414-429.Sparks, B. A., and Wang, Y. 2014. "Natural and Built Photographic Images: Preference, Complexity, andRecall," Journal of Travel and Tourism Marketing (31:7), pp. 868-883.Thake, C. L., Bambling, M., Edirippulige, S., and Marx, E. 2017. "A Psychoevolutionary Approach toIdentifying Preferred Nature Scenes with Potential to Provide Restoration from Stress," HealthEnvironments Research and Design Journal (10:5), pp. 111-124.Tifferet, S., and Vilnai-Yavetz, I. 2017. "Phyophilia and Service Atmospherics: The Effect of IndoorPlants on Consumers," Environment and Behavior (49:7), pp. 814-844.Twohig-Bennett, C., and Jones, A. 2018. "The Health Benefits of the Great Outdoors: A SystematicReview and Meta-Analysis of Greenspace Exposure and Health Outcomes," EnvironmentalResearch (166), pp. 628-637.Ulrich, R. 1993. "Biophilia, Biophobia, and Natural Landscapes," in The Biophilia Hypothesis, S.R.Kellert and E.O. Wilson (eds.). Washington, DC: Island Press, pp. 73-137.Vilnai-Yavetz, I., and Tifferet, S. 2013. "Promoting Service Brands Via the Internet," The ServicesIndustries Journal (33:15-16), pp. 1544-1563.Wang, Y., and Sparks, B. A. 2016. "An Eye-Tracking Study of Tourism Photo Stimuli: ImageCharacteristics and Ethnicity," Journal of Travel Research (55:5), pp. 588-602.Wilson, E. O. 1984. Biophilia. Cambridge, Mass.: Harvard University Press.Yeh, Y. S., and Li, Y. M. 2014. "Design-to-Lure in the E-Shopping Environment: A Landscape PreferenceApproach," Information & Management (51:8), pp. 995-1004.AcknowledgementsThis research was supported by an Australian Government Research Training Program (RTP)Scholarship.Copyright: © 2019 Rendell, Adam & Eidels. This is an open-access article distributed under theterms of the Creative Commons Attribution-NonCommercial 3.0 Australia License, which permits non-commercial use, distribution, and reproduction in any medium, provided the original author and ACISare credited.
Page 1
Australasian Conference on Information SystemsAljohani & Chandran2019, Perth Western AustraliaAdoption of M-Health Applications180Adoption of M-Health Applications: The Saudi ArabianHealthcare PerspectivesResearch in ProgressNasser AljohaniFaculty of Engineering and Information TechnologyUniversity of Technology SydneyEmail: nasser.aljohani@student.uts.edu.auDaniel ChandranFaculty of Engineering and Information TechnologyUniversity of Technology SydneyEmail: daniel.chandran@uts.edu.auAbstractDespite the vital role that mobile applications will play in the implementation of healthcare plans in theSaudi Vision 2030, several factors may influence the process. Due to the conflict of interest, lack ofexposure, resistance to change, as well as limited technical knowledge of the apps, the Saudi Arabiansociety may inadvertently impede the government’s objectives. All the challenges could be related toindividual perceptions, technical complexities, social influence, as well as organizational reliability andpreparedness. The earlier the authorities identify the issues and respond to them, the faster it will be tosucceed in the implementation of mobile health (m-health) and subsequent attainment of the Vision2030 health goals. This study conducted a review of literature in this context. The proposed model andfactors identified will be tested to understand patients’ perceptions of m-health applications. The resultswill be beneficial to increase the adoption rates of m-health in Saudi Arabia.Keyword Saudi Arabia, adoption, m-health, applications, vision
Page 2
Australasian Conference on Information SystemsAljohani & Chandran2019, Perth Western AustraliaAdoption of M-Health Applications1811 INTRODUCTIONMobile applications will play an instrumental role in the delivery of efficient healthcare services aroundthe world. Most countries have already implemented the technology since it is associated with positiveoutcomes. For example, Yousaf et al. (2019) find that the intervention is already proving its effectivenessin helping people with dementia to undergo cognitive training, screening processes, socializing,navigating, as well as monitoring of their conditions. As for Saudi Arabia, the achievement of the Vision2030 requires the integration of such programs. The country has since adopted several healthapplications, although the prevalence has not reached the anticipated levels (AlBar and Hoque 2018).Even as the Saudi Arabian government strives for full penetration of mobile health (m-health)applications (MOH 2017), this paper argues that the adoption of such interventions would enhance thecountry's Vision 2030 health goals, but the authorities must first identify individual, technical, social, aswell as organizational challenges that may hinder its plans. To understand and validate the adoptionand acceptance of m-health applications, we conducted a review of literature in this context. Moreover,this paper explores the current state of m-health in Saudi Arabia by intensively searching for previousstudies to lead why there is low adoption of such technology by patients. Moreover, a review of literatureon m-health adoption remains poorly defined even though numerous studies on m-health that havepublished in the last few years. Such a review represents an essential discovery in the development of aresearch field that can help to achieve and assist the Saudi’s Vision 2030. Therefore, this study aims tocontribute to this growing area of research by exploring a detailed review of m-health adoption. Besides,to examine and analyse the current state of m-health and its adoption by patients living in Saudi Arabia.This paper is organized as follows: Section 2 describes problem statements and research questions.Section 3 defines the aim and objectives of the research. Section 4 presents the literature review thatexplores mobile health definitions and indicate the current state of m-health applications in SaudiArabia. Section 5 presents the findings and discussion, a conceptual research model, and all factors ofthe research model are also discussed. Lastly, the research is concluded by giving a conclusion andrecommendation for future research contribution.2 PROBLEM STATEMENTDue to continuing growing number of patients in Saudi Arabia, there is a need for advanced technologysolutions to improve and overcome patients’ requests of healthcare services. The Saudi Ministry ofHealth (MOH) in 2018 designed a new application that performs online medical consultations withpersonal physicians and to book primary health care appointments (MOH 2018). However, accordingto the 2019 half-annual report by the National Digital Transformation Unit (2019), many people in thecountry are still unaware of its importance. There are various challenges of m-health and its adoption inSaudi Arabia. This includes technological challenges, infrastructural issues, and security concerns(Albabtain et al. 2014). Apart from these problems, there are many individuals, social, andorganizational challenges. The challenges of adopting m-health raise the following questions:•What is the current status of m-health applications in Saudi Arabia?•What is the role of m-health apps in the Saudi’ 2030 Vison?•What are the factors that influence the implementation of m-health apps?3 AIMS AND OBJECTIVESThe aims of this research are:•To identify the current state of m-health applications in Saudi Arabia.•To highlight and understand factors that influence patients’ adoption and acceptance of m-health services.•To validate the hypotheses developed in the next research stages.The objectives of this research are:•To provide the knowledge needed to ensure a successful implementation of m-healthapplications by a healthcare provider.•To identify the main factors that need to be examined within Saudi patients in the next researchstages.
Page 3
Australasian Conference on Information SystemsAljohani & Chandran2019, Perth Western AustraliaAdoption of M-Health Applications182• To provide a model that connects all factors related to m-health applications adoption.4 LITERATURE REVIEW4.1 Mobile Health (M-Health)Over the past century, the marvellous spreading of mobile technologies and applications related tohealth concerns has boosted a new field of electronic health (e-health), known as mobile health (m-health) (Hoque 2016). The term m-health was invented by Prof. Robert Isteparian, who indicated thatit involves the use of mobile technologies such as smartphones, tablets, and portable digital devices toimprove health practices (Kariuki and Okanda 2017). M-health related applications have added theopportunity to save and help a wide range of targeted users with health issues. Compared to traditionalhealth services, m-health applications have a lot of opportunities over general health since they allowpatients to customize their health services (Zhang et al. 2017). M-health provides access to treatmentand prescriptions anytime, health status, and rapidly obtain patients information, thereby removinglong queues and waiting times (Crico 2018). M-health context involves the use of several mobileplatforms that used by patients for health practices, including short text message services (SMS), videoconferencing, as well as smartphone applications (apps) (O'Shea et al. 2017). Healthcare plans have beenreformed in different developing countries, such as Malaysia, Thailand, Singapore, and India, toinfluence healthcare providers to use approaches in technology that can address health problems(Hussein et al. 2017).4.2 Current State of M-Health Applications in Saudi ArabiaDespite the benefits associated with mobile-based applications, many people in the country are stillunsure of its significance. In the study that examined the penetration and usage of m-health apps inSaudi Arabia, 46% of people with mental disability proved that they had between one and two healthcareapplications on their cell phones (Atallah et al. 2018). Although the research focused on a small groupof people, its findings portray the usage of such applications as substantially low. Therefore, the stateagencies should educate the members of the public on the usefulness of the programs. Additionally, theSaudi government needs to work with healthcare practitioners to accept technology in all fields.Although the mobile-based programs have the potential of influencing every health department, someexperts have not adopted it in full. For instance, m-health apps have not been widely used in addressingcardiovascular diseases (Abid 2019). Even in the mental health area, not everyone has fully embracedtechnology. Considering the morbidity of related diseases, it is vital that the authorities encourage alldoctors as well as other primary stakeholders to sensitize patients and the public in general on theimportance of the applications (Abid 2019). The incorporation of m-health into the Saudi Vision 2030was considered and approved by the government to help in the realization of its objectives. In 2016, theSaudi government unveiled its plans to achieve various goals relating to the economy. The healthcaredepartment was mentioned as an essential contributor due to emerging opportunities. Through thevision, the ministry of health envisages reformed and restructured primary healthcare, the partnershipbetween the public and private healthcare, increased capacity and quality of related education, as wellas improved collaborations with insurance companies. The government is committed to the initiative asaffirmed by its ambitious plan to have as many as 70% of its citizens' records transferred and storeddigitally (Abid 2019).As a start point to follow the Saudi Vision 2030, the Ministry of Health (MOH) in 2017 developed itsnew smartphone applications that performs online medical consultations between personal physiciansor appropriate specialists and patients remotely (MOH 2017). Even though there is a high percentage ofmobile phone users in Saudi Arabia, the utilization of this indication by healthcare providers in bothpublic and private sectors have little attention. Currently, most of the standard procedures, includingdoctor appointments and purchase of drugs, are conducted manually. Al-Hanawi et al. (2018) argue thatsuch tedious processes contribute to the high cost of healthcare services in the country. Additionally,patients with chronic diseases have to visit health centres for treatment. In contrast, developedcountries, such as the United States, provide care using different interventions, including mobile phoneapplications. The highlighted problems will possibly be addressed through technology because peoplewill not have to visit doctors. Instead, they will contact the physicians using their mobile phone apps(Al-Hanawi et al. 2018). The integration of m-health could also facilitate efficient service delivery inhealthcare. Like other countries, Saudi Arabia is always committed to maintaining effective public healthsince the factor is integral to economic growth (Alhowaish 2014). In recent times, the country hasachieved significant goals, which will be improved under the m-health initiative. For example, unlike inthe past when a high number of people did not have proper medications, Aljuaid et al. (2016) point out
Page 4
Australasian Conference on Information SystemsAljohani & Chandran2019, Perth Western AustraliaAdoption of M-Health Applications183that the access to reliable healthcare has increased in Saudi Arabia in the last three decades. While suchimprovements are commendable, recent studies show that the sector is still facing the problem ofefficiency as evidenced by the low availability of drugs, long waiting times as well as tedious appointmentprocesses (Al-Hanawi et al. 2018). Such challenges will possibly be addressed with the help of m-healthsince clients will book appointments via phone and access drug information using the same gadgets (Al-Hanawi et al. 2018). The prevalence of m-health will be instrumental in enhancing the Vision 2030 goalof improved health education. In its 2030 targets, the government intends to collaborate with theMinistry of Education to help young people to acquire pertinent information. The Saudi residents seemto be supportive of the idea judging by the high number of people who show interest with theapplications. As previously indicated, more than 86% of doctors and 64% of other people have an appon their phones, which they use to access educative health-related information (Al Ghamdi 2018; Atallahet al. 2018). The increased knowledge will enhance the attainment of health goals on which Vision 2030is predicated.5 FINDINGS AND DISCUSSION5.1 Factors and ChallengesBased on the problem statements and results from the literature review, this section examined moreliterature on factors, including individual perceptions, technical complexity, social influences andtrends, and organizational reliability and readiness, to create hypotheses for testing.5.1.1 Individual PerceptionsPerceptions refers to the feelings of users towards a system (Aiyebelehin and Omekwu 2019). IndividualPerceptions are those that were discovered in the literature as crucial factors related to the role playedby the individuals' adoption in this context (Talukder 2012). The perceptions of the members of thepublic towards technology will play a significant role in the implementation of m-health (Dutta 2015).Like people in other nations, Saudis' most considerable concern about the use of applications pertainsto their security (Almubarak 2017). The government will have to mitigate such eventualities by insistingthat mobile app companies adhere to strict principles of ethics and governance (Al Ghamdi, 2018).Currently, most Saudi nationals support the government's initiative to integrate mobile technology intohealth matters (Atallah et al., 2018). This trend is positive, but it will only remain that way if the appsare reliable in providing accurate and up-to-date data (Atallah et al., 2018). Members of the public areoften seen as the only constituents that need sensitization and motivation to new technologies. However,group psychology may affect even the most unimaginable group in the healthcare sector, the doctors(Al-Ghamdi, 2018). The attitude may negatively influence patients and other practitioners if not solvedas quickly as possible. Therefore, the following hypothesis is posited:H1: Individual Perceptions have direct influences on users' adoption of m-health applications.5.1.2 Technical ComplexitiesThe success of m-health in Saudi Arabia will also be influenced by the exposure that people have totechnical complexities that come with mobile apps. The technical complexities are those identified asmost expected technical barriers to influence the adoption of m-health applications (Gagnon et al. 2015).On these factors, the current situation suggests that the authorities may not incur substantial cost ortime in educating the public since most of them are already exposed to technology. For example, Al-Ghamdi (2018) established that at least 86.3% of the interviewed physicians used mobile phone appswithout any challenges. Similarly, Atallah et al. (2018) confirmed that as many as 64% of therespondents could seamlessly use their medical apps. In essence, most of the residents will not have anyproblem using the applications if the relevant technology companies simplify them to match the currentprograms. However, Cajita et al. (2019) found that technical complexities such as poorly designedinterface and technology cost have a negative influence towards m-health adoption. Consequently, thefollowing hypothesizes are proposed:H2: Technical complexities have negative influences on users' adoption of m-health applications.5.1.3 Social Influences and TrendsSocial influence is defined as “the extent to which consumers perceive that important others (e.g., familyand friends) believe they should use a particular technology” (Venkatesh et al 2012). Executiveinnovation-decision created by a regularity authority will remove the optional decision to adopt aninnovation by individuals (Rogers 2003). However, some studies indicate some negative attitudes ofpart of the social towards new health technologies (Gorini et al. 2018; Hoque and Bao 2015), other
Page 5
Australasian Conference on Information SystemsAljohani & Chandran2019, Perth Western AustraliaAdoption of M-Health Applications184highlights the importance and positivity of social member towards a technology (Hoque and Sorwar2016; Tavares and Oliveira 2018). For that reason, the ministry of health should analyze and change thesocial impacts and patterns that tend to impede the penetration of m-health apps (Hoque and Bao 2015).Even though a high number of Saudis understand the usefulness of the applications, there is a cultureof boredom that occurs to some of them. As an example, 86.3% of physicians possess and even utilizemobile applications, but only 50% of them want to access the app for only one time per day (Al-Ghamdi,2018). The failure to respond to social influences and trends may hamper the plan to implement thetechnology and subsequently affect Vision 2030. Thus, the following hypothesis is proposed:H3: Social Influences and Trends have direct influences on users' adoption of m-health applications.5.1.4 Organizational Reliability and ReadinessOrganizational readiness for change is described as the degree to which individuals of an organizationbehavioural and physiological willingness to perform the organizational transformation (Touré et al2012). Despite the government’s willingness, the implementation of m-health may encounter additionalcomplications from the companies that will be used in the process. An obvious critical factor for m-health readiness is government plans and regulation (Khatun et al 2015). Introducing a new m-healthapp within an organization will affect and change work responsibilities; thus, healthcare providers andapp developers should primarily consider job responsibilities while developing the app, which reflects avital readiness of the organization (Feroz et al. 2018). The app developers should ensure that theapplications do not have extended outages. Typically, the lack of reliability is one of the challenges thataffect most app companies, with some of them failing to update the relevant information (Yousaf et al.,2019). While app developers bear the most significant responsibility, other insurance companies shouldalso be prepared to support the policy. Without organization preparedness, reliability, and consensus,the implementation process will be hindered. The responsible government entities should start workingwith the relevant stakeholders early enough to avert such a possibility (Yousaf et al., 2019). As per thediscussion, the following hypothesis is proposed:H4: Organizational Reliability and Readiness has a positive influence on users' adoption of m-healthapplications.6 PROPOSED MODELThis study has developed a model that has been prepared based on a critical analysis of the literaturerelates to the adoption and acceptance of health-related technologies. The proposed model, shown inFigure 1, structured from four different perspectives that have been identified as crucial factors in theliterature. Moreover, it contains one dependent variable: Adoption to use m-health applications, andfour independent variables: Individual Perceptions, Technical Complexities Organizational Reliabilityand readiness, and Social Influence and Trends, The principal purpose of this model is to explore andinvestigate the adoption of m-health applications by patients in Saudi Arabia to increase the adoptionrate of m-health applications and achieve the Saudi Vision 2030 in the healthcare area.Figure 1: The proposed model
Page 6
Australasian Conference on Information SystemsAljohani & Chandran2019, Perth Western AustraliaAdoption of M-Health Applications1857 CONCLUSION AND RECOMMENDATIONS FOR FUTURE RESEARCHSaudi Arabia’s Vision 2030 health-related goals will require the integration of m-health applications.However, the government needs to solve individual perceptions, social influences, technicalcomplexities, as well as organizational issues that may inhibit the process. Current problems should alsobe addressed for a successful implementation. Some of the common difficulties pertain to thepenetration of mobile phone apps and unsupportive attitudinal issues. The authorities ought toencourage the members of the public to install and use the apps to access all the pertinent information.To overcome organizational issues, the relevant ministry should employ change efficacy and changevalence. If the highlighted matters are sufficiently resolved, the Saudi Arabian government couldinevitably achieve its Vision 2030 and the National Transformation Program that seeks to change thequality, accessibility as well as the cost of healthcare. The attainment of such goals would place SaudiArabia on the same level as other developed countries and satisfy its citizens. The proposed model andfactors identified will be tested in Saudi Arabia to understand and interpret patients’ perceptions of m-health applications. The results will be beneficial to increase the adoption rates of m-health in SaudiArabia.8 REFERENCESAbid, A. 2019. "M-Health Technologies to Be Key Enablers of Saudi Digital Transformation Process:Expert." Retrieved 12 April, 2019, from http://www.arabnews.com/node/1465921/saudi-arabiaAlbabtain, A. F., AlMulhim, D. A., Yunus, F., and Househ, M. S. 2014. "The Role of Mobile Health in theDeveloping World: A Review of Current Knowledge and Future Trends," Journal of SelectedAreas in Health Informatics (4:2), pp. 10-15.AlBar, A. M., and Hoque, M. R. 2018. "Patient Acceptance of E-Health Services in Saudi Arabia: AnIntegrative Perspective," Telemedicine and e-Health (25:9), pp. 847–852.Al-Ghamdi, S. 2018. "Popularity and Impact of Using Smart Devices in Medicine: Experiences in SaudiArabia," BMC Public Health (18:1), p. 531.Al-Hanawi, M. K., Alsharqi, O., Almazrou, S., and Vaidya, K. 2018. "Healthcare Finance in the Kingdomof Saudi Arabia: A Qualitative Study of Householders’ Attitudes," Applied Health Economics andHealth Policy (16:1), pp. 55-64.Alhowaish, A. K. 2014. "Healthcare Spending and Economic Growth in Saudi Arabia: A GrangerCausality Approach," International Journal of Scientific & Engineering Research (5:1), pp. 1471-1476.Aljuaid, M., Mannan, F., Chaudhry, Z., Rawaf, S., and Majeed, A. 2016. "Quality of Care in UniversityHospitals in Saudi Arabia: A Systematic Review," BMJ Open (6:2), p. e008988.Almubarak, S. S. 2017. "Factors Influencing the Adoption of Cloud Computing by Saudi UniversityHospitals," International Journal of Advanced Computer Science and Application (8:1), pp. 41-48.Atallah, N., Khalifa, M., El Metwally, A., and Househ, M. 2018. "The Prevalence and Usage of MobileHealth Applications among Mental Health Patients in Saudi Arabia," Computer Methods andPrograms in Biomedicine (156), pp. 163-168.Cajita, M. I., Hodgson, N. A., Lam, K. W., Yoo, S., and Han, H.-R. 2018. "Facilitators of and Barriers tomHealth Adoption in Older Adults with Heart Failure," CIN: Computers, Informatics, Nursing(36:8), pp. 376–382.Chen, H., Chai, Y., Dong, L., Niu, W., and Zhang, P. 2018. "Effectiveness and Appropriateness ofmHealth Interventions for Maternal and Child Health: Systematic Review," JMIR MhealthUhealth (6:1), p. e7.Crico, C., Renzi, C., Graf, N., Buyx, A., Kondylakis, H., Koumakis, L., and Pravettoni, G. 2018. "mHealth andTelemedicine Apps: In Search of a Common Regulation," ecancerMedicalscience (12), p. 853.Dutta, A., Krishnan, D., Ramanathan, S., Roy, R., Seetharaman, P., and Veppathur Mohan, R. 2015."Emr Adoption: A User Perception Study," in: 21st Americas Conference on Information Systems(AMCIS2015). Puerto Rico: AIS.
Page 7
Australasian Conference on Information SystemsAljohani & Chandran2019, Perth Western AustraliaAdoption of M-Health Applications186Feroz, A., Kadir, M. M., and Saleem, S. 2018. "Health Systems Readiness for Adopting mHealthInterventions for Addressing Non-Communicable Diseases in Low-and Middle-IncomeCountries: A Current Debate," Global Health Action (11:1), p. 1496887.Gagnon, M.-P., Ngangue, P., Payne-Gagnon, J., and Desmartis, M. 2015. "m-Health Adoption byHealthcare Professionals: A Systematic Review," Journal of the American Medical InformaticsAssociation (23:1), pp. 212-220.Gorini, A., Mazzocco, K., Triberti, S., Sebri, V., Savioni, L., and Pravettoni, G. 2018. "A P5 Approach toM-Health: Design Suggestions for Advanced Mobile Health Technology," Frontiers in Psychology(9).Hoque, M. R. 2016. "An Empirical Study of Mhealth Adoption in a Developing Country: The ModeratingEffect of Gender Concern," BMC Medical Informatics and Decision Making (16:1), p. 51.Hoque, M., and Sorwar, G. 2015. "Factors Influencing Mhealth Acceptance among Elderly People inBangladesh," in: Australasian Conference on Information Systems. Adelaide, AustraliaHoque, M. R., and Bao, Y. 2015. "Cultural Influence on Adoption and Use of E-Health: Evidence inBangladesh," Telemedicine and e-Health (21:10), pp. 845-851.Hussein, Z., Oon, S. W., and Fikry, A. 2017. "Consumer Attitude: Does It Influencing the Intention toUse Mhealth?," Procedia Computer Science (105), pp. 340-344.Kariuki, E. G., and Okanda, P. 2017. "Adoption of M-Health and Usability Challenges in M-HealthApplications in Kenya: Case of Uzazi Poa M-Health Prototype Application," in: 2017 IEEEAFRICON. Cape Town, South Africa: IEEE, pp. 530-535.Khatun, F., Heywood, A. E., Ray, P. K., Hanifi, S. M. A., Bhuiya, A., and Liaw, S. T. 2015. "Determinantsof Readiness to Adopt Mhealth in a Rural Community of Bangladesh," International journal ofMedical Informatics (84:10), pp. 847-856.Ministry of health (MOH). 2017 "MOH Initiatives 2030" https://www.moh.gov.sa/en/Ministry/MediaCenter/News/Pages/News-2017-12-14- 007.aspx Retrieved 28 Jul. 2019.National Digital Transformation Unit. 2019 "Biannual-Report2019" https://ndu.gov.sa/images/uploads/content/BiAnnualReport-2019.pdf Retrieved 1 Aug. 2019.O'shea, C. J., McGavigan, A. D., Clark, R. A., Chew, D. P., and Ganesan, A. 2017. "Mobile Health: AnEmerging Technology with Implications for Global Internal Medicine," Internal medicine journal(47:6), pp. 616-619.Rogers, E.M. 2003. "Diffusion of innovations" (5th ed.). New York: Free Press.Talukder, M. 2012. "Factors affecting the adoption of technological innovation by individual employees:An Australian study", Procedia-Social and Behavioral Sciences (40), pp 52-57.Touré, M., Poissant, L., and Swaine, B. R. 2012. "Assessment of organizational readiness for e-health ina rehabilitation centre," Disability and rehabilitation (34:2), pp 167-173.Venkatesh, V., Thong, J.Y. and Xu, X., 2012. "Consumer acceptance and use of information technology:extending the unified theory of acceptance and use of technology." MIS Quarterly (36:1), pp 157-178.Yousaf, K., Mehmood, Z., Saba, T., Rehman, A., Munshi, A. M., Alharbey, R., and Rashid, M. 2019."Mobile-Health Applications for the Efficient Delivery of Health Care Facility to People withDementia (Pwd) and Support to Their Carers: A Survey," BioMed research international 2019.Zhang, X., Han, X., Dang, Y., Meng, F., Guo, X. and Lin, J. 2017, "User acceptance of mobile healthservices from users’ perspectives: The role of self-efficacy and response-efficacy in technologyacceptance," Informatics for Health and Social Care (42: 2), pp. 194-206.Copyright: © 2019 Aljohani and Chandran. This is an open-access article distributed under the termsof the Creative Commons Attribution-NonCommercial 3.0 Australia License, which permits non-commercial use, distribution, and reproduction in any medium, provided the original author and ACISare credited.
Page 1
Australasian Conference on Information SystemsOmar, Beydoun, Win, Shukla & Baker2019, Perth Western AustraliaManaging Type II Diabetes840Socio-Technical Perspective on Managing Type II DiabetesFull PaperAdel OmarUniversity of Technology SydneySydney, AustraliaEmail: adel.omar@student.uts.edu.auGhassan BeydounUniversity of Technology SydneySydney, AustraliaEmail: ghassan.beydoun@uts.edu.auKhin Than WinUniversity of WollongongWollongong, AustraliaEmail: win@uow.edu.auNagesh ShuklaUniversity of Technology SydneySydney, AustraliaEmail: nagesh.shukla@uts.edu.auGeorge BakerFederation UniversityBallarat, AustraliaEmail: georgekeithbaker@gmail.comAbstractSocial attributes such as education level, family history or place of residence all place a strong role in theprobability of a person developing type II diabetes later in life. The aim of this paper is to develop aknowledge system based to use social attributes to estimate the prevalence of type II diabetes in a givenarea in Australia to support public health policymaking. The focus of this paper is towards answeringthe research question How can social determinants associated with type II diabetes, be used toincrementally develop a supporting knowledge-based system (KBS)? The contribution of this paper istwo folds: 1. The problem domain is analysed and a suitable KBS development framework is chosen 2.A prototype is developed and presented. Initial results with preliminary data confirm the validity of theapproach.Keywords: Social Determinants, Type II Diabetes, Decision Support Systems, Knowledge-BasedSystems, Ripple Down Rules
Page 2
Australasian Conference on Information SystemsOmar, Beydoun, Win, Shukla & Baker2019, Perth Western AustraliaManaging Type II Diabetes8411 INTRODUCTIONThe epidemic of type II diabetes is growing globally. It affected 366 million people in 2011 (Sim et al.2017) and this is expected to increase to 552 million people by 2030 (Sim et al. 2017). Here in Australiadiabetes costs the nation an estimated $14.6 billion annually (2015a). This was a considerable rise fromthe $10.6 billion (Lee et al. 2013) in 2005. In the same timeframe, Australia had spent $161.6 billion onhealth (2016). This indicates that just under 10% of the Australian Government’s Health expenditure isspent on diabetes. The numbers above indicate a significant cost to the Australian people, as indeed theworld at large. This would indicate that there’s significant evidence to suggest that working on betterways of dealing with and managing type II diabetes would be of benefit.Social determinants are a set of conditions that a person is born into, lives within and conducts his/hernormal activities within. These include but not limited to, lifestyle, education level, culture and povertyetc. During a person’s life, some people manage to change their social determinants others are confinedwithin the one that they were born into for the duration of their natural life (2018a).Type II diabetes means that a person’s pancreas is no longer producing enough good insulin. In otherwords, the body has built a resistance to it. It’s a chronic condition that develops over years (2019).The question that the paper focusses on is “How can these social determinants be used to develop aKnowledge-Based System (KBS) incrementally”? As the research involves predictive analysis, somewidely accepted predictive techniques were examined and compared for suitability for the task at hand.As a result of the literature review and liaison with the various academic & industry experts in predictiveanalysis techniques, four available Knowledge-Based Development techniques were chosen for furtherinvestigation. They are Artificial Neural Networks (ANN), Bayesian Networks, Markov Networks andRipple Down Rules (RDRs).The expected users of the proposed decision support system (DSS) are health insurance companies, lifeinsurance companies etc., government ministers and their advisors etc., It is likely that many of thoseusers have limited IT skills. Hence, in order to maximise the DSS’s effectiveness, a socio-technicalapproach is used rather than a purely technical one. An appropriate Human-Computer Interface (HCI)must be considered in the development of the DSS. That is the communication between system andusers (Sommerville and Dewsbury 2007) is a central consideration in the design decisions of the DSS. Asocial-technical perspective also addresses the challenge of constellation users from the variousorganisations and/or departments consolidated for a more integrated care delivery (Dessers et al. 2019).A most salient consideration in our application is that the contributors of domain knowledge cannot beexpected to be able to program when they communicate their knowledge to the system. We alsoanticipate the knowledge is constantly liable for change so such users cannot be burdened with codingor programming effort when s such, the proposed KBS & DSS that the domain expert isn’t required tohave a lot of technical skills to develop and maintain the KBS required to drive the DSS.Taking the above into account, the research approach is unique in two ways. Firstly, to develop the KBSduring the knowledge acquisition phase not afterwards as is common. Secondly, the very nature ofRDRs, the selected DSS development technique forces the subject matter expert to explain anydiscrepancies in the data for similar cases leading to different conclusions.2 LITERATURE REVIEWSocial determinants have an effect on other health conditions/issues as well as diabetes Sauliune andKalediene (2015), (Smith et al. 2016) and (Hill et al. 2013a). Some research has suggested that socialdeterminants are associated with type II diabetes (Schwerdtle 2016). Whilst the literature also indicatesthat KBSs have previously been used in healthcare (Sim et al. 2017). This all supports the idea that socialdeterminants could be used to develop a KBS. The exact determinants are not yet identified limitingdata availability. The novelty of this research will be to develop the KBS incrementally. Therefore, anappropriate tool is required for the incremental development of the proposed KBS. These tools arediscussed for their suitability as the framework for the proposed KBS. However, we first review theliterature related to identifying the social determinants in the first place.2.1 Social DeterminantsHill, Nielsen & Fox (2013) propose a sociobiological cycle of type II diabetes (Figure 1). That is, socialdeterminants such as income, education, housing, and nutritious food access etc, lead to poor dietarychoices. In turn, these factors then lead to inadequate lifestyle factors such as physical activity habitsas well as primary health care access. Combined, these factors can affect a diabetic’s work and/or
Page 3
Australasian Conference on Information SystemsOmar, Beydoun, Win, Shukla & Baker2019, Perth Western AustraliaManaging Type II Diabetes842education attainment productivity (Hill et al. 2013b). Further limiting a diabetic’s ability to secure theresources required to gain suitable income streams to access better healthcare. Hence, the cyclecontinues. This process results in and contributes to adverse outcomes (Hill et al. 2013a). It may leadto the retardation in the progression rate of type II diabetes. The current interventions, covered under"Biological and Psychological Responses" in figure 1, refer to monitoring and controlling factors such asblood pressure and blood glucose levels etc. The social determinants of discussed above and indicatedin figure 1 below, are not adequately addressed in managing a chronic condition such as type II diabetes(Hill et al. 2013b). These social determinants will continue to be a barrier to improving populationhealth unless they are adequately addressed (Hill et al. 2013b).This could be addressed at a policy level as Schwerdtle (2016) suggested. Therefore, it stands to reasonthat a system that can manipulate the various data from the various sources to provide useful andaccurate information would be a handy tool for policymakers and their advisors. Hence, the purpose ofa KBS in this area. Figure 1 suggests that there’s currently a biological and psychological response to thedisease, ie. a clinical perspective. However, figure 1 also suggests that more social response is requiredfor the disease. This is in direct correlation with the current research, adding further justification. Sofar, the discussion has been establishing and utilising social determinants. This gives rise to the questionof “Once the social determinates are established and utilised to develop a KBS how is the output used?”Ramaprasad et al. (2016) discuss developing an ontological mapping of Australia’s National HealthPrograms. The health programs discussed here are mapped onto an ontology with five dimensions. Thefive dimensions are: policy-scope, policy-focus, outcomes, type of care and population served(Ramaprasad et al. 2016). Hence, the purpose of this research and the proposed development of a KBS.In fact, the Ramaprasad et al. (2016) article discusses traditional policy analysis from variousperspectives. These perspectives include policymakers, doctors and allied staff, etc. Namely beingcommunicable diseases, maternal, child health and geographical location (Ramaprasad et al. 2016).What the article doesn’t discuss is how to obtain this information and present it in a usable manner.Hence, the development of the proposed KBS would fulfil this gap.There is research suggesting that particular dietary patterns are associated with type II diabetes (Galbeteet al. 2018). (Galbete et al. 2018) conducted research on associations between Ghanaian adults’ dietarypatterns and type II diabetes. They concluded that the various diets of Ghanaian adults were inverselyassociated with type II diabetes. This indicates that dietary habits are social determinant that could beused in this research. Galbete et al. (2018) and Schwerdtle (2016) suggested that particular diets areassociated with type II diabetes. Diabetes Australia, Petersen et al. (2011), Gurka Filipp and Debor(2018) suggested a link between type II diabetes and geographic regions. These two findings could becombined by establishing the normal dietary patterns of people living in particular regions, which couldbe viewed as a composite social determinant. The determination of this social determinant could bedone with the aid of the proposed KBS will be discussed next.Figure 1. The sociobiological cycle of diabetes. Source: Hill, Nielsen & Fox (2013)Schwerdtle (2016), discusses the social determinants of health, in particular, type II diabetes. The articlediscusses reducing the risk of diabetes is about addressing the social determinates that could lead to itsonset. Schwerdtle (2016) suggests that type II diabetes could also be combated by policymakers and
Page 4
Australasian Conference on Information SystemsOmar, Beydoun, Win, Shukla & Baker2019, Perth Western AustraliaManaging Type II Diabetes843should be “addressed at a policy level”. Schwerdtle (2016) lists a few social determinants associated withtype II diabetes but discusses more the importance of addressing them. Schwerdtle (2016) does mentionhow social determinates of health are economic, social and political systems do shape the conditions ofdaily life, but not the specific ones that could be associated with type II diabetes. Schwerdtle (2016),suggests a strong emphasis on education in conjunction with strategies to promote a healthier lifestyle,incorporating more exercises and healthier food, etc. Schwerdtle (2016) also suggests that nurses arewell placed to advocate for these changes to occur.Petersen et al. (2011) discuss the potential of a geodemographic system for targeting particularneighbourhoods for health issues. The system was the London Output Area Classification (LOAC),which was then compared to six other systems from various sources, both government and commercial(Petersen et al. 2011). This classified people on the basis of where they lived in order to decide where toallocate the various healthcare resources. Evidence of the geographic aspect of type II diabetes can beseen on the Diabetes Australia website. The diabetes Australia website has a diabetes risk calculator toprovide a person with the probability of developing type II diabetes within the next 5 years. This riskcalculator works by asking 11 short questions (2015b). As part of this research, the authors conduct ashort experiment on a diabetes risk calculator. The Diabetes Australia risk calculator findings were thata person born in the Middle East has a much higher chance of developing type II diabetes than a personborn in Australia (2015b). This provides further evidence that geographic location plays a role in theonset of type II diabetes. This suggests that perhaps future research could include incorporating theKBS developed in the course of this research with a Geographic Information System (GIS). Geographiclocation has also been flagged as a potential social determinant that has a correlation to MetabolicSyndrome (MetS) in the US (Gurka et al. 2018). MetS is a collection of disorders that increase the riskof type II diabetes and other diseases (2018b). Gurka, Filipp and Debor (2018) made adjustments forthe variation in sexes, age groups and ethnicities in various locations in the US. The findings were thatthe prevalence of MetS was >= 35% in West North Central, West South Central, and East South Centralwhilst 30% in the Pacific, New England, and Mid-Atlantic areas (Gurka et al. 2018).2.2 Knowledge-Based Systems (KBS) & Decision Support Systems (DSS)A Knowledge-Based System (KBS) reduces a large body of domain knowledge down to a set of rules andfacts (Gaines and Shaw 1993). Knowledge-based systems are developed by a knowledge expert,sometimes referred to as a knowledge engineer, working closely with the domain expert to elicit domainknowledge into an appropriate structure to ensure an acceptable performance by the KBS. This is donethrough a diverse range of activities as indicated in figure 2 below (Gaines and Shaw 1993).Figure 2. The basic knowledge Engineering. Source: Gaines & Shaw (1993)The rules and facts referred to earlier, are then fed into a Decision Support System (DSS) to makeinferences about particular cases. The process in figure 2 can be broken down into the following steps(Gaines and Shaw 1993):1. The knowledge engineer works closely with the domain expert to elicit their knowledge;2. The knowledge engineer then encodes the extracted knowledge for the knowledge base;3. The shell then utilises the knowledge base to obtain inferences regarding particular cases andthese inferences are used to obtain advice on particular cases.
Page 5
Australasian Conference on Information SystemsOmar, Beydoun, Win, Shukla & Baker2019, Perth Western AustraliaManaging Type II Diabetes844The essence of a DSS is to maximise the effectiveness of decision-makers (Vogel 1985). DSSs are notdesigned to replace decision-makers, but to support their decisions (Vogel 1985). In the absence ofadequate supporting tools, the structure of large databases and content may overwhelm decision-makers (Vogel 1985). The knowledgebase provides that structure and supportive tool.The process of decision making involves 3 phases (Vogel 1985):1. Intelligence, that is searching for a condition suitable for a condition;2. Design, that is, determining a possible course of action; and3. Choosing the appropriate course of action.KBSs and DSSs have often been used in healthcare, but not in a population health managementperspective. The KBS sought here would then be used by policymakers. This would play a significantrole in reducing the above-mentioned costs and impact associated with type II diabetes along with socialimpact, both within Australia and around the globe.Clearly, there’s work being conducted regarding utilizing social determinants to address and managetype II diabetes around the globe (Hill et al. 2013a); Kalediene and Sauliune and Kalediene (2015); and(Schwerdtle 2016). In recent research, it was established that adverse pregnancy outcomes are morecommon among Aboriginal and Torres Strait Islander women than non-indigenous women (Gibson-Helm et al. 2018). It was also established that later in life expectancy of Aboriginal and Torres StraitIslander women varied from non-indigenous women due to non-communicable diseases (NCDs)(Gibson-Helm et al. 2018). Using continuous quality improvement (CQI), various social determinantswere identified as barriers to high-quality health care to Aboriginal and Torres Strait Islander women inAustralia (Gibson-Helm et al. 2018). Four primary barriers identified were smoking, alcohol,psychosocial wellbeing and nutrition. Also using CQI Gibson-Helm et al. (2018) also identified priorityevidence-practice gaps in Aboriginal and Torres Strait Islander maternal health care (Gibson-Helm etal. 2018). Various strategies were then developed to address these findings. They include upskillinghealth staff, advocating healthy food and partnering with communities on health promotion projects,etc.Other research has been conducted on social determinants associated with various health conditions.For example, research has been conducted developing a toolkit that helps health workers to ask theirpatients about their social determinates of health and consequently referring them to non-clinicalsupport resources (Naz et al. 2016). There have been conferences on discussing the various cross-cultural social determinants and how these could be used to implement healthcare KBSs. However, dueto the size constraints of this paper, these will not be discussed now. It suffices to say that they arefurther evidence of the importance of social determinants used to develop health care KBSs.However, the exact social determinants associated with type II diabetes are not fully understood. It isanticipated that some of these determinates may be revealed whilst this research is in progress. Hence,the IS tool required must be flexible enough to accommodate the incremental nature of informationacquisition. That is, the KBS will be developed iteratively. Therefore, an incremental knowledgeacquisition process is required for the proposed development of this KBS. Hence, the novel concept inthis research.Shukla et al. (2018) proposed a robust data analytical model to provide a better understanding of thefactors associated with cancer patient survivability. The system comprised of a large data set used toidentify patterns of survivability of cancer patients (Shukla et al. 2018). The work then proposedsegmentation of patient historical data could be formulated into clusters and identifying patterns withinthese clusters. The appropriate course of action could then be taken to improve the patient’ssurvivability. The system operates with little to no input from an expert (Shukla et al. 2018). In anotherhealth KBS, Sangi et al. (2015) developed a risk advisor model to predict the chances of type II diabetescomplications with respect to changes in risk factors. This was done using regression analysis andArtificial Neural Networks (ANN) (Sangi et al. 2015).3 KNOWLEDGE-BASED SYSTEMS DEVELOPMENT TECHNIQUESAs discussed by Hill, Nielsen & Fox (2013), there has been considerable biological and psychologicalintervention directed at managing type II diabetes. However, Hill, Nielsen & Fox (2013) suggests thatthere should be more of a focus on social interventions. That is, shifting the context into the publichealth domain, adding a new perspective in managing type II diabetes. As suggested by Hill, Nielsen &Fox (2013), this area requires more work. The difficulty is, that there is not a significant amount of this
Page 6
Australasian Conference on Information SystemsOmar, Beydoun, Win, Shukla & Baker2019, Perth Western AustraliaManaging Type II Diabetes845kind of data available in this domain, here in Australia. In fact, according to the NSW Ministry of Health,a lot of it does not exist (Taylor 2019).That being the case, social determinants data availability is scarce and incomplete. Therefore, decisionsare currently perhaps made on the basis of deep medical expertise. However, who are the people makingthese decisions? Doctors, researchers, healthcare professional etc. All of which rely heavily on clinicalexpertise. As suggested earlier, if the context is shifted to public health, there is a new set of decision-makers. These include health insurance companies, statisticians, other researchers, and variouspoliticians etc. However, as discussed by the NSW Ministry of Health (Taylor 2019), this kind of data isscarce and limited. As they’re not sure about the knowledge, decision-makers using this data havelimited ability to make decisions. As it’s a domain that still requires more work (Hill et al. 2013b), they’relikely to make various assumptions till validated by actual data. This may be pertinent to test the limitof the incremental development approach.Most predictive analysis methods such as Artificial Neural Networks (ANNs), Bayesian Networks andMarkov Networks, etc., require an abundance of data. They use this data to look for patterns and trendsin the data for the purpose of predicting future outcomes. As discussed earlier, such data on socialdeterminants associated with type II diabetes are scarce and limited. Therefore, what’s required is apredictive analysis method capable of incremental development due to the current absence of availabledata. Ripple Down Rules (RDRs) does have those capabilities and is well suited to the incrementaldevelopment technique. RDRs will be discussed further in the following section.4 RIPPLE DOWN RULES (RDRs)RDRs knowledge base development relies on updates based on knowledge from an expert (Galani et al.2015; Beydoun and Hoffmann 20oo). The resultant product is a collection of an interconnected set ofrules organized in a binary tree structure where every node has 2 branches, a “True” and a “False” branch(Beydoun and Hoffmann 2000; Beydoun and Hoffmann 2001). Based on expert feedback, particularrules written to satisfy cases traverse the binary tree to reach a node. The process works by selecting acase and comparing its condition to the expected condition. If “True” then the conclusion is “DefaultConclusion”. If the case returns a “False” comparison, then the last condition satisfied by a rule isreturned to the knowledge base. This process begins at the root node and is repeated until reaching aleaf node (Beydoun and Hoffmann 2013). Assuming t to indicate a leaf node required, then ‘ripplingdown’ to t guarantees that at least the default rule is satisfied, hence, returning a conclusion. The RDR’supdate policies are based upon the concept that when a knowledge-based system makes an incorrectconclusion, a new rule r is added to correct the conclusion in the same context that the incorrectconclusion was made (Beydoun and Hoffmann 2013) (see Figure 3).In an RDR KBS, rules are added only in the context of their desired application (Beydoun and Hoffmann2013). Rules are added to satisfy a case for which the original sequence of rules failed, excluding casescovered by its predecessor rule (Beydoun and Hoffmann 2013). Rules are never removed or modifiedbecause the corrected case is actually in the newest added rule (Beydoun and Hoffmann 2013). Due tothe way conditions of new rules are added, correction made by the expert is guaranteed to be valid(Beydoun and Hoffmann 2013). Should an expert disagree with a knowledge base conclusion, then theknowledge base has failed and requires modification. Hence the proposed monitoring duringdevelopment concept. This modification can be carried out directly by the expert due to the simplicityof its modification (Beydoun and Hoffmann 2013). There are two main features of RDRs whichformulate the simplicity of its modification. Firstly, that the cause of failure is automatically determineddue to the knowledge base’s tree-like structure. That is, a new rule is added to the leaf node and isattached to the last visited rule prior to the knowledge base’s failure (Beydoun and Hoffmann 2013).Secondly, the knowledge base’s framework ensures that every newly added rule is consistent with itscorresponding new case, without creating any inconsistencies with previously classified cases. That is,each new rule added is justified for a case classified by the expert (Beydoun and Hoffmann 2013).
Page 7
Australasian Conference on Information SystemsOmar, Beydoun, Win, Shukla & Baker2019, Perth Western AustraliaManaging Type II Diabetes846Figure 3. A classification RDR tree. A case to be classified starts at the root default node andripple down to a leaf node. Source: Beydoun and Hoffmann (2013)As our data required tends to come in subsets where uniform distribution is likely, ie. Assume 2 differentpeople randomly selected in any given postcode will have the same probability of having type II diabetes.Hence, there is coverage but, very limited density and volume. The coverage is provided by the expertknowledge of the individual(s) developing each RDR (see figure 4).Figure. 4 RDR Convergence: added accuracy is diminished with size. The number ofinstances that individual rules classify drops quickly as the knowledge base converges.Source: Beydoun and Hoffmann (2000, 2001)Predictive analysis techniques normally allow for discrepancies in conclusions based on datasets withoutoffering any real explanation on the reason. For example, if you had a dataset said indicated a personover 50 years of age, born in the Middle East with a family history of type 2 diabetes and the conclusionof that person is a diabetic. In the same dataset, along comes a person with the same social determinantsand the conclusion is that that person is not a diabetic. Most predictive analysis techniques would putthat down to discrepancies in the dataset and often ignore it. However, the very nature of RDRs wouldforce the subject matter expert to examine other social determinants to establish why this discrepancyoccurred and develop a new rule to account for it. That is, unlike other predictive analysis techniques,RDRs force the subject matter expert to justify discrepancies in the dataset.5 SUMMARY AND DISCUSSIONThis paper has proposed an alternative way to deal with and manage the consequences of type IIdiabetes. It discussed the Australian Government’s national expenditure and other statistics fromaround the globe on the disease. It further explores possible social as well as the financial impacts ofthe disease. This notion is supported by citing the various articles from researchers around the worldmaking similar claims based on their research. The authors propose that a Knowledge-Based System tobe used by policymakers would be a beneficial tool in better utilization of the limited available resources.This, in turn, would lead to a reduction of the above-mentioned government expenditure on type IIdiabetes. Not to mention improving other social implications of type II diabetes.
Page 8
Australasian Conference on Information SystemsOmar, Beydoun, Win, Shukla & Baker2019, Perth Western AustraliaManaging Type II Diabetes847Four common techniques are discussed for suitability as the framework for the proposed KBS. They areANNs, Bayesian Networks, Markov Network, and RDRs. Results of this review are shown in Table 1. Asdiscussed earlier, ANNs operate by comparing known variables with unknown variables in largedatasets. Bayesian Networks operate by a series of probabilities with conditions sourced through largedatasets. Similar to Bayesian Networks, Markov Networks also require large datasets and require somecomplex probability knowledge. All these techniques use various validation techniques through theirrespective large dataset. RDRs operate by traversing a decision tree during the data acquisition phase.Validation using RDRs is also conducted during the data acquisition phase.RequirementAnalysis ModelANNsBayesianNetworksMarkovNetworksRDRsEasy to use by non-IT Professional√Requires minimum IT professionalintervention for maintenance√Does not rely on large datasets√Provides validation during development√Contains built-in decision-making process√√√√Reaches accurate decisions√√√√Provide explanation for abnormalities inconclusion√Table 1. Analysis Model against RequirementRDRs is the method that most closely aligns with the situation at hand. That is, firstly, not a lot of datais available on social determinants associated with type II diabetes, some social determinants will bedetermined either through this research and/or while this research is being conducted and/or even afterthe conclusion of this research. Hence the need for incremental development of the proposed KBS.Secondly, due to time constraints, data validation must be done quickly. Incremental development usingRDRs allows data validation to be done during the knowledge acquisition process, eliminating the needto backtrack to validate acquired data. Ultimately, speeding up the development process. Finally, RDRsallow the update and maintenance of the KBS to be carried out by subject matter experts not IS experts.This is due to the lack of complex probability, clustering, and/or database knowledge required as in theother prediction methods. Leading to minimal intervention by IT expert, hence reducing operating coststo end-user.Based on the above comparison, RDRs were chosen as the framework for the KBS. This decision issupported this suggestion by reference to other researchers using RDRs in the medical domain allaround the globe, most of which are in a clinical sense. The authors propose that these concepts couldbe modified to the RDRs being utilized in the development of KBSs in a non-clinical perspective, that is,a population health perspective.Using the above-mentioned RDRs a prototype Diabetes Decision Support System (DDSS) has beenimplemented. Using SQL & C Sharp, it provides an easy to use interface for domain experts to use.Below are some screenshots of the DDSS indicating its ease of use with functionality.Figure 5. The RDR interface, providing domain users various options on their next course ofaction
Page 9
Australasian Conference on Information SystemsOmar, Beydoun, Win, Shukla & Baker2019, Perth Western AustraliaManaging Type II Diabetes848Figure 6. The rule builder interface, allowing domain users to construct rules to matchcases and their own subject matter expertiseFigure 7. A graphical output display indicating the percentage of type II diabetes patients ina geographic region by postcode.The DDSS is not yet complete. However, there are promising early signs. These signs indicate theapparent validity of the research.6 CONCLUSION AND FUTURE WORKThis paper has proposed an alternative way to deal with and manage the consequences of type IIdiabetes. The research indicates that type II diabetes has major implications on society both financiallyand otherwise. It was also established that a lot of response from a clinical perspective, but not a lotfrom a social perspective (Hill et al. 2013a). This research aims to provide a tool to fill this gap.Ripple Down Rules are proposed as the favoured method for developing a KBS to support policydevelopment for type II diabetes management. This is due to two main characteristics of RDR: Firstly,there is a lack of available data about social determinants associated with type II diabetes.proposed that in the absence of large data sets, the incremental development of a KBS wastheoretically possible using RDRs. Secondly, the incremental nature of the data acquisition process
Page 10
Australasian Conference on Information SystemsOmar, Beydoun, Win, Shukla & Baker2019, Perth Western AustraliaManaging Type II Diabetes849lends itself to the constructed knowledge about this domain. Finally, the nature of RDRs forces thedomain subject expert to justify any abnormalities in conclusions.The next phase of the research is to complete the above-mentioned porotype DDSS. It’s anticipated thatdata from various sources will be gathered and fed into the DDSS and analyse findings. Results of theDDSS will be further be analysed for validation the proposed theories discussed in this paper.7 REFERENCES2015a. "Diabetes in Australia." Retrieved 17/08/2017, 20172015b. "Diabetes Risk Calculator." Retrieved 27.04.2017, 2017, from https://www.diabetesaustralia.com.au/risk-calculator2016. "Health Expenditure Australia 2014-15," Canberra.2018a. "About Social Determinants of Health." Retrieved 2.06.2018, 2018, fromhttp://www.who.int/social_determinants/sdh_definition/en/2018b."MetabolicSyndrome."2018,fromhttps://www.betterhealth.vic.gov.au/health/conditionsandtreatments/metabolic-syndrome2019. Retrieved 19 July 2019, 2019, from https://diabetesnsw.com.auBeydoun, G., and Hoffmann, A. 20oo. "Monitoring Knowledge Acquisition, Instead of EvaluatingKnowledge Bases" Proc. of the European Knowledge Acquisition Conference (EKAW2000) Juan-les-Pins, France, LNCS 1937, Springer, 2000.Beydoun, G., and Hoffmann, A. 2001. "Theoretical basis for hierarchical incremental knowledgeacquisition" International Journal of Human Computer Studies (54:3), pp. 407-452.Beydoun, G., and Hoffmann, A. 2013. "Dynamic Evaluation of the Development Process of Knowledge-Based Information Systems" Knowledge and Information Systems (35:1), pp. 233-247.Dessers, E., Mohr, B. J., Lawer, C., and Kostadinova, T. 2019. "Designing Integrated Care at theEcosystem Level," International Journal of Integrated Care (IJIC) (19:S1), pp. 1-2.Gaines, B. R., and Shaw, M. L. G. 1993. "Eliciting Knowledge and Transferring It Effectively to a Knowledge-Based System," IEEE Transactions on Knowledge and Data Engineering (5:1), pp. 4-14.Galgani, F., Compton, P., Hoffmann, A. 2015, “Lexa: Building knowledge bases for automatic legalcitation classification”, Expert Systems with Applications (42:17-18), pp.6391-6407Galbete, C., Nicolaou, M., Meeks, K., Klipstein-Grobusch, K., De-Graft Aikins, A., Addo, J., Amoah, S.K., Smeeth, L., Owusu-Dabo, E., Spranger, J., Agyemang, C., Mockenhaupt, F. P., Beune, E.,Stronks, K., Schulze, M. B., and Danquah, I. 2018. "Dietary Patterns and Type 2 Diabetes amongGhanaian Migrants in Europe and Their Compatriots in Ghana: The Rodam Study," Nutritionand Diabetes (8:1).Gibson-Helm, M. E., Bailie, J., Matthews, V., Laycock, A. F., Boyle, J. A., and Bailie, R. S. 2018."Identifying Evidence-Practice Gaps and Strategies for Improvement in Aboriginal and TorresStrait Islander Maternal Health Care," PLoS ONE (13:2).Gurka, M. J., Filipp, S. L., and Deboer, M. D. 2018. "Geographical Variation in the Prevalence of Obesity,Metabolic Syndrome, and Diabetes among Us Adults Article," Nutrition and Diabetes (8:1).Hill, J., Nielsen, M., and Fox, M. H. 2013a. "Understanding the Social Factors That Contribute toDiabetes: A Means to Informing Health Care and Social Policies for the Chronically Ill," ThePermanente Journal (17), pp. 67 -72.Hill, J., Nielsen, M., and Fox, M. H. 2013b. "Understanding the Social Factors That Contribute toDiabetes: A Means to Informing Health Care and Social Policies for the Chronically Ill," ThePermanente journal (17:2), pp. 67-72.Lee, C., Cohort, R., and Magliano, D. 2013. "The Cost of Diabetes in Adults in Australia," Diabetes resClin), pp. 385 - 390.Naz, A., Rosenberg, E., Andersson, N., Labont�, R., Andermann, A., Bari, T. I. A., Adams, A.,Andermann, L., Awan, S., Betancourt, T., Chung, R., Denov, M., Dini, H. S., Dorman, P., Douma,D., Evans, T., Grais, R. F., Gunn, S., Iriart, J., Johnson, K., Khan, J., King, N., Laporta, M.,
Page 11
Australasian Conference on Information SystemsOmar, Beydoun, Win, Shukla & Baker2019, Perth Western AustraliaManaging Type II Diabetes850MacAulay, A., Naseer, F., Nasrullah, M., Netto, G., Rasanathan, K., Salim, L., Santana, V., Servili,C., Souley, I. B., and Thomas, J. 2016. "Health Workers Who Ask About Social Determinants ofHealth Are More Likely to Report Helping Patients: Mixed-Methods Study," Canadian FamilyPhysician (62:11), pp. e684-e693.Petersen, J., Gibin, M., Longley, P., Mateos, P., Atkinson, P., and Ashby, D. 2011. "Geodemographics asa Tool for Targeting Neighbourhoods in Public Health Campaigns," Journal of GeographicalSystems (13:2), pp. 173-192.Ramaprasad, A., Win, K. T., Syn, T., Beydoun, G., and Dawson, L. 2016. "Australia's National HealthPrograms: An Ontological Mapping,").Sangi, M., Win, K. T., Shirvani, F., Namazi-Rad, M. R., and Shukla, N. 2015. "Applying a NovelCombination of Techniques to Develop a Predictive Model for Diabetes Complications," PLoSONE (10:4).Sauliune, S., and Kalediene, R. 2015. "Health Profile of the Urban Community Members in Lithuania:Do Socio-Demographic Factors Matter?," International Journal of Epidemiology (44:suppl_1),pp. i81-i81.Schwerdtle, P. 2016. "Prevalence, Distribution and Impact," Australian Nursing and MidwiferyJournal).Shukla, N., Hagenbuchner, M., Win, K. T., and Yang, J. 2018. "Breast Cancer Data Analysis forSurvivability Studies and Prediction," Computer Methods and Programs in Biomedicine (155),pp. 199-208.Sim, L. L. W., Ban, K. H. K., Tan, T. W., Sethi2, S. K., and Loh, T. P. 2017. "Development of a ClinicalDecision SupportSystem for Diabetes Care: A Pilot Study," Plos One).Smith, C., McNaughton, D., and Meyer, S. 2016. "Client Perceptions of Group Education in theManagement of Type 2 Diabetes Mellitus in South Australia," Australian Journal of PrimaryHealth).Sommerville, I., and Dewsbury, G. 2007. "Dependable Domestic Systems Design: A Socio-TechnicalApproach," Interacting with Computers (19:4), pp. 438-456.Taylor, L. 2019. "What Kind of Social Determinants Data Is Available on Diabetes Patients?." Sydney,Australia: Epidemiology and Biostatistics | Centre for Epidemiology and Evidence, NSW Ministryof Health.Vogel, L. H. 1985. "Decision Support Systems in the Human Services: Discovering Limits to a PromisingTechnology," Computers in Human Services (1:1), pp. 67-80.AcknowledgementsThe authors would like to acknowledgement emeritus professor Paul Compton (UNSW) for hisassistance and support to Adel Omar during this research.Copyright: © 2019 Omar, Beydoun, Win, Shukla & Baker. This is an open-access article distributedunder the terms of the Creative Commons Attribution-NonCommercial 3.0 Australia License, whichpermits non-commercial use, distribution, and reproduction in any medium, provided the originalauthor and ACIS are credited.
Page 1
Australasian Conference on Information SystemsSoliman, Rinta-Kahila, & Kaikkonen2019, Perth Western AustraliaExploring Crowdsourcing Discontinuance467Why Is Your Crowd Abandoning You? ExploringCrowdsourcing Discontinuance through the Lens ofMotivation TheoryFull PaperWael SolimanFaculty of Information TechnologyUniversity of JyvaskylaJyväskylä, FinlandEmail: wael.soliman@jyu.fiTapani Rinta-KahilaUQ Business School & Australian Institute for Business and EconomicsThe University of QueenslandBrisbane, QueenslandEmail: t.rintakahila@uq.edu.auJoona KaikkonenFaculty of Information TechnologyUniversity of JyvaskylaJyväskylä, FinlandEmail: kaikkonen.joona@gmail.comAbstractA typical crowdsourcing platform connects organisations in need for workforce to individuals willing towork for a compensation. Considering that a motivated crowd constitutes a vital resource of suchplatforms, nurturing it becomes a crucial managerial consideration. Yet, little is known of why individualworkers abandon crowdsourcing platforms after long periods of usage. Therefore, we set out to explorehow crowd-workers’ motivations change during a platform’s usage lifecycle, from initial usage, tocontinued use, to its eventual abandonment. To this end, we conducted an in-depth qualitative inquiryinto a popular crowdsourcing platform in the software-testing domain. Leveraging self-determinationtheory and IS use lifecycle as sensitising devices, we interviewed crowd-workers who had adopted, used,and eventually abandoned the platform. As a result, we propose a stage model in which individuals’extrinsic and intrinsic motivations emerge and interact over time, resulting in discontinued use. Weprovide implications to both theory and practice.Keywords: crowdsourcing, IS discontinuance, motivation theory
Page 2
Australasian Conference on Information SystemsSoliman, Rinta-Kahila, & Kaikkonen2019, Perth Western AustraliaExploring Crowdsourcing Discontinuance4681 INTRODUCTIONCrowdsourcing has become one of the most recognisable examples of modern-day two-sided markets(Eisenmann et al., 2006; Rysman, 2009), whereby a platform (the agent) orchestrates the demand for,and supply of, the crowd’s under-utilised resources (Soliman 2015), be it funds (Ordanini et al. 2011),expertise (Ebner et al. 2009), or general belongings they are willing to share (Constantiou et al. 2017;Malhotra and Van Alstyne 2014). As spatiotemporal distances vanish, organisations, on the one hand,gain access to vast numbers of potential workers; and the public, on the other hand, are provided withnovel opportunities to employ themselves. To retain and nurture their indispensable human resources,crowdsourcing platforms must understand their crowd-workers, what motivates them, what disturbsthem, and above all, must treat them as co-creators of value (Grönroos 2008; Lusch et al. 2008). Strictlyspeaking, there is nothing preventing a displeased or frustrated crowd-worker from discontinuing orabandoning a crowdsourcing platform. However, it is not clear what frustrates or displeases a workeron a crowdsourcing platform. This is where information systems (IS) use discontinuance becomes atopic of interest. Whereas various phenomena around IS use have been studied quite extensively in thepast (see, e.g., Bhattacherjee 2001; Davis 1989; Venkatesh 2000), the focus has largely been on adoptionand continued use of technologies, leaving the final stage of the use life cycle only passingly addressedat best, and completely neglected at worst. Yet, IS discontinuance represents behaviour that is arguablydifferent from adoption and use continuance, governed by its own distinct mechanisms (Turel 2014).Individual users constitute a vital asset of any IS, especially when it comes to people-poweredcrowdsourcing platforms, rendering IS use discontinuance a critical strategic issue to be acknowledgedand addressed by managers (Xu et al. 2014). Accordingly, recent decades have witnessed a surge instudies tackling the IS discontinuance phenomenon (see Soliman and Rinta-Kahila 2019).Although interest in the IS discontinuance topic is growing, comprehensive understanding of individualIS discontinuance is only starting to emerge, leaving crowdsourcing discontinuance a completelyuncharted area. In our attempt to contribute to this development, we drill into the topic leveragingmotivation theory. Current research on IS discontinuance, and IS use phenomena in general, tends tofocus on explaining behavioural outcomes such as continued or discontinued use with the IS users’psychographic factors and their perceptions of the IS. However, the role of users’ motivations hasreceived less attention in the literature. Moreover, research on individual-level IS use discontinuancehas mostly investigated hedonic, leisure-time systems, such as social networking systems (Luqman etal. 2017; Maier, Laumer, Eckhardt, et al. 2015; Maier, Laumer, Weinert, et al. 2015), focusing on theirstress-inducing and addictive properties. Yet, information systems come in wide varieties, andcrowdsourcing systems in particular represent IS use that stems from a complex interplay of varioustypes of motivations, thus challenging the traditional hedonic/utilitarian distinctions (Soliman andTuunainen 2015). In this study, we view crowd-workers’ IS use behaviour as a temporal process, thatevolves over time. Therefore, our interest lies in examining how workers’ extrinsic and intrinsicmotivations evolve and interact over the course of the platform usage lifecycle, ultimately leading todiscontinued use. Self-determination theory (Deci and Ryan 1985, 2000; Ryan and Deci 2000a)provides important insights especially in a non-compulsory context where, unlike in manyorganisational-specific technologies, workers can exercise freedom of choice. The research questions tobe answered are the following: 1. How do motivational factors change during crowdsourcing serviceuse life cycle from adoption to discontinued use?; and 2. What are the demotivational factors drivingcrowd-workers to abandon the crowdsourcing platform they volitionally adopted?In order to shed light on the connection between discontinued IS use and users’ motivation, an empiricalinquiry was performed. As we strived to tap onto individual users’ lived experiences, a natural choicewas of method was to conduct a series of in-depth interviews with workers who had used acrowdsourcing platform and passed through the entire use lifecycle from adoption, through continueduse to discontinuance. This paper is organised as follows: the next chapter will discuss crowdsourcingand self-determination theory as our theoretical framework. After this, we disclose our researchapproach and methods. The last two sections are dedicated to reporting on findings and discussing theirimplications, respectively.2 THEORETICAL BACKGROUND AND RELATED WORKSelf-determination theory (SDT) is a broad socio-cognitive theoretical framework focusing mainly onunderstanding the study of human motivation, and its impact on behaviour and wellbeing (Deci andRyan 1985, 2000; Ryan and Deci 2000a). SDT distinguishes two types of motivations based on the originof motivation (i.e., a locus of causality): intrinsic motivation and extrinsic motivation. Intrinsicmotivation is “... the doing of an activity for its inherent satisfactions rather than for some separable
Page 3
Australasian Conference on Information SystemsSoliman, Rinta-Kahila, & Kaikkonen2019, Perth Western AustraliaExploring Crowdsourcing Discontinuance469consequence.” (Ryan and Deci 2000a, p. 56). It is the kind of motivation that is present when a persondoes something out of his or her pure, inherent interest, enjoy or challenge (Ryan and Deci 2000a). SDTfurther argues that certain context-specific events, for example rewards, communication and feedbackwhich promote individual’s feeling of competence, are the factors that can reinforce or increaseindividual’s motivation towards certain activity. Here, the intrinsic motivation is perceived to satisfyindividual’s inborn psychological needs of competence and autonomy (Deci and Ryan 2000).Furthermore, the experience of efficacy or competence alone does not enhance intrinsic motivation.Individuals must feel a sense of autonomy over the activity being performed. The behaviour must beexperienced as self-determined by the individual. As Ryan and Deci (2000a, p. 58) put it, “... for a highlevel of intrinsic motivation people must experience satisfaction of the needs both for competence andautonomy.”By contrast, extrinsic motivation involves some separable outcome (e.g. money, food, prizes) to beaccommodated with an activity. More precisely, an activity is performed because it can yield someinstrumental value to the person performing the activity (Ryan and Deci 2000b). Ryan and Deci (2000a)argue, that extrinsic motivation gains more ground on individuals’ lives as they grow up, since socialdemands and roles force people to engage with activities that are not intrinsically motivating. Accordingto SDT, extrinsic motivation varies in the level of autonomy. For example, an individual might performan activity because he or she wants to avoid a sanction (i.e., the instrumental value of the activity) whichoccurs, if the individual did not perform that particular activity. Or individual wants to perform certainactivity because it can help her work career, school some other aspect in the individual’s life. In the latterexample, the instrumental value of the activity is the positive outcome obtained from performing theactivity (e.g. individual’s work career is being uplifted). Both are examples of individuals acting byextrinsic motivation and to gain instrumental value (instead of performing out of pure interest towardsthe activity), but the level of autonomy varies from compliance with some external control/authorityand the latter involves individual’s personal choice (Ryan and Deci 2000a). If an individual is notmotivated at all (he or she does not have extrinsic or intrinsic motivation), the person is said to beamotivated and as a result of this, does not act at all or acts without an intent. Amotivated person isunwilling to commit certain behaviour. Thus, in this respect, an individual who is forced to do somethingthat he or she is not willing to do, would be amotivated and would act without an intent.SDT has been instrumental in shaping our understanding of crowd motivation in various crowdsourcingcontexts (Antikainen et al., 2010; Brabham, 2008, 2010; Ebner et al., 2009; Väätäjä, 2012; Zheng, Li, &Hou, 2011). A central assertion these studies build on is that crowd work is generally motivated by bothintrinsic and extrinsic motivations (Hossain, 2012). Despite its importance in explaining workers’motivation in crowdsourcing platforms, this line of research has largely ignored a critical questionunaddressed: how do changes in motivation impact the crowd’s willingness to contribute over time? Infact, motivation research emphasises the dynamic nature of motivations, and that their strengths andimpact may change over time (Pink 2009). While different motivations may co-exist over time, theirrespective strengths can lead to varying behaviours. What this means is that different motivationalfactors are expected to be responsible for initiating and driving different behaviours at different times.For instance, recent crowdsourcing research indicates that pre-adoption motivations differ significantlyfrom post-adoption motivations (Alam and Campbell 2016; Soliman and Tuunainen 2015). This newresearch direction which takes temporality at its core is discussed next.IS scholars have recently acknowledged that the IS use phenomenon is best portrayed as a lifecycle(Furneaux and Wade 2010, 2011) going through various stages. Most notably, three major stages havebeen widely recognized: the adoption stage, the continued usage stage, and the termination stage (Maieret al., 2015). These stages reflect an archetypical lifecycle process, with its phasic transitions throughinception, growing and maturing, before its eventual termination (Van de Ven 1992; Van de Ven andPoole 1995). This lifecycle process has been recently translated into what Maier et al. (2015) describe asthe user transformation model. This model describes a typical lifecycle that begins with an IS beingadopted, after which it transits to being continuously or repeatedly used, and as the process matures,usage is eventually discontinued. In the current work, we adopt this generic three-stage view to guideour research exploration into the process by which crowd workers go through the phases of adoption,continued use, until they decided to discontinue the using the platform.3 RESEARCH APPROACHConsidering the nascence of the phenomenon under investigation (i.e., IS discontinuance, in general,and crowdsourcing platforms, in particular), our endeavour is mainly premised on gaining an in-depthand subjective understanding of a particular worldview through the eyes of those who actually live inthat world. This emphasis on subjective interpretation generally steers the research direction to be
Page 4
Australasian Conference on Information SystemsSoliman, Rinta-Kahila, & Kaikkonen2019, Perth Western AustraliaExploring Crowdsourcing Discontinuance470explorative in nature (Klein and Myers 1999; Walsham 1995), rather than testing theory in thetraditional explanatory sense (Dubé and Paré 2003; Lee 1989). Thus, our interpretive work draws onsome central tenets from phenomenology and case study research. On the one hand, fromphenomenology, we adopt the notion that a phenomenological study ‘describes the meanings for severalindividuals of their lived experiences of a concept or a phenomenon’ (Creswell 2007, p. 57). As such, theemphasis here is on examining what the individuals experienced and how they experienced it (ibid, p.58). Motivations (or lack thereof) are manifestations of abstract human experience, just like anger,excitement, or insomnia, suggesting that a phenomenological approach is an appropriate perspectivefor studying individuals’ motivations to IS use over its lifecycle. On the other hand, from case studyresearch we adopt the emphasis on the examination of a phenomenon in its naturalistic context, withthe purpose of confronting theory with the empirical world (Piekkari et al. 2009; Keutel et al. 2014).3.1 Data Collection and AnalysisThis study was conducted in the context of a crowdsourcing service called uTest (www.utest.com), whichis an established, global online community of software testers with a platform for managingcrowdsourced software testing. To gain a comprehensive understanding of the context in which thephenomenon of interest occurs, we conducted an in-depth inquiry to the uTest platform before engagingin the actual data collection. This investigation gave us a good understanding on how the service works,allowing us to prepare an adequate interview protocol and to sensitise us to possible behaviouralpatterns that could emerge in this context. Moreover, in accumulating our contextual understanding,we were also able to leverage the expertise of one of this paper’s authors as he had previously beeninvolved in software testing business.Before a tester can participate in paid projects on uTest, he or she must be qualified by completing anaudition known as Sandbox arranged by the platform. This allows uTest to evaluate new starting testers(or testers who are not yet rated) and familiarize them with the service. In short, new testers are invitedto upcoming Sandbox test cycles. Once he or she gets invited, the tester is assigned with a so-calledSandbox team lead who supervises and provides all the necessary resources and guidance for the tester.Then the tester follows given Sandbox test cycle (an unpaid test cycle mimicking real, paid test cycles)instructions, submit one test case and one-to-two software bugs from a pre-defined website. New testers’performance in the Sandbox Program is linked to their profiles and thus, to some extent, dictates thetester’s ability to receive paid projects. Activity level is determined by lifetime participation level (whichis determined by quality of participation factors, e.g., number of reported bugs, number of approvedbugs), recent participation level during the previous three, six or twelve months and reliability (i.e., thetester reports test cases and bugs for projects that have a Test Cycle Agreement checked). Workers arecompensated with monetary payments and the size of the payment depends mainly on the tester’sperformance. Thus, the better the tester the higher the payment. It should be noted that from here on,the terms “information system” and “service” are used interchangeably to describe the crowdsourcingplatform that the crowd-testers on uTest are interacting with.We conducted in-depth, semi-structured interviews with former uTest crowd-workers. All intervieweeswere first contacted via LinkedIn or by email. The potential interviewees to be contacted were selectedbased on three main criteria: (1) The person had worked for uTest according to his or her LinkedInprofile; (2) The person is no longer working for uTest according to his or her LinkedIn profile; and (3)The end-date of the uTest work period is not more than three years ago. This requirement was set toensure the person still remembered the time he or she was using the crowdsourcing platform and herebywas able to reflect meaningfully on the interview questions. The sampling process led to theidentification of five such users who had used the service and later discontinued it. An interview protocolwas developed to cover the main themes we wanted to cover with each case. Altogether three temporalstages were covered: adoption stage, usage stage, and discontinuance stage. While our interviewprotocol included a list of questions for the interviewees, we refrained from posing the questions in astrict and ordered manner. Rather, we started by sensitising the interviewees’ lived experiences and afterthis allowed the interviews to flow naturally, using the interview protocol as a frame of reference. Thus,more detailed questions could be presented in varying order, some questions were added, and some leftout from one interview session to another (Myers and Newman 2007). The approach gives freedom forthe interviewer to adapt the course of the interview and to acquire deeper insights, which may benecessary when studying individuals’ motivations. The interviews were conducted using a computer-mediated communication strategy (Kazmer and Xie 2008) via Skype, because the interviewees wereglobally distributed and lived in different countries. The interviews were carried out using either instantmessaging (i.e. typing), voice call without video, or video call. The chosen communication technique(i.e., typing, voice call, or video call) was determined by the interviewees’ preferences and abilities to use
Page 5
Australasian Conference on Information SystemsSoliman, Rinta-Kahila, & Kaikkonen2019, Perth Western AustraliaExploring Crowdsourcing Discontinuance471that particular medium. Each interview was recorded and transcribed. Table 1 presents the study’sparticipants.PekkaMattiEevaKimmoTiinaGenderMaleMaleFemaleMaleFemaleAge (in their …)30s30s20s40s30sLocationS. EuropeS. AmericaCentral EuropeN. EuropeN. EuropeSoft. eng. educationNoYesNoYesYesFrequency of use(approximate)Usage (stage) duration5-7 days perweek36 months1-3 days perweek18 months4-7 days perweek12 months1-5 days perweek7 months0-3 days perweek12 monthsTable 1: Overview of study participantsThe analysis was informed by the theoretical framework discussed earlier, with its emphasis on thecapturing how different motivations interact and how they evolve across the three stages of adoption,continued use, and discontinuance. Each stage was examined closely via the lens of SDT, whichsensitised us to distinctions between intrinsic and extrinsic motivations, as we highlighted significantstatements that were found to convey interviewees’ motivations within each stage. This analysisprocedure yielded a matrix which consisted of the identified motivational factors related to each usagelife cycle stage. Eventually, every interview was then transformed into a narrative (Creswell 2007, p. 61)of interviewees’ experiences. Using the textual descriptions based on the interviews and the structuraldescription based on both the interviews and our own inquiry to the research context, we then strivedto capture essence of the phenomenon by describing the interviewees’ shared experience of motivationswithin the context of software testing crowdsourcing service. This process was informed by theexperiences of one of the authors who had been involved in software testing. The transcriptions werecomplemented by studying the recordings and typing down descriptions on how the interviewee seemedto react to our questions. Next, we report on our findings.4 FINDINGS4.1 AdoptionThe interviewees found out about uTest from various sources, such as software testing publications,through googling, from colleagues, and testing related blog sites. The reasons to try the service varied tosome extent, indicating that both intrinsic and extrinsic motivations contributed to the adoptiondecision. Specifically, the intrinsic factor of curiosity and enjoyment, together with the extrinsic factorsof financial reward and self-development were central to entering the adoption stage. Curiosity to trythe service, to see how it works, and to see what it could offer was one of the main intrinsic motivationsinfluencing the initial interest in trying out service. “... I was curious to see whether it would work.”(Kimmo) and “... I wanted to try it.” (Pekka) were the typical answers the interviewees gave when askedabout why they decided to try the service. Enjoying this type of work (software development) was equallyimportant in this stage. Interviewees expressed enjoyment as a driving factor of adopting the service bystating explicitly that they enjoyed their work and role in the platform.The prospect for financial rewards was the major driving extrinsic motivation of the service adoption.All of the interviewees stated that financial rewards played an important role when it comes to initialtrial of the service and later on adopting it, as exemplified by Pekka who said: “... Probably wouldn’thave ever started if the service did not offer any money.” Self-development and the desire to learn moreabout software development and testing after becoming exposed to it was another critical driving forceat this stage.4.2 Continued UseSimilar to the adoption stage, the continued usage stage was driven by a combination of both intrinsicand extrinsic motivational factors, although of differing nature. Specifically, in terms of intrinsicmotivation, self-actualisation and enjoyment were the main driving forces, followed by curiosity, andaltruistic factors. Self-actualisation reflected a general tendency to appreciate the sense of mastery andautonomy: “the freedom to do it anywhere and when you wanted...” (Eeva); and “... a sense ofautonomy regarding project selection.” (Matti). Altruism was also identified as a key contributing factorby some. Altruism here reflected a general sense of belonging to the community and a tendency toappreciate helping its members. For example, Pekka stated being rather active in the uTest community
Page 6
Australasian Conference on Information SystemsSoliman, Rinta-Kahila, & Kaikkonen2019, Perth Western AustraliaExploring Crowdsourcing Discontinuance472helping new members and educating old ones; while Eeva had written an instructional article for thecommunity to help other members in the testing work.In terms of extrinsic motivation, financial rewards and self-development remained as the leadingextrinsic factors. However, these were complemented with other factors, namely, non-monetarypersonal gains, and to a lesser extent social pressure and a desire for publicity. Non-monetary personalgains reflected self-oriented extrinsic motivational factors such as the possibility to add the workexperience gained during the service use to one’s CV. Some form of social pressure also contributed asan extrinsic factor to use the service. For instance, Tiina implied that the society’s expectations madeher feel that doing work via uTest is better than doing nothing at all, considering that at that time shehad no permanent job. Finally, publicity was identified as a key contributing factor to Pekka sinceworking with uTest helped him gain a reputation as one of the best testing engineers in the community.Interestingly, the findings suggest that both intrinsic and extrinsic motivations are necessary for drivingworkers to participate in the crowdsourcing platform. However, whereas these motivations remainsimilar in terms of origin or what SDT describes as perceived locus of causality (i.e., being intrinsic orextrinsic); the direction of these motivations have changed from the adoption to usage stages.Specifically, whereas all identified motivational factors at the adoption stage were self-oriented (i.e.,aimed at the self), with no indication that social motivational factors played a role at this stage; thecontinuance stage was found to be driven by a mix of both self-oriented and socially-oriented (i.e., aimedat others) motivational factors (see, Soliman and Tuunainen 2015). Next, we present the final stage:discontinuance.4.3 Discontinuance and AbandonmentUnderstanding motivations is important especially in the last stage of IS use lifecycle, considering thatuTest is a service that relies on voluntariness. The motivational factors discussed above explain thechanges in the driving forces behind the adoption and continuance stages, which may be described aspositive forces. By contrast, discontinuing/abandoning uTest was driven by a decrease in those positiveforces, combined with the emergence of negative forces. Specifically, we found that abandoning uTestwas mainly driven by the dwindling influence of enjoyment, self-development, and financial rewards,combined with the emergence of frustration towards uTest practices, feeling of injustice, and a sense ofpersonal life constraint. We discuss these next.When it comes to intrinsically oriented demotivating factors, we find that decrease of enjoyment wasone of the major demotivating factors that triggered or sparked individuals’ decision to discontinue theuse of the service. “The projects did not feel interesting anymore...” (Matti) and “... it didn’t feel like funanymore.” (Pekka) are examples of how the service use was experienced in the end of the usage life cyclefor some of the interviewees. Per interviewees’ reflections on their motivations, the service use had lostits excitement along the way since the projects or tasks did not offer any novelties. In some cases, theenjoyment was hampered by the service’s practicalities, as the interviewees got eventually fed up withcertain quirks in the service use. Thus, increased frustration with uTest’s practicalities, instructions, andoperative management were identified as key intrinsically oriented demotivating factors. Frustrationagainst practicalities and instructions stemmed from confusing and vaguely expressed instructions. Forexample, Matti lamented the “... confusing terms regarding payments...” referring them as one reasonhe had decided to quit using uTest. Further, Matti described frustration with the operative managementas follows: “... the project’s technical leader was never present”, which affected his perception of theservice’s support and consequently contributed into his discontinuance decision.Two of the interviewees’ stated that they had felt themselves being mistreated during the service use,reflecting a sense of injustice. This emerged as rather sensitive and emotionally loaded topic to theinformants, who stated having discontinued the service fairy quick after being confronted by the feeling.As elaborated by Pekka: “I was professionally frustrated with uTest because I was working hard there,had earned great ratings, received awards and was active in the community but I was always left outfrom role promotions and high value contracts”. Pekka’s testimony implies that he did not feel himselfas being valued by the service superiors. Sense of injustice, together with the lack of novelty, resulted indecreased self-actualisation. In Pekka’s situation, self-actualisation was suppressed by not providing anopportunity to develop his skills further, fulfill his talent and expertise, and feel competent over theactivity. In other words, the need for competence was not satisfied and was, in fact, undermined.On the other hand, personal life constraints were identified as the most prominent extrinsicallyoriginated demotivational factor affecting individual’s decision to quit using the service in par withlimited learning and self-development possibilities. Personal life constraints emerge through changes inone’s life situation. Here, acquiring a full-day job resulted in a lack of time for uTest use, as did a decision
Page 7
Australasian Conference on Information SystemsSoliman, Rinta-Kahila, & Kaikkonen2019, Perth Western AustraliaExploring Crowdsourcing Discontinuance473to move from one’s testing work to another environment. Moreover, the service’s learning and self-development possibilities emerged as another extrinsic demotivator: limitations encountered inprospects of learning and self-development affected individuals’ use behaviour in a negative manner: “...I felt like I achieved my top regarding learning from projects...” (Matti), “The possibility to learn newdecreased...” (Pekka).Although still perceived as a significant factor, the informants did not find financial rewards asimportant in the end stage of service use lifecycle as they had been in adoption and continuance stages.Per interviewees, the monetary compensations offered by uTest were perceived too inadequatecompared to the amount of time and work the interviewees had put on the service use: “As I put moreeffort I started expecting more [money] but that never really happened...” (Pekka) and “... if the paywas vastly different, then I might have reconsidered...” (Kimmo). Disappointing monetarycompensations decreased motivation towards using the service, especially in the absence of otherintrinsic motivators that could have compensated for the small pay, leading individuals to quit using theservice.Figure 1 summarises the findings in terms of the typical stages of the crowdsourcing platform usagelifecycle and its underlying motivations.Figure 1: Stages in the Crowdsourcing Platform’s Usage Lifecycle and the UnderlyingMotivations5 DISCUSSIONFrom a static perspective (i.e., stage-agnostic), our work corresponds well to the accumulating body ofknowledge on crowd motivation in that both intrinsic and extrinsic motivations play a critical role indriving crowd-workers to join and participate in crowdsourcing platforms (Antikainen et al., 2010;Brabham, 2008, 2010; Ebner et al., 2009; Väätäjä, 2012; Zheng, Li, & Hou, 2011). In this regard,curiosity, enjoyment, self-actualisation, and altruism were among the most salient intrinsic motivationalfactors; while money, self-development, social pressure, and publicity were among the salient extrinsicmotivational factors. Whereas previous research has reported signs of the ‘crowding out’ phenomenon(e.g., Lee et al. 2015) where controlled extrinsic motivation (such as financial compensation) wouldundermine intrinsic motivation when adopting technology; this was not the case in our study. Bycontrast, the participants in our study reported the synergic co-emergence of extrinsic and intrinsicmotivations in both adoption and continued use stages, and there was no evidence that financialcompensation had a negative impact on the workers enjoying their assigned work.When we adopt a temporal perspective (i.e., stage-orientation), however, we start to see some interestingobservations. First, throughout the stages of adoption, continuance and discontinuance, both intrinsicand extrinsic are critical. For instance, as Figure 1 illustrates, in the adoption phase, the emergence ofboth extrinsic and intrinsic motivations triggered the participants to start using the service to inspect ifit actually offers what it promises. However, at this stage, the crowd-workers were fundamentally drivenby selfish reasons. Specifically, curiosity and the possibility to gain financial rewards were the chief
Page 8
Australasian Conference on Information SystemsSoliman, Rinta-Kahila, & Kaikkonen2019, Perth Western AustraliaExploring Crowdsourcing Discontinuance474drivers, although the latter’s role was more pronounced. Gradually, over continuous use of the service,intrinsic motivations started to outweigh extrinsic ones in significance, and also new, additional intrinsicmotivations emerged, especially socially oriented ones (e.g., sense of community and social pressure).Towards the end of the usage life cycle, demotivating factors emerged, and previously motivating factorsdiminished, triggering users to discontinue using the service and abandon it. Once again, discontinuanceat this point was driven by both intrinsic and extrinsic motivations, although, just in like the adoptionstage, they were all selfish in nature. As the users’ motivations disappeared, they ultimately returned toa state of amotivation. However, amotivation after discontinuance is different from pre-use amotivationsince the former exists already before the individual’s actual use behaviour whereas post-useamotivation depends on the individual’s use behaviour. Very few studies in the domain of crowdsourcingadopt a temporal lens and aim to capture the dynamic nature of motivations and how they evolve acrossthe different usage stages (see, Alam and Campbell 2016; Soliman and Tuunainen 2015). Our workcomplements this line of research by a) noting the changing nature of motivational factors from selfishto social during the transition from adoption to usage stage, and b) noting that, just like in the adoptionstage, discontinuance stage is mainly driven self-oriented reasons, although different in substance.Overall, we enrich the current knowledge of motivations in IS use by examining the interplay of differenttypes motivations over time, showing how the motivational factors change and interact during the usagelife cycle. Our findings provide interesting perspective to the stage of continued IS use. While continueduse has traditionally been treated as a unidimensional construct (e.g., Bhattacherjee 2001), some havenoted that it in fact consists of various dimensions, such as frequency, intensity, and breadth of use(Turel 2015; Venkatesh et al. 2008). Our temporal analysis revealed that the dimensions of use are notonly closely connected to one’s intentions to continue using the IS but also to the expected outcomes ofIS use. This aspect has not received much attention in the prior literature. Expectations of IS use’soutcomes are in the core of the IS use continuance model (Bhattacherjee 2001), which posits thatconfirmed or positively disconfirmed expectations result in continued use of IS through increasedsatisfaction. In our study, the users were first relatively satisfied (i.e., expectations confirmed) with theservice (uTest), but when they increased their use frequency and intensity, they consequently revisedtheir expectations of financial and career-related rewards to a higher level to correspond the increasedwork input. However, the service did not yield the expected benefits, resulting in negativedisconfirmation and dissatisfaction. This indicates a crucial connection between varying levels of use,confirmation of expectations, discontinued use, and their joint interplay over time, which can be difficultto capture with static, factor-based models. Our temporal treatment provides a starting point developingricher ways to examine how variations in IS use cause expectations to evolve over time.Retaining individual users is a paramount issue in managing crowdsourcing platforms. Our studyindicates that lack of managerial and operative support can make users feel insecure about their workand alienate them from the service and its community. As the user does not work face-to-face withsuperiors and peers, it would be important to strive for making everyone feel as part of the communityand invest in the availability of support. Moreover, the interviewed ex-uTesters reported that they wereinterested in learning new things about software testing and earning a bit of money amidst an activitythey enjoy. However, negatively disconfirmed expectations diluted the effect these extrinsic motivatorsand resulted in the emergence of intrinsically oriented demotivators, such as feelings of injustice. In thelight of this, crowdsourcing platforms should critically evaluate their incentive systems by enablingcontinuous learning and progression on their platforms. Monetary rewards act as a tempting kicker toget users onboard but are not a lasting solution to retain them.We acknowledge that our findings come with certain limitations. Firstly, our study builds on insightsfrom five participants. Although this number falls within the range of adequate sample size forqualitative studies aiming for an in-depth understanding of subjects’ lived experiences (Creswell 2007,p. 61); we do not claim that our findings are generalisable in the statistical sense. Furthermore, wecontend that when conducting explorative qualitative research on a phenomenon that is in its nascence,it may be advisable to focus on gaining an in-depth understanding of individuals’ lived experiencesinstead of trying to strive for generalisability. Secondly, we could not include all the analyses andsupporting materials that we would have wished to include in this manuscript, mainly due to spacelimitation. However, we encourage the reader to contact the authors if any further clarifications aredesired.This study was set in a context of crowdsourced software engineering service that is specifically focusedon software testing. In this light, it would be interesting to conduct future research on a service that isless specialised in a small niche and offers wider array of crowdsourcing activities. Moreover, studyingthe evolution of motivations over the course of IS lifecycle in other IS contexts could prove fruitful. Forinstance, currently trending research on social networking systems (SNS) discontinuance could benefit
Page 9
Australasian Conference on Information SystemsSoliman, Rinta-Kahila, & Kaikkonen2019, Perth Western AustraliaExploring Crowdsourcing Discontinuance475from considering the temporal dimension of IS use together with motivations, as the main focus in thatstream has been on testing factorial models in cross-sectional settings. In this paper, we considered boththe hedonic and utilitarian aspects of crowdsourcing as both were found prominently present in thiscontext and thus neither could be ignored. However, we find that, especially in the case of SNS research,the focus has almost solely been on the hedonic and social aspects that incite excessive use and feelingsof overload. Yet, SNS, as well as various other IS with prominent hedonic character can be used forpurely utilitarian purposes too. We find that future research could benefit from a more inclusive andmultidimensional understanding of how and why individuals use their IS artefacts. Finally, we note thatthe way in which diminishing extrinsic motivations invited the emergence of new intrinsically orienteddemotivators in the stage of discontinued use is interesting and warrants further investigation.6 CONCLUSIONIn this paper, we explored the motivational journey through the lived experiences of five crowd-workerswho adopted, used, and then later on abandoned the crowdsourcing platform uTest. To complementprevious research that has charted workers’ motivations to adopt and continue using crowdsourcingplatforms, we examined the IS use lifecycle in its entirety, including the previously neglected final stage:discontinuance. Our results indicate that while both extrinsic and intrinsic motivations are presentthroughout the IS use lifecycle, their content and relative importance change over time. In the adoptionphase, the emergence of both extrinsic and intrinsic motivations triggered the interviewees to start usingthe service to inspect if it actually offers what it promises. However, at this stage, the crowd-workerswere fundamentally driven by selfish reasons. Specifically, curiosity and the possibility to gain financialrewards were the chief drivers, although the latter’s role was more pronounced. Over continuous use ofthe service, intrinsic motivations started to outweigh extrinsic ones in significance, and additionalintrinsic motivations emerged, especially socially oriented ones. Towards the end of the usage life cycle,demotivating factors emerged, and previously motivating factors diminished, triggering users todiscontinue using the service and abandon it. Once again, discontinuance at this point was driven byboth intrinsic and extrinsic motivations, although, just like in the adoption stage, they were all selfish innature.7 REFERENCESAlam, S. L., and Campbell, J. 2016. “Understanding the Temporality of Organizational Motivation forCrowdsourcing,” Scandinavian Journal of Information Systems (28:1), pp. 91–120.Antikainen, M., Mäkipää, M., and Ahonen, M. 2010. “Motivating and Supporting Collaboration in OpenInnovation,” European Journal of Innovation Management (13:1), pp. 100–119.Bhattacherjee, A. 2001. “Understanding Information Systems Continuance: An Expectation-Confirmation Model,” MIS Quarterly (25:3), pp. 351–370.Brabham, D. C. 2008. “Moving the Crowd at IStockphoto: The Composition of the Crowd andMotivations for Participation in a Crowdsourcing Application,” First Monday (13:6).Brabham, D. C. 2010. “Moving the Crowd at Threadless,” Information, Communication & Society(13:8), pp. 1122–1145.Constantiou, I., Marton, A., and Tuunainen, V. K. 2017. “Four Models of Sharing Economy Platforms,”MIS Quarterly Executive (16:4), pp. 231–251.Creswell, J. W. 2007. “Qualitative Inquiry and Research Design: Choosing among Five Approaches (2ndEd.),” SAGE Publications Inc (Vol. 114), SAGE Publications Inc.Davis, F. D. 1989. “Perceived Usefulness, Perceived Ease of Use, and User Acceptance of InformationTechnology,” MIS Quarterly (13), pp. 319–340.Deci, E., and Ryan, R. 1985. “The General Causality Orientations Scale: Self-Determination inPersonality,” Journal of Research in Personality (19), pp. 109–134.Deci, E., and Ryan, R. 2000. “The ‘What’ and ‘Why’ of Goal Pursuits: Human Needs and the Self-Determination of Behavior,” Psychological Inquiry (11:4), pp. 227–268.Dubé, L., and Paré, G. 2003. “Rigor in Information Systems Positivist Case Research: Current Practices,Trends, and Recommendations,” MIS Quarterly (27:4), pp. 597–636.
Page 10
Australasian Conference on Information SystemsSoliman, Rinta-Kahila, & Kaikkonen2019, Perth Western AustraliaExploring Crowdsourcing Discontinuance476Ebner, W., Leimeister, J. M., and Krcmar, H. 2009. “Community Engineering for Innovations: The IdeasCompetition as a Method to Nurture a Virtual Community for Innovations,” R&D Management(39:4), pp. 342–356.Eisenmann, T., Parker, G., and Van Alstyne, M. 2006. “Strategies for Two-Sided Markets,” HarvardBusiness Review (84:10), p. 12.Furneaux, B., and Wade, M. 2010. “The End of the Information System Life: A Model of ISDiscontinuance,” The DATA BASE for Advances in Information Systems (41:2), pp. 45–69.Furneaux, B., and Wade, M. 2011. “An Exploration of Organizational Level Information SystemsDiscontinuance Intentions,” MIS Quarterly (35:3), pp. 573–598.Grönroos, C. 2008. “Service Logic Revisited: Who Creates Value? And Who Co‐creates?,” EuropeanBusiness Review (20:4), pp. 298–314.Hossain, M. 2012. “Users’ Motivation to Participate in Online Crowdsourcing Platforms,” inInternational Conference on Innovation, Management and Technology Research (ICIMTR),IEEE, pp. 310–315.Kazmer, M. M., and Xie, B. 2008. “Qualitative Interviewing in Internet Studies: Playing with the Media,Playing with the Method,” Information, Communication & Society (11:2), Routledge, pp. 257–278.Keutel, M., Michalik, B., and Richter, J. 2014. “Towards Mindful Case Study Research in IS: A CriticalAnalysis of the Past Ten Years,” European Journal of Information Systems (23:3), pp. 256–272.Klein, H. K., and Myers, M. D. 1999. “A Set of Principles for Conducting and Evaluating InterpretiveField Studies in Information Systems,” MIS Quarterly (23:1), pp. 67–94.Lee, A. S. 1989. “A Scientific Methodology for MIS Case Studies,” MIS Quarterly (13:1), pp. 33–50.Lee, Y., Lee, J., and Hwang, Y. 2015. “Relating Motivation to Information and CommunicationTechnology Acceptance: Self-Determination Theory Perspective,” Computers in HumanBehavior (51), pp. 418–428.Luqman, A., Cao, X., Ali, A., Masood, A., and Yu, L. 2017. “Empirical Investigation of FacebookDiscontinues Usage Intentions Based on SOR Paradigm,” Computers in Human Behavior (70),Elsevier Ltd, pp. 544–555.Lusch, R. F., Vargo, S. L., and Wessels, G. 2008. “Toward a Conceptual Foundation for Service Science:Contributions from Service-Dominant Logic,” IBM Systems Journal (47:1), pp. 5–14.Maier, C., Laumer, S., Eckhardt, A., and Weitzel, T. 2015. “Giving Too Much Social Support: SocialOverload on Social Networking Sites,” European Journal of Information Systems (24:5), pp.447–464.Maier, C., Laumer, S., Weinert, C., and Weitzel, T. 2015. “The Effects of Technostress and SwitchingStress on Discontinued Use of Social Networking Services: A Study of Facebook Use,”Information Systems Journal (25:3), pp. 275–308.Malhotra, A., and Van Alstyne, M. 2014. “The Dark Side of the Sharing Economy … and How to LightenIt,” Communications of the ACM (57:11), pp. 24–27.Myers, M. D., and Newman, M. 2007. “The Qualitative Interview in IS Research: Examining the Craft,”Information and Organization (17:1), pp. 2–26.Ordanini, A., Miceli, L., Pizzetti, M., and Parasuraman, A. 2011. “Crowd-Funding: TransformingCustomers into Investors through Innovative Service Platforms,” Journal of ServiceManagement (22:4), pp. 443–470.Piekkari, R., Welch, C., and Paavilainen, E. 2009. “The Case Study as Disciplinary Convention: Evidencefrom International Business Journals,” Organizational Research Methods (12:3), pp. 567–589.Pink, D. H. 2009. Drive. The Surprising Truth About What Motivates Us, New York: Penguin Group(USA).Ryan, R., and Deci, E. 2000. “Intrinsic and Extrinsic Motivations: Classic Definitions and NewDirections,” Contemporary Educational Psychology (25), pp. 54–67.
Page 11
Australasian Conference on Information SystemsSoliman, Rinta-Kahila, & Kaikkonen2019, Perth Western AustraliaExploring Crowdsourcing Discontinuance477Ryan, R. M., and Deci, E. L. 2000a. “Intrinsic and Extrinsic Motivations: Classic Definitions and NewDirections,” Contemporary Educational Psychology (25), pp. 54–67.Ryan, R. M., and Deci, E. L. 2000b. “Self-Determination Theory and the Facilitation of IntrinsicMotivation, Social Development, and Well-Being,” American Psychologist (55:1).Rysman, M. 2009. “The Economics of Two-Sided Markets,” Journal of Economic Perspectives (23:3),pp. 125–143.Soliman, W. 2015. “People-Driven, ICT-Enabled Innovation: Crowdsourcing,” Helsinki: AaltoUniversity Publication Series; Doctoral Dissertations 173/2015, 2015.Soliman, W., and Rinta-Kahila, T. 2019. “Toward a Refined Conceptualization of IS Discontinuance:Reflection on the Past and a Way Forward,” Information and Management (May), Elsevier, pp.1–19.Soliman, W., and Tuunainen, V. K. 2015. “Understanding Continued Use of Crowdsourcing Systems: AnInterpretive Study,” Journal of Theoretical and Applied Electronic Commerce Research (10:1),pp. 1–18.Turel, O. 2014. “Quitting the Use of a Habituated Hedonic Information System: A Theoretical Modeland Empirical Examination of Facebook Users,” European Journal of Information Systems(24:4), Nature Publishing Group, pp. 1–16.Turel, O. 2015. “An Empirical Examination of the ‘Vicious Cycle’ of Facebook Addiction,” The Journalof Computer Information Systems (55:3), pp. 83–91.Väätäjä, H. 2012. “Readers’ Motivations to Participate in Hyperlocal News Content Creation,” in GROUP2012. ACM, Florida, pp. 309–311.Van de Ven, A. H. 1992. “Suggestions for Studying Strategy Process: A Research Note,” StrategicManagement Journal (13:Special Issue: Strategy Process: Managing Corporate Self-Renewal(Summer, 1992)), pp. 169–191.Van de Ven, A. H., and Poole, M. S. 1995. “Explaining Development and Change in Organizations,”Academy of Management Review (20:3), pp. 510–540.Venkatesh, V. 2000. Determinants of Perceived Ease of Use : Integrating Control , Intrinsic Motivation, and Emotion into the Technology Acceptance Model, (1997), pp. 342–365.Venkatesh, V., Brown, S. A., Maruping, L. M., and Bala, H. 2008. “Predicting DifferentConceptualizations of System Use: The Competing Roles of Behavioral Intention, FacilitatingConditions, and Behavioral Expectation,” MIS Quarterly (32:3), pp. 483–502.Walsham, G. 1995. “The Emergence of Interpretivism in IS Research,” Information Systems Research(6:4), pp. 376–394.Xu, Y. (Calvin), Yang, Y., Cheng, Z., and Lim, J. 2014. “Retaining and Attracting Users in SocialNetworking Services: An Empirical Investigation of Cyber Migration,” The Journal of StrategicInformation Systems (23:3), Elsevier B.V., pp. 239–253.Zheng, H., Li, D., and Hou, W. 2011. “Task Design, Motivation, and Participation in CrowdsourcingContests,” International Journal of Electronic Commerce (15:4), pp. 57–88.Copyright: © 2019 Soliman, Rinta-Kahila, & Kaikkonen. This is an open-access article distributedunder the terms of the Creative Commons Attribution-NonCommercial 3.0 Australia License, whichpermits non-commercial use, distribution, and reproduction in any medium, provided the originalauthor and ACIS are credited.
Page 1
Australasian Conference on Information SystemsKochar, Watson & Ouyang2019, Perth Western AustraliaEffect of ties on employee collaboration711Network Ties and Their Effect on Employee Collaborationin Enterprise Social Networks: A Review and ResearchAgendaResearch in ProgressShilpa KocharSchool of Information SystemsQueensland University of Technology (QUT)Brisbane, AustraliaEmail: s.kochar@qut.edu.auJason WatsonSchool of Information SystemsQueensland University of Technology (QUT)Brisbane, AustraliaEmail: ja.watson@qut.edu.auChun OuyangSchool of Information SystemsQueensland University of Technology (QUT)Brisbane, AustraliaEmail: c.ouyang@qut.edu.auAbstractIn recent years, there has been a rapid growth and widespread adoption of social media technologiesacross all industries. Despite the growing importance of enterprise social networks (ESN) in creatingsocial capital and facilitating innovation, there has been limited research in examining the role ofemployee relationships (ties) in these networks. Earlier studies have reported that the network structurecan enhance or restrict employee behaviour to a great extent and that those who implement ESN shouldconsider how network structure can support positive collaboration behaviour, rather than restrict it.Accordingly, we propose that the understanding of network ties and their effect on employees’collaboration will help in influencing the design and use of ESN in a way that it will help in buildingproductive and sustainable employee collaboration. This paper reviews the existing literature tounderstand the relationship between network ties and collaboration outcomes and proposes a futureresearch agenda.Keywords: Enterprise social network, network ties, organisational social media, employeecollaboration, collaboration outcomes
Page 2
Australasian Conference on Information SystemsKochar, Watson & Ouyang2019, Perth Western AustraliaEffect of ties on employee collaboration7121 INTRODUCTIONThere is a growing research into the adoption of social networks by organisations as well as implicationsof these networks. In this study we use the term “Enterprise Social Networks” (ESN) to refer to the socialnetworking technologies used by an organisation. Prior research has made some important contributionin explaining the performance impacts of ESN at the organisational level (Kane 2017; Leonardi et al.2013). Yet, research related to the implications of ESN at the employee level, within the context of anorganisation is not significant. To name a few, Kane (2015) and Kuegler et al. (2015) have done studiesto explore impact of ESN on employees. The links or connections that exist between the participatingactors (e.g. employees) are commonly referred to as ties in social networks (Burt 2004; Granovetter1973). Kane et al. (2014) suggested that the ESN structure can enhance or restrict employee behaviourto a great extent, so it is important to ensure that ESN structure supports employee collaboration, ratherthan restricting it, and in order to achieve this, it is necessary to obtain an understanding of therelationship between ESN structure, which is mainly formed by ties, and employees’ collaboration in anorganisation. Implications of ties on the employee collaboration in ESN is hardly researched before,therefore, this study is planning to propose a model to enhance employee collaboration through ESNties.Collaboration takes place when a group of individuals or stakeholders with an interest of achievingcommon goals engage in an interactive process using shared rules, norms and structures (Wood andGray 1991). Collaboration can be investigated at individual, team, intra-organisational or inter-organisational level (Blomqvist and Levy 2006). While inter-organisational collaboration has receivedgreat attention in several research disciplines, research providing insights into intra-organisationalcollaboration is not very significant, in particular at the employee level and this study aims to contributetowards addressing this gap.This research-in-progress investigates existing literature to understand the role of network ties onemployees’ collaboration manifested by collaboration outcomes within an organisation. Hence, thestudy aims to address the following research question: how do network ties in enterprise social networkaffect employees’ collaboration? To gain an in-depth understanding of ties and collaboration, a mixedmethod research will be conducted.The rest of this paper is structured as follows to present in order:• a literature review focusing on network ties and collaboration in an organisational setting,• research framework synthesised from literature review,• research methodology proposed to carry out required research activities,• research gaps identified, and finally concluding remarks of expected contribution of the study2 LITERATURE REVIEW2.1 Network TiesIn network structure, it is the network ties that define the nature of the relationship between individualsand therefore a foci for researchers investigating collaboration networks (Scherngell et al. 2016). Tiesare often defined as links or relations between two or more individuals or actors (Liu and Moskvina2016). Social network analysis tools like Gephi, Pajek and UCINET have the capability to enable usersto visualise network structure underpinned by ties and to analyse social networks.Not all ties in a network are similar and they tend to have different effect on actors in the network. Intheir seminal study of social networks, Borgatti et al. (2009) categorised tie types in social networksas similarities, relationships, interactions and flows (see Figure 1). This study will be focussingon analysing these tie types, namely similarities, relationships, interactions and flows among employeesin an intra-organisational social network.Ties between actors in a network can be measured in terms of strength. Strong ties refer to the tiesbetween nodes or individuals that are close to each other in a network (i.e. distance) and interact often(i.e. frequency), while weak ties refer to ties between nodes that are loosely connected to each othervia indirect ties and they don’t interact frequently; although weak ties are still helpful in reaching morepeople and gaining access to diverse knowledge. (Granovetter 1973).
Page 3
Australasian Conference on Information SystemsKochar, Watson & Ouyang2019, Perth Western AustraliaEffect of ties on employee collaboration713Figure 1. Typology of ties (Taken from (Borgatti et al. 2009))Network ties can also be evaluated as being positive or negative. Individuals or nodes having a positivetie share positive feelings for each other, and individuals having a negative tie share negative feelingstowards one another (Marineau et al. 2016). Negative ties are predominantly linked with outcomes suchas lower individual performance, decreased satisfaction, and conflicts, however, a recent study suggeststhat indirect negative ties can be beneficial for a person’s job performance (Marineau et al. 2016). Whileresearch has been done to address the positive ties and their effects, negative ties have often beenignored.2.2 Collaboration OutcomesIn the context of ESN, collaboration can be defined as “facilitation of the co-creation of a particular,defined outcome (solution, product or service). That is, social media used in this way are notprimarily facilitating communication, but primarily facilitating action and work” (Schlagwein and Hu2017, p. 201). Many studies have been done to quantify collaboration success or the outcomes ofcollaboration. Lee et al. (2015) classified collaboration outcomes into subjective and objectivedimensions. Objective dimension is used to quantify direct and tangible outcomes such as creating newproducts or services or process improvements; subjective dimension relates to indirect and intangibleoutcome of collaboration e.g. financial and non-financial satisfaction experienced by individualsinvolved in the collaboration process (Lee et al. 2015). In this study we focus on employee collaborationand thus will use the following collaboration outcomes identified in literature for further research:knowledge creation, competitive advantage, trust, innovation, satisfaction and social capital(Blomqvist and Levy 2006; Claiborne and Lawson 2005; Lee et al. 2015).Collaboration in online communities often leads to “sharing, transfer, accumulation, transformation,and co-creation of knowledge”, a concept also known as knowledge collaboration (Faraj et al. 2011, p.1224). Managers often find it challenging to promote an environment for knowledge creation that isconsistent with the collaboration objectives of organisation (Inkpen 1996). Besides creating knowledge,employees in an organisation are believed to be a source of competitive advantage and sustainedcompetitive advantage is achieved by the effective collaboration between employees (Wright et al. 1994).Competitive advantage is defined as “valuable, inimitable, rare, and non-substitutable capabilitiesthat help organisations outperform their competitors” (Allred et al. 2011, p. 129). When employeescollaborate, they develop a capability to understand each other and to accept negotiations i.e. mutualtrust is built among the employees, which has tremendous effect on collaboration performance(Hurmelinna et al. 2005). Innovation is often a result of information-intensive activity comprising ofinformation collection and information processing (Ahuja 2000). Collaboration networks present anincreased opportunity for innovation by providing access to various information in the network. Theliterature suggests that active employee collaboration leads to positive attitudes toward innovation, aswell as employee satisfaction (Santos-Vijande et al. 2016). Collaboration success is often measured interms of goal achievement and satisfaction is a widely accepted indicator of goal achievement(Hurmelinna et al. 2005). Social capital is also derived from effective collaboration (Riemer et al.2015). “The actual content that members exchange creates a web of cooperative relationships thatbreed norms, trust, common purpose, and coordination—that is, social capital” (Kane et al. 2014, p.277). ESN helps in creating social capital by improving connections and interactions among employees(Riemer et al. 2015).3 PROPOSED RESEARCH FRAMEWORKTo answer the research question, this study adopts a ‘collaboration outcome’ perspective to identify andconceptualise relationship between network ties and collaboration in an intra-organisational socialnetwork. A research framework has been developed taking into consideration the Borgatti’s typology ofties, namely similarities, relationships, interactions and flows (discussed in section 2.1 earlier). Aliterature review was conducted to explore the effects of ties on collaboration outcomes. It was found
Page 4
Australasian Conference on Information SystemsKochar, Watson & Ouyang2019, Perth Western AustraliaEffect of ties on employee collaboration714that despite the relationship between network ties and collaboration outcomes not being fully exploredat the intra-organisational level studies retrieved from alternate contexts do indicate a strong connectionbetween ties and collaboration outcomes. These relationships form the conceptualisation and start pointfor this research and are presented in a discussion of tie types and characteristics below.Similarities: Findings from existing inter-organisational research confirmed that too little similaritybetween nodes can decrease inter-organisational knowledge exchange, and too much similarity in anyform can impede innovative performance and can create lock-in (Lazzeretti and Capone 2016). Also,actors with diverse ties have more chance to collaborate with people from different communities anddevelop more competitive advantage (Shi et al. 2011).Relationships: Established online relationships between employees help in building trust andcontributes to positive outcomes of the project like knowledge sharing, commitment, team satisfaction,and team performance (Buvik and Rolfsen 2015). Theoretical and empirical evidence confirms thatperceived quality of relations affects the association between relationship type and well-being, therebyaffecting overall satisfaction (Fiori et al. 2006).Interactions: A recent study shows that social interaction ties are a significant predictor of satisfactionin virtual communities and active interaction in these communities leads to higher collaborativeinnovation (Xue et al. 2018). In the context of intra-organisational networks, it is not sufficient to knowwho interacts with whom, rather it is important to understand the effectiveness with which a groupexchanges information to ensure that these interactions will lead to innovation (Cross et al. 2002).Flows: Also known as tie content, flows refer to what passes through the nodes when they interact(Borgatti and Halgin 2011). Tie content (ideas, norms and beliefs) play an important role in determiningthe social capital manifested in a social network (Adler and Kwon 2002). It can also be seen that withthe increased affection in online community and with an increase in information value, the satisfactionin online community also increases (Yang et al. 2016). Cross et al. (2002) did a social network analysisof an organisation and revealed that in typical social networks, expertise of employees is not properlycaptured and some employees seem to have occupied positions such that they create bottlenecks insharing information. Such networks need interventions to realise optimal collaboration outcomes.Figure 2 Conceptual framework for the studyTie Strength and evaluation: It is perceived that the relationship between network ties andcollaboration outcome is also moderated by tie strength and its evaluation i.e. whether a tie inconsideration being strong, weak, positive or negative will also have an impact on collaboration
Page 5
Australasian Conference on Information SystemsKochar, Watson & Ouyang2019, Perth Western AustraliaEffect of ties on employee collaboration715outcome. For example, in the context of similarity ties, strong positive ties are more likely to be formedbetween highly similar nodes due to network homophily. One the one hand, this indicates that peopletend to interact more with whom they are similar with or feel more comfortable with. On the other hand,such strong positive ties may form a close network around a person thus leading to restricted knowledgeof the outside world beyond his/her network (Granovetter 1973). As actors with strong ties are often inconstant touch, the information they share might become redundant after a while and will not becontributing significantly to social capital of a network, whereas weak ties (e.g. acquaintances) have apotential to contribute novel information to the network, and can possibly lead to innovation(Granovetter 1973). Similarly, Negative ties have often been associated with stress, reducedperformance, lack of trust and decrease in performance (Marineau et al. 2016), while positive ties areconsidered to be highly beneficial as they increase the trust in the perceived value of knowledge and it’ssource, enabling positive collaboration outcomes (Levin and Cross 2004).Based on the above findings from literature, this study develops the conceptual framework shown infigure 2. The framework proposes relationship between network ties and collaboration outcomes in thecontext of ESN, and this relationship will be studied in detail in the next phase of the research.Knowledge about how network ties impact collaboration outcomes will not only help contribute to ESNtheory but also assist stakeholders in leveraging on the use of enterprise social networks to facilitate,support, and strengthen employee collaboration (GC Kane, 2014).4 RESEARCH METHODOLOGYThe research will be conducted using case study research method. An organisation using ESN for intra-organisational collaboration (i.e. within the organisation will be selected and the research participantswill be actors (employees at managerial and executive level) participating in these networks. This studywill use mixed-method approach, in which both qualitative and quantitative research methods will beused. “Mixed methods designs create special opportunities for improving data quality, therebyincreasing the significance of results” (Hollstein 2014)Social network data will be visualised and analysed using one of the network visualisation tools such asR, Gephi, UCINET etc. The tool will be selected based on its compatibility with the case organisations’social network. Several measurements will be employed to evaluate network properties. At theindividual level, degree, centrality, closeness, and betweenness will be used to identify actors of interestand their interaction patterns. We are particularly interested in identifying actors representing differenttypes of ties in the network, including those located at the periphery of the network via weak ties.Additionally, other measures like density, centralization, and size will be used to describe the wholenetwork. If needed, the network data may undergo further investigation using statistical analysis.Qualitative and descriptive data will be gathered by conducting interviews and performance ratings foremployees will also be collected. With respect to network research, qualitative methods are mostappropriate for investigating network practices and network perceptions and interpretations (Hollstein2014). This data will provide better understanding of the actors involved in the network and will help todraw relations between their virtual workplace ties and resulting collaboration outcomes. Qualitativedata will be analysed using Nvivo, a popular qualitative data analysis computer software package. NVivosoftware helps in textual data analysis and theory construction (Walsh 2003).5 RESEARCH GAPSA review of existing literature led to the identification of following limitations:• A wealth of research has explored how ties affect interactions and behaviours of individuals in asocial network (Borgatti et al. 2009; Faraj et al. 2015; Marineau et al. 2016). However, in the contextof ESN, the effects of ties on employees’ collaboration needs to be explored in depth.• The extant literature essentially reflects one way relationship between ties and collaboration. Webelieve that it’s a two-way relationship and studying the effect of collaboration on ties can be apotential research area for ESN studies.• “Most social network research in organizations has focused exclusively on the structure of ties,without regard to the content or flows which make their way through those ties” (Marineau et al.2016, p. 245). These studies provide a partial view of social network and a holistic approach isneeded that will reveal what is truly happening within the network.• What positive effects negative ties may have on collaboration outcomes is still an open question,and more research is warranted to study negative ties to determine their full potential.
Page 6
Australasian Conference on Information SystemsKochar, Watson & Ouyang2019, Perth Western AustraliaEffect of ties on employee collaboration7166 CONCLUSION AND FUTURE DIRECTIONSAfter an initial study of the literature, several research gaps have been identified in understandingnetwork ties and their association with collaboration at intra-organisational ESN level. Due to thelimitation of existing research findings in the identified area, further study is needed to answer ourresearch question. In the next phase of this study, a case study research will be carried out to addressthe research question and will provide an in-depth insight into the impact of network ties oncollaboration. Exploring tie types and their characteristics and associated collaboration outcomes indepth might aid in developing strategies to ensure productive and sustainable employee collaborationin ESN. In a broader sense, the work is expected to make important theoretical contributions to ISresearch that continuously seek to explore the role of technology in affecting job performance atworkplace.7 REFERENCESAdler, P. S., and Kwon, S.-W. 2002. "Social Capital: Prospects for a New Concept," Academy ofmanagement review (27:1), pp. 17-40.Ahuja, G. 2000. "Collaboration Networks, Structural Holes, and Innovation: A Longitudinal Study,"Administrative science quarterly (45:3), pp. 425-455.Allred, C. R., Fawcett, S. E., Wallin, C., and Magnan, G. M. 2011. "A Dynamic Collaboration Capabilityas a Source of Competitive Advantage," Decision sciences (42:1), pp. 129-161.Blomqvist, K., and Levy, J. 2006. "Collaboration Capability–a Focal Concept in Knowledge Creation andCollaborative Innovation in Networks," International Journal of Management ConceptsPhilosophy (2:1), pp. 31-48.Borgatti, S. P., and Halgin, D. S. 2011. "On Network Theory," Organization science (22:5), pp. 1168-1181.Borgatti, S. P., Mehra, A., Brass, D. J., and Labianca, G. 2009. "Network Analysis in the Social Sciences,"Science (323:5916), pp. 892-895.Burt, R. S. 2004. "Structural Holes and Good Ideas," American journal of sociology (110:2), pp. 349-399.Buvik, M. P., and Rolfsen, M. 2015. "Prior Ties and Trust Development in Project Teams–a Case Study fromthe Construction Industry," International Journal of Project Management (33:7), pp. 1484-1494.Claiborne, N., and Lawson, H. A. 2005. "An Intervention Framework for Collaboration," Families inSociety (86:1), pp. 93-103.Cross, R., Borgatti, S. P., and Parker, A. 2002. "Making Invisible Work Visible: Using Social NetworkAnalysis to Support Strategic Collaboration," California Management Review (44:2), pp. 25-46.Faraj, S., Jarvenpaa, S. L., and Majchrzak, A. 2011. "Knowledge Collaboration in Online Communities,"Organization science (22:5), pp. 1224-1239.Faraj, S., Kudaravalli, S., and Wasko, M. 2015. "Leading Collaboration in Online Communities," MISQuarterly (39:2), pp. 393-411.Fiori, K. L., Antonucci, T. C., and Cortina, K. S. 2006. "Social Network Typologies and Mental Healthamong Older Adults," The Journals of Gerontology Series B: Psychological Sciences SocialSciences (61:1), pp. 25-32.Granovetter, M. S. 1973. "The Strength of Weak Ties," American Journal of Sociology (78:6), pp. 1360-1380.Hollstein, B. 2014. "Mixed Methods Social Networks Research: An Introduction," Mixed methods socialnetworks research: Design and Application (1), pp. 3-34.Hurmelinna, P., Blomqvist, K., Puumalainen, K., and Saarenketo, S. 2005. "Striving Towards R&DCollaboration Performance: The Effect of Asymmetry, Trust and Contracting," CreativityInnovation Management (14:4), pp. 374-383.Inkpen, A. C. 1996. "Creating Knowledge through Collaboration," California management review(39:1), pp. 123-140.Kane, G. C. 2015. "Enterprise Social Media: Current Capabilities and Future Possibilities," MISQuarterly Executive (14:1), pp. 1-16.Kane, G. C. 2017. "The Evolutionary Implications of Social Media for Organizational KnowledgeManagement," Information and Organization (27:1), pp. 37-46.
Page 7
Australasian Conference on Information SystemsKochar, Watson & Ouyang2019, Perth Western AustraliaEffect of ties on employee collaboration717Kane, G. C., Alavi, M., Labianca, G. J., and Borgatti, S. 2014. "What’s Different About Social MediaNetworks? A Framework and Research Agenda," MIS Quarterly (38:1), pp. 275-304.Kuegler, M., Smolnik, S., and Kane, G. C. 2015. "What's in It for Employees? Understanding theRelationship between Use and Performance in Enterprise Social Software," Journal of StrategicInformation Systems (24:2), pp. 90-112.Lazzeretti, L., and Capone, F. 2016. "How Proximity Matters in Innovation Networks Dynamics Alongthe Cluster Evolution. A Study of the High Technology Applied to Cultural Goods," Journal ofBusiness Research (69:12), pp. 5855-5865.Lee, Y., Cho, I., and Park, H. 2015. "The Effect of Collaboration Quality on Collaboration Performance:Empirical Evidence from Manufacturing Smes in the Republic of Korea," Total QualityManagement & Business Excellence (26:9-10), pp. 986-1001.Leonardi, P. M., Huysman, M., and Steinfield, C. 2013. "Enterprise Social Media: Definition, History,and Prospects for the Study of Social Technologies in Organizations," Journal of Computer-Mediated Communication (19:1), pp. 1-19.Levin, D. Z., and Cross, R. 2004. "The Strength of Weak Ties You Can Trust: The Mediating Role of Trustin Effective Knowledge Transfer," Management science (50:11), pp. 1477-1490.Liu, J. M., and Moskvina, A. 2016. "Hierarchies, Ties and Power in Organisational Networks: Model andAnalysis," Social Networks Analysis and Mining (6:1), p. 106.Marineau, J. E., Labianca, G. J., and Kane, G. C. 2016. "Direct and Indirect Negative Ties and IndividualPerformance," Social Networks (44), pp. 238-252.Riemer, K., Finke, J., and Hovorka, D. 2015. "Bridging or Bonding: Do Individuals Gain Social Capitalfrom Participation in Enterprise Social Networks?," in: Thirty Sixth International Conference onInformation Systems. Fort WorthSantos-Vijande, M. L., López-Sánchez, J. Á., and Rudd, J. 2016. "Frontline Employees’ Collaboration inIndustrial Service Innovation: Routes of Co-Creation’s Effects on New Service Performance,"Journal of the Academy of Marketing Science (44:3), pp. 350-375.Scherngell, T., Wanzenbock, I., and Berge, L. 2016. "Bridging Centrality: A New Indicator to Measurethe Positioning of Actors in R&D Networks," 21st International Conference on Science andTechnology Indicators, I. Rafols, J. MolasGallart, E. CastroMartinez and R. Woolley (eds.),Valencia, pp. 1106-1116.Schlagwein, D., and Hu, M. 2017. "How and Why Organisations Use Social Media: Five Use Types and TheirRelation to Absorptive Capacity," Journal of Information Technology (32:2), pp. 194-209.Shi, Q., Xu, B., Xu, X., Xiao, Y., Wang, W., and Wang, H. 2011. "Diversity of Social Ties in ScientificCollaboration Networks," Physica: A Statistical Mechanics and its Applications (390:23), pp.4627-4635.Walsh, M. 2003. "Teaching Qualitative Analysis Using Qsr Nvivo," The Qualitative Report (8:2), pp.251-256.Wood, D. J., and Gray, B. 1991. "Toward a Comprehensive Theory of Collaboration," The Journal ofApplied Behavioral Science (27:2), pp. 139-162.Wright, P. M., McMahan, G. C., and McWilliams, A. 1994. "Human Resources and SustainedCompetitive Advantage: A Resource-Based Perspective," International journal of humanresource management (5:2), pp. 301-326.Xue, X. L., Zhang, X. L., Wang, L., Skitmore, M., and Wang, Q. 2018. "Analyzing CollaborativeRelationships among Industrialized Construction Technology Innovation Organizations: ACombined Sna and Sem Approach," Journal of Cleaner Production (173), pp. 265-277.Yang, X., Zhang, X., and Gallagher, K. P. 2016. "The Moderating Effect of Online Community Affiliationand Information Value on Satisfaction with Online Travel Communities in China," Journal ofGlobal Information Technology Management (19:3), pp. 190-208.Copyright: © 2019 Kochar, Watson & Ouyang. This is an open-access article distributed under theterms of the Creative Commons Attribution-NonCommercial 3.0 Australia License, which permits non-commercial use, distribution, and reproduction in any medium, provided the original author and ACISare credited.
Page 1
Australasian Conference on Information SystemsMwenya & Brown2019, Perth Western AustraliaChampioning Accountability in Cloud Computing333Cloud privacy and security issues beyond technology:championing the cause of accountabilityFull PaperJoshua K. MwenyaDepartment of Information SystemsUniversity of Cape TownCape Town, South AfricaEmail: mwnjos003@myuct.ac.zaIrwin BrownDepartment of Information SystemsUniversity of Cape TownCape Town, South AfricaEmail: irwin.brown@uct.ac.zaAbstractCloud computing provides IT service providers increased efficiency of resource utilization whileenabling consumers to benefit from innovative advantages like access to up-to-date IT resources andlow upfront investment. A significant hindrance to adoption of cloud computing is the lack of trustarising from worries over privacy and security when data resources of cloud service consumers arehandled by third parties. A key factor in fostering cloud privacy and security is accountability, whichincreases trust by obligating an entity to be answerable for its actions. This paper uses a hermeneuticliterature review to investigate (i) the prevailing methods and strategies of fostering privacy and securitythrough accountability, (ii) the key actors in championing cloud accountability and (iii) the key barriersto cloud accountability. This literature review provides insight into current practices associated withchampioning cloud accountability and contributes to cloud service provider awareness of ways toimprove cloud computing trustworthiness.Keywords Accountability, Hermeneutic, Privacy, Security, Trust
Page 2
Australasian Conference on Information SystemsMwenya & Brown2019, Perth Western AustraliaChampioning Accountability in Cloud Computing3341 INTRODUCTIONIn recent years cloud computing has emerged as a key information technology (IT) service deliveryparadigm and a major innovation driver that offers a new business model that suits both IT serviceconsumers and IT service providers (Aguez et al. 2016). For IT service providers, cloud computingprovides new opportunities, such as realization of economies of scale by increasing efficiency of resourceutilization (Sunya and Schneider 2013). For IT service consumers, cloud computing is a technology thatallows organizations to selectively adopt specific resources from a wide range of cloud-based servicesand to outsource their entire IT based businesses process so they can concentrate more on their corebusiness (Diener et al. 2016; Khana and Al-Yasiri2016). However, evidence indicates that migration tothe cloud paradigm is often hampered by concerns over security and privacy (Coppolino et al. 2017;Mazhar et al. 2015). A major impediment to cloud adoption is the lack of trust by potential customers,arising from the worry over privacy, security, and data protection when data resources are handled bythird parties and accessed via networks (Adjei 2015; Habib et al. 2012; Ko et al.2011; Pearson 2011).Trust in a cloud service provider (CSP) is an important issue and the lack of this trust is considered oneof the biggest concerns preventing cloud computing from quickly attaining its full technical, social, andeconomic potential (McLeod and Gormly 2017). For cloud computing to earn the full trust it deserves,cloud service consumers should be able to store their data in the cloud with the same confidence thatthey have when they deposit their money and other valuables in banks (Asadi et al. 2017). Trust iscomprised of four main components: (i) Security - the mechanisms which make it difficult oruneconomical for an unauthorised person to access some information; (ii) Privacy - the protectionagainst the exposure or leakage of personal or confidential data; (iii) Auditability - the relative ease ofauditing a system or an environment; and (iv) Accountability - the obligation and/or willingness todemonstrate and take responsibility for performance in light of agreed-upon expectations (Al-Rashdi etal. 2015; Ko et al. 2011).This paper takes the view that accountability is the main construct and key enabler of trust and thatachievement of accountability in a cloud environment brings about the other three trust components(privacy, security and auditability) as by-products. The paper contributes to research and knowledge oncloud privacy and security by establishing a relationship between accountability and both privacy andsecurity and then addressing privacy and security issues through accountability by investigating theprevailing accountability practices in cloud ecosystems, the key actors in championing cloudaccountability, and the key barriers and challenges associated with championing and implementingcloud accountability. The rest of this paper is organized as follows: The next section highlights theconcept of accountability and its dimensions that are relevant to trust (and, hence, to privacy andsecurity) building and explains how the achievement of accountability brings about privacy, security andauditability. The section ends with identification of our research problem and research questions. Thethird section outlines the research methodology we adopted - a hermeneutic circle based literaturereview. The fourth section presents our research findings while the fifth section gives a brief discussionof the findings and provides suggestions for further research. The paper concludes with section six.2 CONCEPT OF ACCOUNTABILITY2.1 Accountability and its AttributesAccountability is about defining governance to comply in a responsible manner with internal andexternal criteria, ensuring implementation of appropriate actions, explaining and justifying thoseactions and remedying any failure to act properly (Contractor and Patel 2017; Felici et al. 2013). Bovens(2007, p.450) defines accountability as “a relationship between an actor and a forum, in which the actorhas an obligation to explain and to justify his or her conduct, the forum can pose questions and passjudgement, and the actor may face consequences”. In terms of its components, accountability is viewedand interpreted in terms of a set of key attributes and properties, including the following five keyattributes identified by several researchers: transparency, responsibility, remediability, attributabilityand verifiability (Felici and Pearson 2014; Al-Rashdi et al. 2017; Jaatun et al. 2018). Viewingaccountability in terms of the foregoing key attributes strengthens the view of accountability as a basisfor satisfaction of obligations along the cloud service provision chain, which ensures that all partnersare accountable and that there has been proper allocation of responsibilities along the service provisionchain (Contractor and Patel 2017; Pearson et al. 2012). Implementing accountability by putting its coreattributes into practice is identified as an effective method of addressing issues of privacy, security andtrust (Pearson and Benameur 2010). Achievement of accountability is, thus, a practical and effectivecatalyst for bringing about cloud privacy, security and trust. Table 1 describes the five key accountabilityattributes that organisations put into practice to enforce accountability.
Page 3
Australasian Conference on Information SystemsMwenya & Brown2019, Perth Western AustraliaChampioning Accountability in Cloud Computing335Accountability attributeDescriptionAttributabilityThe possibility to trace a given action back to a specific entity.RemediabilityThe property of a system, organization or individual to take correctiveaction and/or provide a remedy for any party harmed in case of failure tocomply with its governing norms.ResponsibilityThe property of an organization or individual in relation to an object,process, or system of being assigned to take action to be in compliancewith the norms.TransparencyThe property of an accountable system that is capable of giving accountof, or providing visibility of how it conforms to its norms, governing rulesand commitments.Verifiability:The extent to which it is possible to assess norm compliance.Table 1. Key Accountability Attributes2.2 How Accountability brings about Privacy, Security and AuditabilityHow accountability brings about privacy: According to ISO/IEC29100 guidelines, accountabilityrequires a data controller to document policies, procedures and practices, assign the duty to implementprivacy policies to specified individuals in the organization, provide suitable training, inform aboutprivacy breaches, and give access to effective sanctions and procedures for compensations in case ofprivacy breaches (Berthold 2013). Furthermore, full accountability is derived from contracts and othertransparency mechanisms that govern active interactions among cloud stakeholders, all with theprimary objective of reducing the risk of disproportionate harm to the data subjects and permitting theamelioration of negative consequences for the data controllers in case of harms arising from failure toprovide sufficient privacy protection (Pearson and Charlesworth 2009). This means that accountabilityimposes transparency and privacy liability on cloud data controllers and their partners in the servicedelivery chain, ensuring that achievement of accountability by cloud data controllers engenders privacyas a by-product.How accountability brings about security: Cloud security is often compromised by the lack of orabsence of several key attributes, notably confidentiality (ensuring that a customer’s data andcomputation tasks performed on the data are kept confidential from both the cloud service providersand other customers), integrity (data integrity which ensures that a customer’s data is honestly storedon cloud servers and computation integrity which ensures that data manipulation programs areexecuted without being distorted by malware, cloud providers, or other malicious users and that anyincorrect computing is detected), and availability (ensuring that each expected service is available andthe quality of service meets the agreed Service Level Agreement) (Xiao and Xiao 2013). Accountabilityprovides constraints and control mechanisms for cloud data controllers and others in the serviceprovision chain by encompassing the obligation for each one to act as a responsible steward of thepersonal information of others, to take responsibility for the protection and appropriate processing anduse of that information beyond mere legal requirements and to provide remediation in case of failure toensure availability, confidentiality and integrity of the data (Pearson and Wainwright 2013). Thus,achieving accountability engenders cloud security as a by-product.How accountability brings about auditability: Auditability is an enabler of accountability in thatauditability ensures that events are recorded while accountability ensures that events deemed importantare logged and not missed (Doiphod and Channe 2015; Ko 2013). Auditability helps ensure availabilityof evidence required by accountability in determining that both users and cloud service providers at alllevels are in compliance with security and privacy policies. Auditability serves as a retrospective enablerof accountability as auditability furnishes evidence allowing an action to be reviewed against a pre-determined policy, enabling relevant parties to hold accountable the person or organization responsiblefor that action (Ko et al.2011). Thus achievement of accountability requires that auditability be attainedas a by-product.2.3 Research Problem and QuestionsIn practice, implementing accountability in clouds raises several compelling issues that should concerncloud privacy and security researchers. First, accountability has the power to increase cloud trust(Doiphod and Channe 2015), but its implementation can produce contrasting and unintended outcomesfor different actors (Pearson and Benameur 2010). Second, accountability aspects such as regulation canstifle innovation and thwart the desired cloud trust increases if it is not introduced in an intelligent way(Pearson 2011). Third, because its implementation can yield a positive outcome for one cloud provider
Page 4
Australasian Conference on Information SystemsMwenya & Brown2019, Perth Western AustraliaChampioning Accountability in Cloud Computing336while at the same time yielding negative outcomes for others, accountability has been identified asneeding urgent attention (Ko et al. 2011). It is these three key concerns that pose a relevant problem andprovide the motivation for this research study. Research on the aforementioned aspects of accountabilityin clouds enables us to understand emerging relationships among cloud actors and allows us to identifyaccountability based mechanisms and appropriate tools available to support privacy, security andtrustworthiness in cloud ecosystems (Felici et al. 2013).As regards our research focus, we agree with Al-Rashid et al. (2017) that research in the area of cloudsecurity has been largely technical in nature, creating a need for more research focused on non-technicalaspects. Thus, our paper focuses more on non-technical approaches to championing and implementingaccountability through a combination of public law (legislation and regulation), private law (contractsand SLAs) and self-regulation (through standards and certification). Specifically, this study contributesto the body of knowledge on the role of accountability in fostering cloud privacy, security and trust byanswering three related questions:What are the prevailing non-technical approaches to championing the cause of accountability in cloudcomputing?Who are the key actors in championing the cause of accountability in cloud computing?What are the key barriers and challenges to championing the cause of accountability in cloudcomputing?3 RESEARCH METHODOLOGYThis paper adopts a hermeneutic literature review approach, a literature review framework known to bewell-suited for theory generation and knowledge building (Boell and Cecez-Kecmanovic 2014;Greenhalgh et al. 2018). The hermeneutic approach assumes that the meaning emerges through adialogue between a text (a paper) and a reader, through an inherently interpretive process which enablesthe researchers to expand and deepen their understanding of relevant literature as they iterativelyinterpret a paper from their own pre-understanding of literature and then incrementally develop betterunderstanding of the literature based on each interpreted paper (Boell and Cecez-Kecmanovic 2014;Baghizadeh et al. 2019). In this study, the literature searching started with Google Scholar, a popularand flexible scholarly search facility and the hermeneutic circle was implemented through an iterativeprocess. The initial set of articles was obtained by querying Google Scholar for particular keywords andphrases appearing in the title of published articles as shown in the first column of Table 2.Keyword search phraseNumber ofArticlesreturnedMeetinginclusioncriteriaRejectedas lessrelevantNumberincludedallintitle: Accountability for cloud Privacy9514allintitle: Accountability for cloud security3101allintitle: Cloud computing accountability44918allintitle: Cloud computing regulation36606allintitle: Cloud computing trust511372314allintitle: Trust accountability cloud computing9321allintitle: Accountability in cloud1621165allintitle: Cloud trust issues34633allintitle: Cloud computing legal178734Total986853946Table 2. Keyword Search phrases and Results relevant to research issuesThe returned articles totalling 986 were initially reviewed both by title and abstract in order to filter outarticles that were not relevant to the key concepts and to ensure investigating accountability in cloudcomputing mainly from a non-technical perspective, thereby ruling out articles that focused on purelytechnical aspects and solutions. The hermeneutic principles are accomplished through two interlinkedcycles: 1) accessing and interpreting the literature (column 2 and 3 of Table 2), focusing on a systematicbut flexible and iterative searching and 2) understanding and developing an argument (column 3 and 4of Table 2), focusing on recognising emerging ideas and perspectives and rejecting less relevant sourcesthrough progressive focusing (Boell and Cecez-Kecmanovic, 2014). The final part of the review resultedin identification of three main themes that emerged from the findings regarding our research questions:
Page 5
Australasian Conference on Information SystemsMwenya & Brown2019, Perth Western AustraliaChampioning Accountability in Cloud Computing337(i) articles that identify current approaches and practices adopted in championing the cause ofaccountability in cloud computing; (ii) articles that identify the key actors and stakeholders thatchampion the cause of accountability in cloud computing; and (iii) articles that identify key barriers andchallenges to the implementation of accountability in cloud computing. Finally, based on the resultingknowledge generated from hermeneutic literature review, we addressed the argument developmentprocess by identifying emerging issues and providing suggestions for further research and direction.4 RESEARCH FINDINGSOur research findings are presented in the following three tables. To answer our first research question,we took a deep look into the literature for the prevailing non-technical approaches and methods thatstakeholders adopt to help enhance one or more of the five key attributes or properties of accountabilityidentified by Felici and Pearson (2014), Al-Rashdi et al. (2017) and Jaatun et al. (2018). We identifiednine (9) prevailing non-technical approaches as shown in Table 3. It should be noted that even thoughJaatun et al. (2018) found nine (9) core attributes, only five key attributes listed in Table 1 were foundto be relevant for this study. Other attributes were not considered relevant. For example, effectivenessand appropriateness measure technical aspects while observability is an element of transparency.PrevailingAccountabilityPracticeAccountabilityaspect enhancedHow the identified accountabilityproperty is enhanced or enactedLiterature Source(s)LegislationResponsibilityLegislation, such as the EU DataProtection Act, create obligations onservice providers to engage in sounddata governance and stewardship,providing a basis for responsibility.Pearson et al. (2012);Ryan (2013)Auditing byexternalagents/entitiesTransparency,verifiability,attributabilityAuditing allows an action by any cloudactor to be reviewed against a pre-determined policy and to shed light oncompliance.Ryoo et al. (2014)ContractualassurancesResponsibility,remediabilityContractual assurances promoteaccountability by enabling parties to acloud contract to both claim their rightsand fulfil their obligations.Ryan (2013);Seddona and Currie(2013)Third-partyCertificationTransparency,verifiabilityThird-party certification enables cloudproviders to implement accountabilityand give users and other datagovernance actors a way to check andmonitor use of data in clouds.Pearson et al. (2012)Imposition ofPenaltiesAttributability,remediabilityFailure to comply with regulation canlead to costly penalties: e.g. violation ofHIPAA in the USA earns a maximumpossible fine of $1.5 million.Al-Rashdi et al.(2017); Hoover(2013)ComplianceRegulationVerifiability,responsibilityWith strict accountability in place,compliance regulators enforce the lawon the ‘first in the chain’ of cloudproviders in regard to the misdeeds ofanybody in the chain.Pearson andCharlesworth (2009);Takabi et al. (2010)Service levelagreements(SLA)Transparency,verifiability,remediabilityAn accountable CSP is not only able toguarantee service availability via SLA,but must also provide documentation toshow that their service is availablewhen the customer needs evidence.Mazhar et al. (2015);Pearson (2011)Enforcement ofindustrystandardsResponsibility,transparencyStandards such as ISO/IEC 27018 cloudstandard primarily aim at fostering andverification of legal and/or contractualcompliance and transparency.de Hert et al. (2016);Löhe and Blind(2015)Monitoring byspecial interestgroups andmarket places.Transparency,verifiabilityNotably, the A4Cloud project helpspromote accountability by holding CSPsaccountable through an orchestratedset of preventive, detective andcorrective mechanisms.Habib et al. (2012);Pearson et al. (2012)Table 3. Prevailing approaches to championing the cause of accountability
Page 6
Australasian Conference on Information SystemsMwenya & Brown2019, Perth Western AustraliaChampioning Accountability in Cloud Computing338To answer our second question, we examined the literature to identify those actors that played aprominent and active role in championing the cause of accountability in cloud computing. We identifiedsix (6) key actors as shown in Table 4 which indicates a combination of institutions of various types,ranging from national governments to special interest groups.Name of actorType of actorChampioning activitiesSource (s)Nationalgovernments (e.g.USA, German, UK,China, Spain,Russia)GovernmentalbodiesGovernments impose a variety oftailored data protection laws andpenalties (e.g. the HIPAA Omnibus lawin the United States, the German dataprotection Law, Golden Shield Project ofChina, and the Russian data storagelocalization law enacted in 2015).Hoover (2013);Maughan (2016)Millard (2015);Rieger et al. (2013);Yaraghi and Gopal(2018)Office for CivilRights (OCR) inthe USAUnited StatesGovernmentalagenciesEnforces the Health InformationPortability and Accountability Act(HIPAA), which protects the privacy ofindividually identifiable personal healthinformation.Klein (2011); Ryoo etal. (2014); Seddonaand Currie (2013)The EuropeanUnion (EU)Inter-governmentalbodyThe EU develops regulations andstandards which the 27 Member Statesmust embed into their own nationaldata privacy and security laws that applywhenever an individual or institutioncollects personal data related to an EUcitizen.Seddona and Currie(2013)InternationalStandardsOrganization (ISO)Independent/Professional bodiesProduces industry standards such as theISO/IEC-27018, which address the lackof trust and transparency, by developingcontrols and recommendations forCSPs.de Hert et al. (2016]Cloud SecurityAlliance (CSA)-incorporated in theUSANot-for-profitindustryorganizationconcerned withcloud security.Promotes accountability via a toolkitused by key stakeholders to assessclouds against industry established bestpractices, standards and compliancerequirements.Habib et al. (2012)The CloudAccountability(A4Cloud) ProjectA European basedCloud AccountabilityInitiative fullyfunded by EUThe primary focus of this Project isaccountability under data protectionlaws for personal data processed incloud service provision ecosystems:accountability obligations owed by CSPsto other cloud stakeholders.Pearson et al. (2012)Table 4. The key actors responsible for championing the cause of accountabilityTo answer our third question, we examine literature to identify the major challenges and barriers thatmay negatively impact on stakeholder efforts in championing the cause of accountability in cloudcomputing. We identified eight (8) significant barriers and challenges as shown in table 5.Key sources of challengesto accountabilityNature or description of accountabilitybarriers and/or challengesSource (s)Government surveillance orinterventionGovernment’s surveillance or intervention, suchas the USA Patriot Act (UPA) of 2001, may poseserious challenges to cloud accountability byobliging cloud suppliers and service providers,for reasons of national security or other reasons,to provide government agencies access tocustomer data without consent of the customers,thereby violating SLAs.Aguez et al. (2016);Fernandes et al. (2014);Marston et al. (2011)Self-regulatory and industry-targeted regulatoryapproachesSelf-regulatory mechanisms such as the SafeHarbour Agreement are viewed as inadequateand legitimately seen as a way of watering downexisting strong privacy protections, notablythose granted to EU citizens.King and Raja (2012);Pearson et al. (2012);Yang and Borg (2012)
Page 7
Australasian Conference on Information SystemsMwenya & Brown2019, Perth Western AustraliaChampioning Accountability in Cloud Computing339Key sources of challengesto accountabilityNature or description of accountabilitybarriers and/or challengesSource (s)Lack of genuine Transparencyand verifiability in CSP servicelevel agreements (SLAs)Most often, Service Level Agreements lacktransparency as they are made using non-negotiable standard contracts which mainly dealwith protecting the rights of the CSP, neglectingconsumer needs. This leads to distrust of cloudstakeholders and diminishes accountability.Fernandes et al. (2015);Khan (2016); Ryan(2013); Sfondrini et al.(2015)Challenges due to cloud datalocation.CSPs ensure efficient service availability byreplicating data in multiple data centres. Thus,cloud based data is stored on the CSP’s serversin undisclosed locations, which could be in theUSA, Europe, or anywhere else. This key tenet ofthe cloud business model conflicts with variouslegal requirements, notably in EU and Russia.AbuOliem (2013);Hon and Millard(2018); Takabi et al.(2010); Yaraghi andGopal (2018)Challenges to enforcement ofISO standardsStandards like the ISO/IEC 27018 act as non-legal forms of regulation by complementing legalregulations. However, the audit and certificationof compliance with ISO/IEC 27018 is not drivenby public authorities, but by private entities.This tends to leave open a choice for some CSPsto ignore key aspects like interoperability.de Hert et al. (2016);Löhe and Blind (2015)Securing the accountability ofsubcontractors and CSPemployees not guaranteedAlthough a contract may exist to forbid the CSPfrom disclosing the data to third parties, it maybe difficult to implement because employees andsubcontractors of the CSP may not be lockedinto the contract too, making it very hard tooblige them all to the terms and standardsrequested by the data owner.Felici et al. (2013);Mazhar et al. (2015;Ryan (2013)Conflicting legal structures ofdifferent countries.Incompatible in legislative regimes of differentcountries pose serious challenges toaccountability. For example, the USA PATRIOTAct is known to conflict with both the PersonalInformation Protection and ElectronicDocuments Act (PIPEDA) of Canada and EUData Protection Directive.Fernandes et al. (2014);Pearson andCharlesworth (2009)Regulatory and Compliancechallenges in highly regulatedsectorsHighly regulated sectors like banking andhealthcare face unique cloud accountabilitychallenges. Many banking regulators requirethat financial data for banking customers stay inhome country regulations require that bankingdata does not get intermixed with other data.Bejju (2014); Maughan(2016); Ryoo et al.(2014); Young and Borg(2012)Table 5. Key challenges to championing the cause of accountability5 DISCUSSION AND IMPLICATIONSThis literature review provides several insights on aspects of cloud accountability: Firstly, the literaturerevealed five key attributes of accountability and current practices that cloud service providers need tobe aware of if they are to be considered accountable CSPs who contribute to making cloud computingmore trustworthy. Secondly, unlike some of the previous waves in computing, cloud computing raisessignificant challenges in identifying who are the responsible entities, in order to assign accountabilityobligations. Multiple actors are involved at various levels and, thus, cloud computing also demands athoughtful and coordinated response from governmental agencies. Appropriate domestic laws can thenbe applied in order to ensure service providers protect sensitive data in certain sectors. Thirdly, somekey actors in championing accountability are also seen as sources of challenges to accountability inclouds. For example the USA government is a leading champion of accountability through theintroduction and enforcement of various relevant and progressive regulations and Acts such asSarbanes-Oxley and HIPAA Omnibus laws. The US PATRIOT Act of 2001, on the other hand, passedand enforced by the USA government, is cited as an example of an Act that fails to adequately protectdata privacy by forcing the disclosure of data to government entities without the consent of data owners.Acts of this type may pose challenges to cloud accountability. They may lead to CSP violations of existingterms of SLAs and subsequent breakup of the chain of trust created between cloud customers and cloudproviders. Fourthly, the literature revealed that conflicts and inconsistencies among legislative regimes
Page 8
Australasian Conference on Information SystemsMwenya & Brown2019, Perth Western AustraliaChampioning Accountability in Cloud Computing340of different countries pose serious challenges to accountability particularly for highly regulated sectorslike banking and healthcare.To complete our hermeneutic review, we address the argument development aspect of the hermeneuticreview by identifying emerging issues and proposing a research agenda for future research directionswith regard to some of the key accountability methods and associated challenges identified in this study.The proposed research agenda is shown in Table 6. For each research issue identified, a description ofthe research concern is presented and pertinent research questions suggested.ResearchIssuesDescription of Research ConcernsSuggested Research questionsGovernmentsurveillanceand access tocloud dataTo prevent and fight cyber related crimes, nationalgovernments have a justifiable need to access cloudbased personal data for purposes such as adducingevidence. However, insights from this studyindicate that government actions in this area have apotential to violate fundamental privacy rights ofindividuals and, in some cases, actions taken by onegovernment may result in violation of anothercountry’s sovereignty. What about the effect ofgovernment cloud data access on CSP obligations?For example, under the EU General Data ProtectionRegulation (GDPR), it is no longer possible for CSPprocessors to excuse themselves as mere processorsand escape the reach of data protection rules bypassing blame to data controllers.What laws exist, or should beformulated, to oblige a government orits agents to follow a lawful procedurewhen seeking to access cloud basedpersonal data?What laws or regulations are therethat oblige cloud service providers tonotify their customers when agovernment or its agents access theircloud-based privacy data?RegulationChallengesInsights from this study indicate that regulation is asource of several issues of research interest. Thecloud computing business model thrives onreducing the levels of control and visibility thatcloud consumers have on their data as data isstored and manipulated away from visibility to dataowners/subjects. On the other hand, a key objectiveof strong regulations such the EU directives ispreservation of such control. Given that other keyplayers like the US have less strict regulation in thisaspect compared to the EU, the resulting regulatoryinconsistencies and fragmentation among variousjurisdictions pose accountability challenges forCSPs.The picture may get even more complex whenregulatory and legal regimes in other jurisdictionsare compared to those of the EU and the US.What efforts are there towardachieving global regulatoryharmonization to promote increasedglobal consistency in accountabilityamong countries and regionaleconomic blocks?How do cloud regulatory frameworksin other countries compare with thosein the EU and the US given that keycloud service providers such asAmazon and Microsoft now host datacentres across the globe?How do global cloud service providersreconcile existing laws of onejurisdiction with contradictory legalrequirements of another?StandardizationchallengesLack of cloud service standardization compromisesinteroperability among cloud platforms, therebyreducing portability of cloud services. In turn, lackof portability promotes vendor lock-in which couldbe harmful to cloud consumers by preventing themfrom moving from one cloud provider to anotherwhen need arises to maximize business. Further,vendor lock-in can become a major problem in caseof bankruptcy of the preferred cloud provider.How do current standardizationefforts and frameworks account forcustomer inputs and interests?How does lack of standardization ofcloud based platforms and servicesinfluence vendor lock-in?Supply chainand InsiderAbuse IssuesInsights from this study indicate that insiders suchas employees and subcontractors can abuse theirposition and compromise a CSP’s contractual andaccountability norms. For example, since the cloudbusiness model thrives on spreading data over anumber of different storage devices while consumerhas reduced visibility regarding where their data isphysically stored, it is not feasible for the consumerto verify secure deletion of their data when the datais deleted. This may create a loophole that insiderssuch as employees can maliciously exploit for dataexfiltration from remnants of unsecured deletions.What mechanisms enable a cloudconsumer or regulatory agent toverify how a cloud service providerenforces compliance of insiders andsubcontractors?What mechanisms ensureaccountability in cases where two ormore cloud providers are involved inproviding a service to one consumersuch as one customer using Microsoft365 that is running on an AmazonECS instance?Table 6. Proposed research directions for cloud accountability
Page 9
Australasian Conference on Information SystemsMwenya & Brown2019, Perth Western AustraliaChampioning Accountability in Cloud Computing3416 CONCLUSIONThe aim of this paper is to contribute to a better understanding of accountability as a key non-technicalmechanism for promoting privacy and security (and hence trust) in cloud computing and, by so doing,contribute to cloud service provider awareness of ways to improve cloud computing trustworthiness andcloud service adoption. This paper focuses more on non-technical approaches to implementation ofaccountability as research in areas of cloud security and privacy has been largely technical. This studyhas applied a hermeneutic literature review to provide new insights regarding the prevailing practicesin championing accountability. It has identified the main actors actively involved in championing thecause of accountability and highlighted the key barriers and challenges to championing accountabilityin cloud computing. As a result of this analysis, a research agenda is proposed for future studies. One ofthe key limitations of the findings is that most of the literature addresses cloud computing issues indeveloped countries, notably the USA, Canada and the EU. The lack of literature from elsewhere offersopportunity to investigate the phenomenon in alternative contexts.7 REFERENCESAbuOliem, A. 2013. “Cloud computing regulation: An attempt to protect personal data transmission tocross-border cloud computing storage services,” International Journal of Computer andCommunication Engineering, (2:4), pp 521-525.Adjei, J.K. 2015. “Explaining the role of trust in cloud computing services,” info., (17:1), pp 54-67.Aguez, E.L.K., Hajji, N., and Barka, H. 2016. “The cloud computing: the impact of regulation onadoption,” International Journal of Computers, (1), pp 22-32.Al-Rashdi, Z., Dick, M., and Storey, I. 2015. “A conceptual framework for accountability in cloudcomputing service provision,” in The Australasian Conference on Information Systems,Adelaide, Australia.Al-Rashdi, Z., Dick, M., and Storey, I. 2017. “Core elements in information security accountability in thecloud,” In Valli, C. (Ed.). 2017. The Proceedings of 15th Australian Information SecurityManagement Conference, pp 125-131 Perth, Australia.Asadi, S., Nilashi, M., Husin, A.R.C., and Yadegaridehkordi, E. 2017. “Customers perspectives onadoption of cloud computing in banking sector,” Information Technology Management, (18), pp305-330.Baghizadeh, Z., Cecez-Kecmanovic, D., and Schlagwein, D. 2019. “Review and critique of theinformation systems development project failure literature: An argument for exploringinformation systems development project distress,” Journal of Information Technology (00:0),pp 1–20. DOI: 10.1177/0268396219832010.Bejju, A. 2014. “Cloud computing for banking and investment services,” Advances in Economics andBusiness Management, (1:2), pp 34-40.Berthold, S., Fischer-Hubner, S., Martucci, L., and Pulls, T. 2013. “Crime and punishment in the cloud- accountability, transparency, and privacy,” in: Pre-Proceedings of International Workshop onTrustworthiness, Accountability and Forensics in the Cloud in conjunction with the 7th IFIP WG11.11 International Conference on Trust Management.Boell S.K., and Cecez-Kecmanovic D. 2014. “A hermeneutic approach for conducting literature reviewsand literature searches. Communications of the Association for Information Systems, (34:12), pp257-286.Bovens, M. 2007. “Analysing and Assessing Accountability: A conceptual framework,” European LawJournal, (13:4), pp 447-468.Contractor, D., and Patel, D. 2017. “Accountability in Cloud Computing by Means of Chain of Trust,”International Journal of Network Security, (19:2), pp 251-259.Coppolino, L., D’Antonio, S., Mazzeo, G., and Romano, L. 2017. “Cloud security: Emerging threats andcurrent solutions,” Computers and Electrical Engineering. (59), pp 126-140.de Hert, P., Papakonstantinou, V., and Kamara, I. 2016. “The cloud computing standard ISO/IEC 27018through the lens of the EU legislation on data protection,” Computer Law & Security Review,(32), pp 16-30.
Page 10
Australasian Conference on Information SystemsMwenya & Brown2019, Perth Western AustraliaChampioning Accountability in Cloud Computing342Diener, M., Blessing, L., and Rappel, N. 2016. “Tackling the Cloud Adoption Dilemma - A User CentricConcept to Control Cloud Migration Processes by Using Machine Learning Technologies,” 11thInternational Conference on Availability, Reliability and Security (ARES), pp 776-785, Salzburg.Felici M., Jaatun M.G., Kosta E., and Wainwright N. 2013. “Bringing Accountability to the Cloud:Addressing Emerging Threats and Legal Perspectives,” in: Felici M. (eds) Cyber Security andPrivacy. CSP 2013. Communications in Computer and Information Science, (182). Springer,Berlin, Heidelberg.Felici, M., and Pearson, S. 2014. "Accountability, Risk, and Trust in Cloud Services: Towards anAccountability-Based Approach to Risk and Trust Governance," 2014 IEEE World Congress onServices, pp 105-112, Anchorage. doi: 10.1109/SERVICES.2014.29Fernandes, A.B., Soares, F.L., Gomes, J.V., Freire, M.M., and Inácio, R.P. 2014. “Security issues in cloudenvironments: a survey,” International Journal of Information Security, (13), pp 113-170.Greenhalgh, T., Thorne, S., and Malterud, K. 2018. “Time to challenge the spurious hierarchy ofsystematic over narrative reviews,” Eur JClin Invest., (48:1). doi.org/10.1111/eci.12931.Habib, S.M., Hauke, S., Ries. S., and Mühlhäuser, M. 2012. “Trust as a facilitator in cloud computing: asurvey,” Journal of Cloud Computing: Advances, Systems and Applications, (2012), pp 1-19.Hon, W.K., and Millard, C. 2018. “Banking in the cloud: Part 3 – contractual issues,” Computer Law &Security, (34), pp 595-614.Hoover, J.N. 2013. “Compliance in the ether: Cloud computing, data security and business regulation,”Journal of Business & Technology Law, (8:1), pp 255-273.Jaatun, M.G., Tøndel, I.A., Moe, N.B., Cruzes, D.S., Bernsmed, K., and Haugset, B. 2018. “Accountabilityrequirements in the cloud provider chain,” Symmetry, (10:4), pp 124-144.Khan, H.M., Chan, G., and Chua, F. 2016. “An adaptive monitoring framework for ensuringaccountability and quality of services in cloud computing,” International Conference onInformation Networking (ICOIN), pp. 249-253, Kota Kinabalu.Khana, N., and Al-Yasiri, A. 2016. “Identifying cloud security threats to strengthen cloud computingadoption framework,” Procedia Computer Science, (94), pp 485-490.King, J.N., and Raja, V.T. 2012. “Protecting the privacy and security of sensitive customer data in thecloud,” Computer Law & Security Review, (28), pp 308-319.Klein, A.C. 2011. “Cloudy confidentiality: Clinical and legal implications of cloud computing in healthcare,” The Journal of the American Academy of Psychiatry and the Law, (39:4), pp 571-578.Ko R.K.L., Lee, B.S., and Pearson, S. 2011. “Towards Achieving Accountability, Auditability and Trust inCloud Computing,” Communications in Computer and Information Science, (193). Springer,Berlin, Heidelberg.Ko, R.K.L. 2013. “Data accountability in cloud systems,” In Security, Privacy and Trust in CloudSystems, Springer-Verlag, Berlin, Heidelberg.Löhe, M.G., and Blind, K. 2015. “Regulation and standardization of data protection in cloud computing,”ITU Kaleidoscope: Trust in the Information Society, pp 1-6, Barcelona.Mazhar, A., Samee, U.K., and Athanasios, V.V. 2015. “Security in cloud computing: Opportunities andchallenges,” Information Sciences, (305), pp 357-383.McLeod, J., and Gormly, B. 2017. “Using the cloud for records storage: issues of trust,” Arch Sci., (17),pp 349–370.Marston, S., Li, Z., Bandyopadhyay, S., Zhang, J., and Ghalsasi, A. 2011. “Cloud computing: The businessperspective,” Decision Support Systems, (51:1), pp 176-189.Millard, C. 2015. “Forced Localization of Cloud Services: Is Privacy the Real Driver?” in IEEE CloudComputing, (2:2), pp 10-14.Maughan, A. 2016. “Cloud computing: A move toward harmonization, or continuation of a tiered serviceprovision?” Intellectual Property & Technology Law Journal, (28:11), pp 9-12.
Page 11
Australasian Conference on Information SystemsMwenya & Brown2019, Perth Western AustraliaChampioning Accountability in Cloud Computing343Pearson S., and Charlesworth A. 2009. “Accountability as a Way Forward for Privacy Protection in theCloud,” In: Jaatun M.G., Zhao G., Rong C. (eds) Cloud Computing. CloudCom 2009. LectureNotes in Computer Science, (5931), Springer, Berlin, Heidelberg.Pearson, S., and Benameur, A. 2010. “Privacy, security and trust issues arising from cloud computing,”IEEE Second International Conference on Cloud Computing Technology and Science, pp 693-702, Indianapolis, IN.Pearson, S. 2011. “Toward Accountability in the Cloud,” IEEE Internet Computing, (15:4), pp 64-69.Pearson, S., Catteddu, D., Südholt, M., Molva, R., Fischer-Hübner, S., Millard, C., Lotz, V., Jaatun, M.G.,Leenes, R., Rong, C., and Lopez, J. 2012. “Accountability for cloud and other future Internetservices,” 4th IEEE International Conference on Cloud Computing Technology and ScienceProceedings, Taipei, pp 629-632.Pearson, S., and Wainwright, N. 2013. “An interdisciplinary approach to accountability for futureinternet service provision”, International Journal of Trust Management in Computing andCommunications, (1:1), pp 52-72.Rieger, P., Gewald, H., and Schumacher, B. 2013. “Cloud-computing in banking: Influential factors,benefits and risks from a decision maker's perspective,” Proceedings of the Nineteenth AmericasConference on Information Systems, Chicago, Illinois.Ryan, D. M. 2013. “Cloud computing security: The scientific challenge, and a survey of solutions,” TheJournal of Systems and Software, (86), pp 2263-2268.Ryoo, J., Rizvi, S., Aiken, W., and Kissell, J. 2014. “Cloud security auditing: Challenges and emergingapproaches,” in IEEE Security & Privacy, (12:6), pp 68-74.Seddona, J.M., and Currie, W.L. 2013. “Cloud computing and trans-border health data: Unpacking U.S.and EU healthcare regulation and compliance,” Health Policy and Technology, (2013:2), pp 229-241.Sfondrini, N., Motta, G., and You, L. 2015. “Service level agreement (SLA) in public cloud environments:A survey on the current enterprises adoption,” 5th International Conference on InformationScience and Technology (ICIST), pp 181-185, Changsha, China.Sunya, A., and Schneider, S. 2013. “Cloud services certification: how to address the lack of transparency,trust, and acceptance in cloud services,” Communications of the ACM, (56:2), pp 33-36.Takabi, H., Joshi, J.B., and Ahn, G. 2010. “Security and Privacy Challenges in Cloud ComputingEnvironments,” in IEEE Security & Privacy, (8), pp 24-31.Yang, Y.T., and Borg, K. 2012. “Regulatory privacy protection for biomedical cloud computing,” BeijingLaw Review, (3), pp 145-151.Yaraghi, N., and Gopal, R.D. 2018. “The role of HIPAA Omnibus rules in reducing the frequency ofmedical data breaches: Insights from an empirical study,” The Milbank Quarterly, (96:1), pp 144-166.Xiao, Z., and Xiao, Y. 2013. “Security and privacy in cloud computing,” in IEEE CommunicationsSurveys & Tutorials, (15:2), pp 843-859.Copyright: © 2019 authors. This is an open-access article distributed under the terms of the CreativeCommons Attribution-NonCommercial 3.0 Australia License, which permits non-commercial use,distribution, and reproduction in any medium, provided the original author and ACIS are credited.