設計行動應用程式以增加運動依從性之研究－使用設計科學方法 - 政大學術集成

全文

(1)國立政治大學資訊管理學(系)研究所碩士學位論文. 設計行動應用程式以增加運動依從性. 治政大之研究－使用設計科學方法立. ‧ 國. 學. A mobile application for adherence improvement on exercise plan－using a design science approach.. ‧. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 指導教授：張欣綠博士研究生：孫若庭撰. 中華民國 106 年 06 月.

(2) 摘要運動依從性在健康管理議題裡是非常重要的一環，現在的人經常感到身體不適而診斷結果卻正常，這種現象根據世界衛生組織的定義為「亞健康」或「健康的灰色地帶」。在高壓的工作環境或不正常的生活作息下，最容易有這些徵狀，儘管大家都知道長期規律的運動可以促進健康，實際實行的狀況卻不如預期的好。為了改善此狀況，本研究透過蘋果公司的套件(ResearchKit)開發行動應用程式來幫助使用者增加運動依從性。此套件內建許多模組供開發者與研究人員使用，如聲明宣告與問卷模組等，讓整個研究、開發流程更為快速有. 政治大. 效。本研究流程遵行設計科學方法論來創造一個設計實體，即為本研究開發的. 立. 應用程式「Active Track」。. ‧ 國. 學. 在設計階段，本研究採用了「說服設計準則」中的「Tailoring」與「Reminder」設計方法，意圖強化、形塑甚或改變使用者對於目標行為的態度，. ‧. 透過本應用程式來激勵使用者改變自身行為，達到目標設定。目前市面上許多. sit. y. Nat. 健康管理的應用程式皆已證實個性與說服科技之間的相關性，因此本研究採用. al. er. io. 了 MBTI 適性分析工具並設計出相對應的激勵文字訊息，期望透過此設計實體. n. 來協助使用者增加運動依從性。. Ch. engchi. i n U. v. 在第一階段的實驗評估，我們於 166 個下載人次當中篩選出 87 個有效樣本來比較樣本之間的表現，其中有 54 人有接收激勵文字訊息，33 人則無。平均而言，那些有收到激勵文字訊息的受測者，完成率較沒有收到激勵文字訊息的受試者高出百分之十五。然而在訊息類型與個性是否相符的比較實驗當中，訊息符合與不符合使用者個性的結果之間並無顯著差異。本研究之結果僅顯示出透過激勵文字訊息可以有效督促使用者完成運動目標。在第二階段的實驗中加入了「訊息重複性」、「回饋機制」因子來改良應用程式，研究結果顯示訊息的重複整體而言可以提升百分之二十三的完成率，其中適性結果為「理性」的受測者則有百分之二十七的提升，然而回饋機制設計在本研究並無統計顯著。. 2.

(3) 在兩階段的設計循環下，本研究證實透過 Active Track 重複地傳送激勵文字訊息可以協助使用者增加運動依從性，進而降低罹患疾病的風險，研究結果對於未來說服系統之開發以及其他醫療領域提升依從性之相關研究也提供了良好的參考價值。. 關鍵字: 說服準則，設計科學，ResearchKit，MBTI 適性測驗. Abstract. 政治大 Nowadays, most people have 立 experienced uncomfortable feelings but diagnostic. Adherence to an exercise schedule is valuable for health management.. ‧ 國. 學. data are normal. The phenomenon is called ‘sub-health’ condition, which is a state between health and disease. People are likely to experience discomfort if their. ‧. working environment is stressful and their lifestyle is unhealthy. Therefore, a long. sit. y. Nat. period of commitment to adhere to physical activity programs is beneficial for. n. al. er. io. people’s health. Although people would benefit from support to increase exercise. i n U. v. compliance, adherence to physical activity plans is often very low. To address these. Ch. engchi. shortcomings, this paper introduces a low-cost method–an iOS application developed using Apple Inc.’s ResearchKit–to help people adhere to their physical activity plans. ResearchKit provides various modules such as consent declaration and survey task for helping researchers create a research app more efficiently. We applied design science methodology to create a design artifact, namely Active Track. By including the “Tailoring” and the “Reminder” persuasive principles in Active Track to develop, strengthen, or change attitudes or behaviors, the design artifact can act as support instruments that stimulate and encourage users to comply with target behavior. Because studies of health-promotion apps have identified the. 3.

(4) correlations between personality and persuasive technology, we used the Myer– Briggs Type Indicator personality assessment to design motivational messages for each type of personality as a text reminder in Active Track. In the first evaluation stage of Active Track, we identified 87 valid participants (54 with motivational messages and 33 without motivational messages) from 166 downloads for performance comparison. On average, the completion rates of participants who were presented with motivational messages were 15% higher, but the difference in message matching experiment was not significant. The results. 政治大 plans by providing users with motivated messages. Therefore, we implemented 立 demonstrated that our design approach is able to improve adherence on exercise. repetitiveness and feedback intervention in a further design iteration and evaluated. ‧ 國. 學. the improvement in Active Track by using these two new factors. The results. ‧. showed that the repetitiveness factor enhanced the completion rate by approximately. sit. y. Nat. 23%; in particular, participants who were identified as having Thinking-type MBTI. io. er. personalities exhibited an improvement of approximately 27% due to repetitive messages. However, feedback information had no significant effect on adherence.. al. n. v i n In summary, the findings C of this study confirmed h e n g c h i U that Active Track can help. individuals to improve their exercise adherence through repetitive motivational. messages, reduce the risk of diseases, and provide useful insights for the future development of persuasive systems and studies into adherence enhancement for health care.. Keywords: ResearchKit, Persuasive principles, Design science, Myer-Briggs Type Indicator. 4.

(5) Table of contents TABLE OF CONTENTS ............................................................................................. I LIST OF TABLES .................................................................................................... III LIST OF FIGURES .................................................................................................. IV CHAPTER1: INTRODUCTION ................................................................................ 1 CHAPTER2: LITERATURE REVIEW ...................................................................... 4 2.1 SELF-MANAGEMENT EFFECTS ON MHEALTH APPS ............................................... 4 2.2 PERSUASIVE PRINCIPLES PRESENT IN MHEALTH APPS .......................................... 5 2.3 PERSONALITY TRAIT AND PERSUASIVE EFFECTS .................................................. 6 2.4 MOBILE PHONE TEXT MESSAGING EFFECTS ON ADHERENCE ................................ 6. 政治大立................................................................................... 9. 2.5 MESSAGE DESIGN AND PERSONALITY ASSESSMENT TOOL ..................................... 7 2.6 MESSAGE DEVELOPMENT. CHAPTER3: RESEARCH APPROACH .................................................................. 11. ‧ 國. 學. 3.1 DESIGN SCIENCE ............................................................................................. 11 3.2 ARTIFACT ....................................................................................................... 13. ‧. 3.3 EXPERIMENTAL DESIGN................................................................................... 15. Nat. y. sit. er. io. 3.3.1 Participants .............................................................................................. 15 3.3.2 Personality assessment ............................................................................... 16 3.3.3 Message design ......................................................................................... 16 3.3.4 Test Design and Hypothesis........................................................................ 17 3.4 ANALYSIS AND ADJUSTMENT ............................................................................ 19. al. n. v i n C hREFINEMENT ................................................. CHAPTER 4. EVALUATION AND 20 engchi U 4.1 ANALYSIS IN THE FIRST DESIGN CYCLE ............................................................. 20 4.2 IMPROVEMENT POSSIBILITIES .......................................................................... 22 4.2.1 The impact of repetitiveness ....................................................................... 22 4.2.2 The impact of feedback .............................................................................. 23 4.3 DESIGN ITERATION: EXPERIMENTAL DESIGN IN CYCLE TWO .............................. 24 4.3.1 Design for Repetitiveness factor .................................................................. 25 4.3.2 Design for Feedback factor ........................................................................ 25 CHAPTER 5 ANALYSIS RESULTS ........................................................................ 26 5.1 Analysis in the second design cycle ................................................................ 26 5.2 Summary of Results and Discussion .............................................................. 29 CHAPTER 6 CONCLUSION ................................................................................... 32 6.1 Summary..................................................................................................... 32 6.2 Limitation and future work ........................................................................... 32 6.3 Contribution ................................................................................................ 34. i.

(6) REFERENCES ........................................................................................................ 35 APPENDIX A .......................................................................................................... 40 APPENDIX B .......................................................................................................... 45 APPENDIX C .......................................................................................................... 49. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. ii. i n U. v.

(7) LIST OF TABLES TABLE 2.1. FOUR DIMENSIONS OF THE MBTI ASSESSMENT ............................................ 8 TABLE 3.1. RESEARCHKIT MODULES USED IN THIS STUDY .......................................... 12 TABLE 3.2. SUMMARY OF VALID PARTICIPANTS IN ACTIVE TRACK ............................... 16 TABLE 4.1. SUMMARY OF INDEPENDENT SAMPLE T-TEST OF USERS’ COMPLETION RATE (IN PERCENT) BASED ON WHETHER USERS RECEIVED MOTIVATIONAL MESSAGES ........ 20 TABLE 4.2. SUMMARY OF INDEPENDENT SAMPLE T-TEST OF USERS’ COMPLETION RATE (IN PERCENT) BASED ON THE MBTI PERSONALITY TYPES ............................................. 21 TABLE 4.3. SUMMARY OF INDEPENDENT SAMPLE T-TEST OF USERS’ COMPLETION RATE (IN PERCENT) BASED ON WHETHER THE MBTI PERSONALITY TYPES MATCHED MOTIVATIONAL MESSAGES ........................................................................................... 21. 政治大. TABLE 4.4. SUMMARY OF INDEPENDENT SAMPLE T-TEST OF USERS’ COMPLETION RATE (IN PERCENT) BASED ON WHETHER “SPECIFIC” MBTI PERSONALITY TYPES MATCHED MOTIVATIONAL MESSAGES ............................................................................................ 22. 立. ‧ 國. 學. TABLE 5.1. SUMMARY OF INDEPENDENT SAMPLE T-TEST OF USERS’ COMPLETION RATE (IN PERCENT) BASED ON DIFFERENT “REPETITIVENESS” FACTORS ................................ 26. ‧. TABLE 5.2. SUMMARY OF INDEPENDENT SAMPLE T-TEST OF USERS’ COMPLETION RATE (IN PERCENT) BASED ON THEY RECEIVED MOTIVATIONAL MESSAGES AND DIFFERENT “REPETITIVINESS” FACTORS.......................................................................................... 27. al. er. io. sit. y. Nat. TABLE 5.3. SUMMARY OF INDEPENDENT SAMPLE T-TEST OF USERS’ COMPLETION RATE (IN PERCENT) BASED ON PERSONALITY TYPES AND DIFFERENT “REPETITIVENESS” FACTORS ....................................................................................................................... 27. n. TABLE 5.4. SUMMARY OF INDEPENDENT SAMPLE T-TEST OF USERS’ COMPLETION RATE (IN PERCENT) BASED ON DIFFERENT “FEEDBACK” FACTORS ......................................... 28. Ch. engchi. i n U. v. TABLE 5.5 SUMMARY OF INDEPENDENT SAMPLE T-TEST OF USERS’ COMPLETION RATE (IN PERCENT) BASED ON WHETHER THEY RECEIVED MOTIVATIONAL MESSAGES AND DIFFERENT “FEEDBACK” FACTORS ................................................................................ 28 TABLE 5.6. SUMMARY OF INDEPENDENT SAMPLE T-TEST OF USERS’ COMPLETION RATE (IN PERCENT) BASED ON PERSONALITY TYPES AND DIFFERENT “FEEDBACK” FACTORS . 28 TABLE 5.7. SUMMARY OF THE HYPOTHESES AND REFINEMENT TESTING RESULTS ......... 29 TABLE B1. THE MBTI’S SCORING SHEET ...................................................................... 48. iii.

(8) LIST OF FIGURES FIGURE 3.1. DESIGN SCIENCE RESEARCH METHODOLOGY ........................................... 11 FIGURE 3.2. OPERATING PROCEDURES OF ACTIVE TRACK ............................................ 14 FIGURE 3.3. TEST DESIGN PROCEDURE IN APPLICATION ................................................ 19 FIGURE 4.1. REDESIGN STRUCTURE OF ACTIVE TRACK. .............................................. 24 FIGURE A1. EXAMPLE OF CONSENT DOCUMENT (ENGLISH / CHINESE). ....................... 40 FIGURE A2. EXAMPLE OF SURVEY QUESTIONS (ENGLISH / CHINESE). ......................... 40 FIGURE A3. MAIN PAGE OF ACTIVE TRACK (ENGLISH / CHINESE)............................... 41 FIGURE A4. ADD NEW PAGE (ENGLISH / CHINESE). .................................................... 41 FIGURE A5. SPORT LIST PAGE (ENGLISH / CHINESE). .................................................. 42. 政治大. FIGURE A6. SHORT SURVEY FOR SATISFACTION (ENGLISH / CHINESE). ....................... 42 FIGURE A7. MESSAGES PRESENTED ON SPORT LIST (ENGLISH / CHINESE). ................. 43. 立. FIGURE A8. MESSAGES PRESENTED ON MAIN PAGE (ENGLISH / CHINESE). ................. 43. ‧. ‧ 國. 學. FIGURE A9. FEEDBACK INFORMATION (ENGLISH / CHINESE). ..................................... 44. n. er. io. sit. y. Nat. al. Ch. engchi. iv. i n U. v.

(9) Chapter1: Introduction In the modern society, the average life span increased drastically during the twentieth century. Management in healthcare has always been considered a critical issue. In recent years, there is a clear phenomenon of people experiencing discomfort despite diagnostic data from medical instruments being normal. According to the definition provided by the World Health Organization (WHO), this situation, between health and illness, is called a “sub-health” condition, which is also called “gray state of health”. As shown by a 2013 survey conducted by the Public Opinion Programme (POP) of the University of Hong Kong (HKU), 97% of 535 respondents reported having experienced at least one sub-health symptom.. 政治大 showed that 76% of white-collar workers in China were in sub-healthy condition. 立 The main causes of the phenomenon are unhealthy lifestyles, such as overtime late Another statistics of sub-health issue from “All-China Women’s Federation” (2012). ‧ 國. 學. at night, eating irregular meals, lack of exercise, working under high pressure environment and so forth. Long-term sub-health status will cause chronic failure of. ‧. bodily organs or systems, weakened immune system, gradually forming hidden diseases. Normal patterns of lifestyles such as adequate sleep and regular exercises. y. Nat. sit. are basic solutions to help people get rid of sub-healthy state. The research results of. al. er. io. the POP (2013) also indicated that respondents who exercised less frequently had a. n. higher frequency of sub-health symptoms than those who exercised on a regular. Ch. i n U. v. basis. This study will focus on how to increase exercise adherence for ordinary. engchi. people to reduce the risk of suffering from illness.. With the advance of science and technology, the growing number of applications are developed for self-managed health projects. The concept of mobile health (mHealth) is increasingly popular, and can improve health conditions and healthcare environments at a low cost through mobile technology (Martin, 2012). According to Pew Research Center, in 2015, nearly two-thirds of American adults are smartphone owners, and one-fifth of them utilize mobile apps to track or manage health (Fox & Duggan, 2012). A wide variety of innovative functions have been developed to help people adhere to diets or exercise regimens. These applications provide various functions and graphical interfaces to assist users in recording data 1.

(10) and keeping tracking of daily reports. Thus far, most health-care applications have been designed for healthcare professionals rather than patients (Mosa, Yoo & Sheets, 2012). From the viewpoint of preventive medicine, people require rigorous and long-term health control to adhere to a regular daily routine. Currently, literatures reveal no evidence to validate the relationship between the effectiveness of mHealth applications and adherence to target activities (Marcano Belisario, Huckvale, Greenfield, Car & Gunn, 2013; Cui, Wu, Mao & Nie, 2016), and few applications have been devoted to strengthening the motivation effect to help people improve their performance in self-control planning.. 治政 several researchers have studied the relationship大between personality and 立 meta-analysis by Barrick and Mount (1991) performance. After a substantial In order to understand how to increase motivation for better self-control,. 學. ‧ 國. identified personality as an effective indicator of job performance, the research topic became more popular than ever. However, only few researchers have studied the. ‧. impact of personality on health-managed performance. Accordingly, this research aims to take the Myers-Briggs Type Indicator (MBTI) instrument (Briggs-Myers &. y. Nat. sit. Briggs, 1985) as the basis to analyze personality and evaluate the effects of. er. io. motivation on self-managed plans through a mobile application. During the design process, this research utilizes design science research methodology described by. n. al. Ch. i n U. v. Peffers, Tuunanen, Rothenberger and Chatterjee (2007) to construct an iOS mobile. engchi. application, “Active Track”. A discipline oriented design science plays an important role to create successful and innovative artifacts (Peffers et al., 2007). In 2015, Apple Inc. introduced ResearchKit, an open source framework that helps developers and researchers create medical applications that are more convenient and powerful. Please refer to Apple Developer website (https://developer.apple.com/researchkit/). The ResearchKit framework offers a variety of customizable modules, including a survey engine, visual consent flow, and active tasks. We will use the ResearchKit framework to design our iOS mobile application called “Active Track.”. Most self-managed applications aim to support patients in improving compliance with physical activity and medication adherence. By including 2.

(11) persuasive principles designed to strengthen, change, or develop attitudes or behavior, health applications can act as support instruments that encourage and motivate users to achieve target behavior (Fogg, 2003). While designing our app, we adopt the “Tailoring” persuasive principle (Harjumaa & Oinas-Kukkonen, 2009) to design appropriate text messages based on different personality traits and use the “Reminder” persuasive principle defined by Harjumaa and Oinas-Kukkonen to send push notifications with motivating messages to participants.. The specific research objective is to examine the completion rate that indicates the improvement in motivation based on the different push notifications sent to. 治政大 experiments to justify if notified messages match personality types. We will conduct 立 this association. people with different personality types. We expect the completion rate to be higher. ‧ 國. 學 ‧. The contribution of this study is the design artifact—Active Track. Not only can the app help people improve their exercise adherence but it also brings useful. Nat. sit. y. information for advanced application development and future studies. Developers. io. al. n. findings of this research.. er. are able to develop more effective and efficient health care apps based on the. Ch. engchi. 3. i n U. v.

(12) Chapter2: Literature Review Since our mHealth app combined personality traits and persuasive text reminders to enhance the adherence to exercise plans, we first review literatures relating with selfmanagement effects on mHealth apps and the summary of persuasive principles used in mHealth apps, followed by literatures relating with personality trait and persuasive effects, and mobile phone text messaging effects on adherence. At last, studies regarding message design, personality assessment tool and message development would be reviewed. 2.1 Self-management effects on mHealth apps. 政治大. The number of health care professionals who use smartphones to facilitate their work has increased dramatically in recent years. Health care technology for mobile. 立. devices has developed rapidly in tandem with this increase. In a systematic review. ‧ 國. 學. of health care apps, Mosa et al. (2012) examined the functionalities of 83 smartphone-based health care apps and classified them into three categories, namely. ‧. those designed for health care professionals, medical or nursing students, and patients. Most of the apps were designed for health care professionals, whereas only. Nat. sit. y. 15 aided patients with disease management. The purpose of patient-oriented apps is. io. er. to improve patients’ quality of life and help them self-manage their illness. In longterm health management, health care professionals must work with patients for. n. al. Ch. i n U. v. improvements in treatment to be possible. Thus, patients’ ability to engage in self-. engchi. control should be reinforced, and apps have been proposed as an effective means of doing this.. Marcano Belisario et al. (2013) assessed the feasibility and impact of using smart technologies, namely tablet and smartphone apps, to promote the self-management of people with asthma. Their results showed no statistical difference in intervention adherence between those who did and did not use the apps, which was not sufficient to support the use of smart technology for delivering asthma-related self-managed applications (Marcano Belisario et al., 2013). Cui et al. (2016) conducted a metaanalysis and systematic review to investigate the effect of self-management apps on patients with type 2 diabetes mellitus (T2DM). They found that mHealth apps 4.

(13) provided only moderate advantages for these patients in managing their health, and that further research with longer follow-up is required to elucidate the relationship between mHealth apps and self-management. 2.2 Persuasive principles present in mHealth apps Interactive information technology designed for the redirection of individuals’ attitudes or behaviors is known as persuasive technology (Fogg, 2003). Harjumaa and Oinas-Kukkonen (2009) discussed the optimal means of designing and evaluating persuasive systems and what functionalities and content mHealth software should include. They proposed four categories of design content, namely the primary task, dialogue, system credibility, and social support, as well as 28. 政治大. design guidelines (Harjumaa & Oinas-Kukkonen, 2009). Those persuasive design rules may become particularly effective in stimulating and persuading users to keep. 立. up with their personal target, and more helpful for developing actual systems.. ‧ 國. 學 ‧. Geuens et al. (2016) analyzed the number and the type of persuasive principles applied to present health apps for chronic arthritis (CA) patients. Rare persuasive. sit. y. Nat. techniques were adopted in engaging patients to facilitate their illness management in digital way, and the research showed that only 5.8 out of 37 persuasive principles. io. n. al. er. were used in the selected apps (Geuens et al., 2016). Moreover, a past review of. i n U. v. mHealth apps that applied persuasive technology to improve physical exercises. Ch. engchi. indicated that users were persuaded and became more involved in disease control or health management by information systems (Matthews, Win, Oinas-Kukkonen & Freeman, 2016). Another research that discussed the effects of health care interventions and influence on compliance by persuasive technology has proved that persuasive technology especially the “Tailoring” principle in the primary task support category has the greatest effect in every health care domain and influences the adherence of an intervention (Hollander, 2016).. The “Tailoring” principle is that if the system provides information which tailored to users’ personalities, interests, potential requirements, usage context, or any factor related to them, the system may be more persuasive (Harjumaa & Oinas-Kukkonen,. 5.

(14) 2009). Matthews et al. (2016) also indicated that the performance of high motivation level for physical activities often presented with persuasive technologies which were dependent on each other. The principles in the primary task support category such as tailoring, tunneling, and personalization are often demonstrated with the principles in the dialogue support category such as suggestion, reminder, and liking (Matthews et al., 2016). Therefore, appropriate persuasive principles are needed to better improve adherence to self-control plans.. 2.3 Personality trait and persuasive effects Past literature reviews of mobile health apps identified the correlations between. 政治大 Halko and Kientz (2010) indicated that there was moderate evidence to validate that 立 using the Big Five personality traits as a persuasive strategy can better conform to personality and persuasive technologies (Halko and Kientz, 2010). For example,. ‧ 國. 學. users’ requirements. Shuttleworth (2015) also found that users had high scored in neuroticism personality trait were more willing to share the social support persuasive. ‧. design principle than the group who scored low. In addition, Sofia, Marianna, George and Panos (2016) empirically supported that using an influential strategy in the. y. Nat. sit. application may positively transfer users’ attitude toward a recommended item, and. er. io. the persuasive effects varied based on diverse personality traits of users.. al. n. v i n C hNassiriyar (2014) proposed Haghshenas, Shahbazi, Zadeh and a similar research using U i e h n c Myers-Briggs Type Indicator (MBTI), gthe another personality assessment tool, namely to distinguish participants’ preferences, and validated the relationship between personality traits and persuasive effects on the audience by using the media. The results showed that the audience was more satisfied with the media content, which met their demands when the content was designed according to their personality temperaments.. 2.4 Mobile phone text messaging effects on adherence Car, Gurol-Urganci, Jongh and Atun (2013) used message reminders by mobile phone to investigate the attendance rate of outpatient appointments. There was low to moderate quality evidence that mobile phone notifications by text messages 6.

(15) increased attendance rate at outpatient appointments, and the attendance rate was higher than those who received postal reminders or no reminders; text reminders on mobile phone can motivate users to achieve physical activities (Car et al., 2013). In the field of health care, strong evidences showed that mobile phone text messaging intervention improves adherence in certain disease control. Finitsis, Pellowski, and Johnson (2014) proved that text messaging supported antiretroviral therapy adherence. A previous study with high-quality evidence identified that compare with standard care, weekly text messaging was helpful for improving adherence to antiretroviral therapy, and mobile phone text messaging at weekly interval was an effective way to enhance HIV viral load suppression (Horvath, Azman, Kennedy & Rutherford, 2012). A recent study has proved that mobile phone. 政治大 short experimental period and self-reported biases on however, because of the 立. text messaging can improve adherence rate in chronic disease from 50% to 67.8%;. ‧ 國. 學. medication adherence, the further research for long-term effects is necessary (Thakkar et al., 2016).. ‧ y. Nat. Furthermore, Goodwin and Friedman (2006) have identified associations between. sit. personality and conscientiousness, including adherence to health, recommended. er. io. treatments and internalization of healthy social norms. Thus, we expect that a mobile. al. n. v i n C heffectively improve participants’ personality traits could e n g c h i U adherence to exercise plans.. health app presenting notifications with motivational text messages based on. 2.5 Message design and personality assessment tool Referring to the persuasive effect of message, in the marketing field, a research discussing the “Tailoring” persuasive appeals to recipients' personality characteristics showed positive results that participants considered the content of advertisements emphasizing a specific motivational feature to be efficacious when the feature was in accordance with their personality preferences (Hirsh, Kang & Bodenhausen, 2012).. In addition, vast literatures discussed the impact of message types on advertising effectiveness. Different appeals such as “Thinking” and “Feeling” content of. 7.

(16) advertisement had different communicated responses; “Thinking” advertisements with more reliable information were more believable than “Feeling” advertisements, but there was no differential impact on purchasing decisions (Zielske, 1982; Golden & Johnson, 1983).. Mayer and Tormala (2010) have also identified that the effect of identical messages varied according to thoughts or feelings from their source. Participants were more willing to process the matched messages than mismatched messages, and they put the feeling down to the messages’ effects of persuasion, and besides, when recipients had cognitive orientations, they were more susceptible to persuasive messages framed in terms of the sources’ thoughts or feelings (Mayer & Tormala, 2010).. 治政大 we considered the MyersTherefore, among various personality assessment tools, 立 Briggs Type Indicator (MBTI) to be appropriate personality assessment. The MBTI is ‧ 國. 學. a widely used and trusted tool, which is a self-report survey questions designed to identify psychological preferences in how people perceive the information and make. ‧. their own decisions. There are four pairs of preferences to identify personality type, namely Extraversion and Introversion (E, I), Sensing and Intuition (S, N), Thinking. y. Nat. er. io. sit. and Feeling (T, F), and Judging and Perceiving (J, P) (see Table 2-1). Table 2.1 Four dimensions of the MBTI assessment (Myers & McCaulley,. n. al. Ch. i n U. v. 1985; Myers, McCaulley, Quenk, & Hammer, 1998/2003) Preferences of personality. e nExplanation g c h i of preferences. Extraversion and Introversion (E,. Attitudes people use to direct their. I). energy.. Sensing and Intuition (S, N). How people perceive and interpret new information.. Thinking and Feeling (T, F). Make rational decisions, based on the. data. received. from. their. information-gathering functions. Judging and Perceiving (J, P). How you like to live your outer life.. 8.

(17) We only selected survey questions about the “Thinking and Feeling” dimension, which exactly conformed to our research context. Based on the literatures above, we could design suitable messages to match each personality type as our control factor to examine the impact of matched messages on motivational level.. The research was to investigate how people process information and decide whether to finish their scheduling items. Vast literatures examining elderly people's participation in exercise classes defined adherence as “completion” and lack of compliance as “termination of participation”; they were measured by percentage of or number of attendances (Williams and Lord, 1995; Estabrooks and Carron, 1999; Tu, Stump, Damush & Clark, 2004; Sullivan-Marx et al., 2011; Hawley-Hague et al.,. 治政大 our design artifact. represented for performance of motivational effects through 立. 2014). Therefore, we applied “completion rate” to measure users’ adherence, which. ‧ 國. 學. Emotional people take the effectiveness of their decisions into account based on others opinions in order to satisfy other people, and are motivated to be appreciated;. ‧. intellectual people often make decisions after deep consideration, naturally seek truth from facts and criticize, and are stimulated by their desire to be successful. Nat. sit. y. (Haghshenas et al., 2014). An empirical evidence that people with “Thinking and. al. er. io. Feeling” preference reacted differently to and prefer different types of psychological. n. therapies (Graff, 1976; Carskadon, 1979; Jinkerson, Masilla & Hawkins 2015).. Ch. engchi. i n U. v. Consequently, our study will take the MBTI assessment to identify the two abovementioned personality traits and design corresponding persuasive text messages in the design artifact; we expect that the design may enhance users’ completion rate and improve the effects of self-management on exercise plans.. 2.6 Message development Based on Kim and Benbasat (2006), they designed and tested arguments according to Toulmin’s model of argumentation (Toulmin, 2003) and proposed that the structure of an argument should be its “claim” plus “data” and “backing”, which contributes to the type of information that can be considered most trustworthy. Namely, the main content of a message (claim) should include some facts (data) to 9.

(18) support the content and indicate reliable source (backing) (Toulmin, 2003). We applied the above-mentioned approach to design “logical” messages for persuading “Thinking” type people.. In contrast, individuals tend to change their minds to meet their emotions, because feelings are seen as evidence by people (Clore & Gasper, 2000). Without being aware of having felt or seen an object, emotional stimuli can have an effect on individuals’ judgment (Murphy & Zajonc, 1993). Therefore, we designed “emotional” messages for “Feeling” type people. The experimental design for personalities and messages will be discussed in the next chapter.. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 10. i n U. v.

(19) Chapter3: Research Approach 3.1 Design Science The current research follows the design science research methodology described by Peffers et al. (2007). Our main goal was to design and develop an application as an artifact to improve exercise adherence in people to reduce their risk of disease. According to its definition, there are six stages during the design process including “problem identification and motivation,” “objectives of a solution,” “design and development,” “demonstration,” “evaluation,” and “communication” (Peffers et al., 2007)(see Figure 3-1). We continuously refined the functions during the process from development to evaluation to ensure that the instantiation we created to improve. 政治大. adherence is the solution to our research objective.. 立. Evaluation. Communication. y. Demonstration. n. al. er. io. sit. Design and development. ‧. ‧ 國. 學. Objectives of a solution. Nat. Problem identification and motivation. Ch. engchi. i n U. v. Figure 3.1 Design Science Research Methodology (Following Peffers et al., 2007) In the first stage, we identified that people with sub-healthy conditions usually lack regular eating habits, exercise, and a healthy lifestyle, and may also have a high potential for suffering from unknown diseases. For the above-mentioned problems, our research objective and solution was to enhance their motivational level to adhere to schedules with our design artifact. With the popularity of mobile health, we designed a mobile application for our target audience to manage their health condition, hoping to cater to users’ habits of using mobile phones and meeting their expectations.. 11.

(20) At the design stage, we developed an iOS application using a new tool–ResearchKit, which is an open-source framework introduced by Apple in 2015. For medical research, ResearchKit framework comes with some predefined sections that are commonly used in research procedure, enabling our iOS app to become a powerful instrument. Through a variety of customizable modules such as informed consent, surveys, and active tasks (see Table 3-1), we could make our app more convenient to use in the research process. Because the research uses the MBTI questionnaire to determine personality types, we mainly applied the “Inform Consent” and “Survey” modules in ResearchKit to develop customizable MBTI questions and designed corresponding messages to randomly assign to each participant.. 政治大 Section/Task. Table 3.1 ResearchKit Modules Used in this Study (O: Applied / X: Unapplied) Overview. O. 學. Data gathering Privacy Time commitment. Nat. io. Tasks Withdrawal. n. al. Ch. review e nConsent h gc i U. O O O. sit. Surveys. O. ‧. Data use. Review. Survey. Artifact. y. 立 Visual. ‧ 國. Inform Consent. Step. er. Module. v ni. O O O O. Agreement. O. Name entry. O. Signature. O. Consent sharing. Consent sharing. X. Instruction. Instruction. O. Question/Form. Single/Multiple questions. O. Answer format. Scale (vertical/horizontal). X. Boolen. X. Value picker. O. Image. X. Text choice (Sindle/Multiple). O. 12.

(21) Active Task. Motor activities Fitness Cognition. 立. Numeric. O. Date/Time of the day. X. Location. X. Text answer(limited/unlimited). O. Email. X. Validated text answer. X. Gait and balance. X. Tapping speed. X. Fitness. X. Timed walk. X. Spatial memory. X. serial addition test 政 Paced治大 Tower of hanoi. X. Reaction time. X. 學. ‧ 國. X. Sustained phonation. X. Audio. Tone audiometry. X. Hole Peg. Hand dexterity. X. Nat. y. ‧. Voice. sit. In the demonstration, evaluation, and communication stages, the initial prototype. er. io. application was published on the App Store in addition to storing and monitoring user. al. n. v i n C hto implement version Connect” platform for developers e n g c h i U control. Therefore, we were data through the Firebase online database. The App Store also provides the “iTunes. able to modify functions from the monitored data and user feedback on the App Store in order to update new versions of the application until the final artifact.. 3.2 Artifact The purpose of the application is to help participants schedule personal exercise plans and ensure that they abide by them. The artifact, called “Active Track,” was designed and developed in two selectable languages–English and Chinese– and is available on the App Store (see screenshots in Appendix A). The principle idea of Active Track is that it is an exercise reminder app that provides a basic planning list and notifications through designed messages to remind users. 13.

(22) View declaration of the study and sign agreement. MBTI assessment and demographic survey. Main Page of Active Track. Scheduling exercise plan. Users receive designed messages. Self-reported completion rate. Weekly satisfaction survey. 政治大. Figure 3.2 Operating procedures of Active Track. 立. ‧ 國. 學. The operating procedures of the application are shown in Figure 3-2. At the onset, participants may be asked for their signature after viewing the research consent document and finishing a series of MBTI survey questions built with the ResearchKit. ‧. framework. The ResearchKit tool provides consent module for us to implement. y. Nat. complicated consent flow. Through the predefined sections that described in the. io. sit. informed consent form (see Table 3-1), users can easily understand the purpose of the. n. al. er. research and sign their declaration to participate. In addition, the survey module. i n U. v. includes several questionnaire types in order that we can utilize it to implement a. Ch. engchi. customized survey –demographic questions and the MBTI test (also see Table 3-1). We were able to easily combine the survey and main application as one artifact rather than have separate paper work as in the past. After providing consent and finishing the survey, users can access the main page of the application and start scheduling an exercise plan. According to the MBTI test results, i.e., different users’ personality types, the system pushes notifications with our designed messages to remind users at scheduled times. The messages are designed for each type of personality and sent to corresponding personalities based on the experimental design in the next section. In order to ensure that the participants receive notifications, each notification can be deferred for five minutes and three times at most, in addition to repeating weekly. Further, we also created a weekly short survey. 14.

(23) to measure participants’ satisfaction with the effects of motivation (see Figure A6 in Appendix A). For evaluating the effectiveness of adherence to exercise, the application also records all scheduled times and reminds users to confirm whether they completed the sport item at the end of the scheduled day. Next, the system automatically calculates the average completion rate of each person and all data are stored in the Firebase online database, which will be analyzed in a later chapter.. 3.3 Experimental Design 3.3.1 Participants. 立. 政治大. In order to validate Active Track’s performance, the application is downloadable for. ‧ 國. 學. free in the App Store and is available to anyone who owns iPhone devices. During the experiment period, Active Track was promoted in Taiwan (most participants. ‧. were Asian). Apart from the free download, the research team also cooperated with coaches in gyms or personal trainers to recruit participants. The coaches only. y. Nat. sit. promoted Active Track to their students or a public audience, and were not allowed. er. al. n. by coaches.. io. to supervise how participants used Active Track with a view to reduce interventions. Ch. engchi. i n U. v. We finally recruited 166 participants during the two-month experiment period; however, in order to avoid bias, we discarded the data of those with data that had 0 and 1 instances of usage. The number 0 or 1 indicates that users did not finish the survey questions from the onset or only used the app once in the first trial, which may result in invalid figures in the analysis. After filtering the data, the final response rate was 52.4% (87 valid participants shown in Table 3-2).. 15.

(24) Table 3.2 Summary of valid participants in Active Track Thinking. Feeling. Gender. Type. Type. Male/Female. 16. 7. 6. 17. 23. 22. 9. 13. 18. 31. 26. 7. 13. 20. 33. 55. 87. Total Matched Group Mismatched Group Control Group Total. 64. 學. ‧ 國. 立. 23 32 政治大. 3.3.2 Personality assessment. The study used the MBTI as the basis to identify the participants’ personality. Our. ‧. research objective was to investigate the motivational effect of messages based on. sit. y. Nat. different personalities, and to understand how participants make decisions when perceiving different types of messages and adhere to their exercise schedule.. er. io. Therefore, we selected individuals with the “Thinking and Feeling” preference as. al. n. v i n C honset of the application included in our questionnaire at the e n g c h i U (see Appendix B). The our research target. All MBTI questions regarding this personality type were. system calculates the total score of each participant to determine their personality type based on the MBTI’s scoring method (also see Appendix B).. 3.3.3 Message design According to Thinking and Feeling personality preferences, we correspondingly designed two types of messages–logical and emotional–and used them as notifications to remind users. The approach mentioned in the literature review to design logical messages with the structure of claim, data, and backing, allows us to provide facts to support why users. 16.

(25) should agree with the claim that these facts are “data.” We then provide “backing” by indicating the source or credibility of the data. For example, “Don’t forget to exercise today! Research shows that even one session of exercise will enhance positive mood.” Here the claim is “Don’t forget to exercise today,” the data is “even one session of exercise will enhance positive mood,“ and the backing is ”research shows.” However, emotional messages do not provide facts to support the claim. They focus on triggering positive emotions. For example, “It’s time for your exercise! You are doing a fabulous job!”. By following the design rules for both logical and emotional messages, we selected. 治政大 in order to fit in a text Appendix C). These messages are shorter than 128 characters 立 message.. 30 motivational messages for each type of exercise from online forums (see. ‧ 國. 學 ‧. 3.3.4 Test Design and Hypothesis. y. Nat. First, according to the literature review, text messaging is an effective way to improve. sit. adherence to therapies for specific diseases in the field of health care. Our first goal. er. io. was to examine if using our designed messages to remind participants of their sport. al. n. v i n C h messages–asUthe foundation and formulated two main categories–with and without engchi. schedule was an effective method to enhance their completion rate. As a result, we set our first hypothesis as follows:. Hypothesis H.1: People are more motivated when presented with notification messages than when they do not receive any type of message. Second, in the category with notification messages, we planned to further investigate the effects of message types on different personalities. In line with the theory, we hypothesized that messages that are logical, i.e., making suggestions based on facts and scientific evidence, are more effective in motivating people with a Thinking-type personality, and that emotional messages that promote positive values (such as persistence, optimism, etc.) and provide emotional support (such as encouragement,. 17.

(26) inspiration, acknowledgment, etc.) are more effective for people with a Feeling-type personality. We divided participants in this category into two groups, which represent two relations between personality and message. One group was composed of people with matched messages. Namely, logical messages were sent to participants who were identified as having Thinking-type MBTI personalities and emotional messages were sent to users with Feeling-type MBTI personalities. The other group was composed of people with mismatched messages. That is, logical messages were delivered to participants with Feeling-type MBTI personalities and emotional messages were transmitted to users with Thinking-type MBTI personalities. Our aim was to verify. 政治大. whether users would be motivated by messages that match their personalities. 立. Hypothesis H.2 (a): People with a Thinking-type personality are more motivated. ‧ 國. 學. when presented with logical messages than emotional messages.. ‧. Hypothesis H.2 (b): People with a Feeling-type personality are more motivated when presented with emotional messages than logical messages.. sit. y. Nat. er. io. The original group was set as the control group, which included both Thinking- and Feeling-type participants. However, participants in the control group received no. n. al. Ch. i n U. v. messages. As a result, we set three groups at the beginning of the experiment. The. engchi. participants were randomly distributed to the three groups when first registering their personal information in the application (see Figure 3-3).. 18.

(27) Thinking-type Control group. Feeling-type. First Registered. Thinking-type. Logical messages. Feeling-type. Emotional messages. Thinking-type. Emotional messages. Feeling-type. Logical messages. match group. mis-matched group. 政治大 Figure 3.3 立Test design procedure in application. ‧ 國. 學. 3.4 Analysis and adjustment. ‧. In the first design cycle, the results were collected after participants’ continuous use. y. Nat. of the application for a month. Active Track recorded users’ first login time, and. io. sit. therefore, we could compare and analyze completion rates among the three groups,. er. and according to analysis results in phase one, allow us to discover factors that could. al. n. v i n C forth. Based on the results in phasehone, e nwegvalidated c h i Uthe effectiveness of providing further improve adherence or influence the final results such as gender, race, and so. motivational messages and the impact of matching messages on participants’ personalities. If any hypotheses were significant, possible improvements would be added into the second design iteration to refine the design. The research results are valuable for future refinement until the final artifact is completed.. 19.

(28) Chapter 4. Evaluation and Refinement 4.1 Analysis in the first design cycle In the first design cycle, we conducted an independent sample t-test to compare mean completion rates of the two independent groups to determine whether there were significant differences to support our hypotheses. The results are as follows. In order to examine whether motivational messages have an effect on completion rate, an independent sample t-test was conducted to compare percentages in the providing messages and not providing messages conditions. As seen in Table 4.1, there is a. 政治大. significant difference in the percentage of completion rate for the providing motivational messages (M = 66.64, SD = 39.31) and not providing motivational. 立. messages (M = 82.45, SD = 24.09) conditions; t(47) = -2.08, p = .021 < .05. These. ‧ 國. 學. results suggest that providing motivational messages affect the completion rate. Specifically, our results suggest that users who receive motivational messages have a. ‧. better completion rate than those who do not. According to these results, H.1 was supported: People are more motivated when presented with notification messages than. er. io. sit. y. Nat. when they do not receive any type of message.. Table 4.1 Summary of independent sample t-test of users’ completion rate (in. n. al. i n U. v. Percent) based on whether users received motivational messages. Ch. w/, w/o. N. i e n g c hStd.. Motivational. w/o. 33. 66.64. 39.31. Messages. w/. 54. 82.45. 24.09. Mean. Deviation. df. t. 47. -2.08. Sig.(1 -tailed) .021. Further, the independent sample t-test illustrated in Table 4.2 shows that there is a significant difference between people with a “Feeling” personality type and those with a “Thinking” personality type . The completion rate in the “Thinking” (M = 71.66, SD = 35.96) and “Feeling” (M = 89.80, SD = 16.39) personality conditions are significantly different; t(78) = -3.31, p = .001 < .01.. Note: * p < .1, ** p < .05, *** p < .01. 20.

(29) Table 4.2 Summary of independent sample t-test of users’ completion rate (in Percent) based on the MBTI personality types. MBTI. Std.. Type. N. Mean. Thinking. 64. 71.66. 34.24. Feeling. 23. 89.80. 16.39. Deviation. df. t. Sig.(2-tailed). 78. -3.31. .001. Based on the positive results above, we further analyzed the group provided with motivational messages. We conducted an independent sample t-test to examine whether there was statistical evidence for significant differences between the two groups–people received messages that matched or mismatched their personalities. As. 政治大 matched and mismatched groups. Therefore, we conducted an independent sample t立 test for people with specific personality types (see Table 4.4); however, there are no. illustrated in Table 4.3, there is no significant difference (p > .05) between the. ‧ 國. 學. significant differences in each group (p > .05 for Thinking-type, p > .05 for Feelingtype). Specifically, the results suggest that users provided with motivational messages. ‧. that match or mismatch their personality types do not affect their completion rate.. y. Nat. sit. Table 4.3 Summary of independent sample t-test of users’ completion rate (in. Message. Ch. Std.. i n U. e n gDeviation chi. N. Mean. matched. 23. 82.61. 31.07. mismatched. 31. 82.33. 17.81. Matching Personality. al. n. messages. er. io. Percent) based on whether the MBTI personality types matched motivational. Note: * p < .1, ** p < .05, *** p < .01. 21. v. df. t. 33. 0.04. Sig.(1 tailed) .485.

(30) Table 4.4 Summary of independent sample t-test of users’ completion rate (in Percent) based on whether “specific” MBTI personality types matched motivational messages Message Matching. N. Mean. Std. Deviation. Thinking. matched. 16. 77.47. 35.96. Type. mismatched. 22. 80.19. 17.65. Feeling. matched. 7. 94.36. 8.66. Type. mismatched. 9. 87.57. 18.12. Sig.(1 -. df. t. 20. -0.28. .391. 14. 0.91. .189. tailed). The results show that the completion rate in those who received messages were not. 政治大 H.2(a) and H.2(b). Therefore, in the next design cycle, we explored new factors to 立 improve motivation based on H1.. influenced by whether the messages matched their personalities, which cannot support. ‧ 國. 學 ‧. 4.2 Improvement Possibilities. y. Nat. The evaluation results and users’ experience from the first design cycle showed that. sit. other factors can be included in the second design stage to improve the effectiveness. er. io. of the motivational messages. The main suggestions from participants were. al. “repetitiveness of notifications” and “feedback information.” The reason for the. n. v i n often C needed time to prepare h e n g c h i U their. former was that users. schedules and the. notifications only appeared at the scheduled time; therefore, appropriately designing reminders prior to the scheduled time should be taken into consideration to improve completion rate. The reason for the latter was that users did not know which schedule they had already finished, and thus, the “number” of completion rate was meaningless. Therefore, we considered that the following two factors should be incorporated into the second design: repetitiveness and feedback.. 4.2.1 The impact of repetitiveness In fields such as social marketing and human decision making, there is a famous phenomenon called the “mere exposure effect,” wherein people possess a preference. 22.

(31) for an object simply because they are more familiar with it (Zajonc, 1968). Zajonc (1968) conducted several experiments for the well-supported explanatory hypothesis that individuals improve their attitude toward an object that is stimulated by mere repeated exposure. There is significant evidence to support the mere exposure effect. For example, Serenko and Bontis (2011) used journal ranking investigations to test the phenomenon of the exposure effect, and the results showed that the quality of a journal perceived by respondents is influenced by participants’ prior exposure to the journal. Another analysis discussing the popularity of candidates presented by Verhulst, Lodge, and Lavine (2010) showed that familiarity with a candidate is a primary reason to. 治政大 the exposure effect. similar analysis on voting in a song competition and validated 立. determine whether an individual will vote for them. Verrier (2012) also conducted a. ‧ 國. 學. Moreover, Malone (1983) investigated principles for computer desktop design and suggested that a useful way to enhance reminder effectiveness is to display high-. ‧. priority tasks more frequently. Based on the above empirical evidence, we assume that the more users are exposed to repetitive motivational messages, the more likely they. Nat. sit. y. are to have a positive attitude toward their schedule and have a better completion rate design iteration for refining of the artifact.. n. al. 4.2.2 The impact of feedback. Ch. engchi. er. io. (Malone, 1983). Therefore, we decided to add the “Repetitiveness” factor into our. i n U. v. In addition to repetitiveness, we found substantial research that showed that feedback intervention can increase performance for both individuals (Kluger & DeNisi, 1996) and teams (Burgio, Engel, Hawkins, McCormick & Scheve, 1990; Goltz, Citera, Jensen, Favero & Komaki, 1989).. The definition of feedback interventions is an external agent taking some actions to provide people with some aspects of their task performance as information (Kluger & DeNisi, 1996). In our research, we specifically focused on knowledge of progress. 23.

(32) (KP) interventions, where the feedback focuses on the reporting of progress and results without suggestions, incentive, or punishment (Ammons, 1956).. We assume that providing feedback information of users’ progress can enhance their motivational level. However, the current research investigated the impact of feedback in our design format, which is described in later sections; the more complicated or other design approaches toward feedback mechanisms could be discussed in future studies. 4.3 Design iteration: experimental design in cycle two. 治政大users and new participants redesigned the structure and randomly assigned current 立 into two groups for each control factor. When new users first downloaded Active. In order to add the new factors (repetitiveness and feedback) in Active Track, we. ‧ 國. 學. Track or current users updated the application, the system randomly assigned each account to one repetitiveness type (normal / one more message) and one feedback. ‧. type (with / without). The new users had to finish the MBTI assessment to determine their personalities but current users did not repeat it. Based on the original design. y. Nat. n. al. er. io. sit. structure in Figure3.3, we can see the redesign structure in the following Figure 4.1.. Ch. engchi. i n U. v. Figure 4.1 Redesign structure of Active Track. 24.

(33) 4.3.1 Design for Repetitiveness factor According to our new design, we set two repetitiveness types. In the first, no repeat, the system would push only one notification to users at scheduled times. In the second, one-time repeat, the first notification appeared at scheduled times, and besides, we added one more notification 30 minutes in advance to remind users of their plans. Regardless of whether the notification was accompanied by a motivational message, each member in the first three groups was randomly assigned to one repetitiveness type.. 4.3.2 Design for Feedback factor. 政治大 Similarly, we also set two feedback types–providing or not providing feedback 立 information to users. The feedback information contained users’ progress of their. ‧ 國. 學. schedules such as their completion rate, scheduled items, completed and failed times, and reasons for failure. The group with feedback information would receive. ‧. notifications along with their completion rate; moreover, detailed feedback information could be accessed by clicking on the percentage button in the main page. y. Nat. sit. (see Figure A9 in Appendix A). However, the other group without feedback. er. io. information received the original notifications, which did not contain any. al. v i n participants after they continuously weeks. C hused the app for two engchi U n. information about the progress of their schedules. We collected results from all. 25.

(34) Chapter 5 Analysis Results 5.1 Analysis in the second design cycle In the second design cycle, we took the new factors into consideration and conducted an independent sample t-test to compare the mean completion rate of the two groups. The results are as follows. First, we conducted an independent sample t-test based on different “Repetitiveness” factors to examine whether users with one more notification have a higher completion rate. According to the results in Table 5.1, the completion rate in the “one-time repeat”. 政治大. condition (M = 86.12, SD = 20.92) is significantly different from the “no repeat” condition (M = 63.39, SD = 38.34); t(52) = -3.27, p < .001. The results showed that. 立. repetitive messages indeed have an effect on improving motivation.. ‧ 國. 學. Table 5.1 Summary of independent sample t-test of users’ completion rate (in. 37. 63.39. 38.34. 50. 86.12. 20.92. io. al. n. w/. Ch. engchi. t. y. w/o. Deviation. df. 52. sit. Mean. er. N. Nat. Repetitiveness. Std.. w/, w/o. ‧. Percent) based on different “Repetitiveness” factors. i n U. -3.27. Sig.(1tailed) .001. v. Furthermore, in order to find whether the impact of the repetitiveness factor can strengthen the effect on the basis of providing motivational messages to participants, we conducted an independent sample t-test in the groups with motivational messages to compare differences with different “repetitiveness.” The results in Table 5.2 show that people who received repetitive motivational messages (M = 89.28, SD = 15.65) had a better completion rate than those who did not receive repetitive motivational messages (M = 72.51, SD = 30.47); t(29) = -2.37, p = .012 < .05. Specifically, receiving repetitive motivational messages had a combined effect on users’ completion rate.. Note: * p < .1, ** p < .05, *** p < .01. 26.

(35) Table 5.2 Summary of independent sample t-test of users’ completion rate (in Percent). based. on. they. receive. motivational. messages. and. different. “Repetitiveness” factors Std.. Repetitiveness. N. Mean. w/o. 22. 72.51. 30.47. w/. 32. 89.28. 15.65. With Motivational Messages. Deviation. df. t. 29. -2.37. Sig.(1tailed) .012. Moreover, there was a significant difference in the repetitiveness factor among Thinking-type people and the results in Table 5.3 show that Thinking-type people. 政治大 completion rate than those who did not receive repetitive motivational messages (M = 立 57.14, SD = 39.25); t(45) = -3.34, p < .001. Additionally, there was no significant. who received repetitive motivational messages (M = 84.46, SD = 22.89) had a better. ‧ 國. 學. difference in Feeling-type people for the repetitiveness factor (p > .05). Specifically, the “Repetitiveness” factor has an obvious effect on completion rate in the group. motivated when they receive repetitive motivational messages.. Nat. sit. y. ‧. composed of people with a “Thinking” personality and these individuals are more. al. er. io. Table 5.3 Summary of independent sample t-test of users’ completion rate (in. n. Percent) based on personality types and different “Repetitiveness” factors. i n U. C hMean Std. e n gDeviation chi. Repetitiveness. N. Thinking. w/o. 30. 57.14. 39.25. Type. w/. 34. 84.46. 22.89. Feeling. w/o. 7. 90.14. 18.46. Type. w/. 16. 89.65. 16.04. v. Sig.(1-. df. t. 45. -3.34. .001***. 21. 0.06. .475. tailed). Second, we conducted an independent sample t-test based on different “Feedback” factors to examine whether users provided with feedback information have a higher completion rate. The “Feedback” factor had no significant difference (p > .05) for completion rate (Table 5.4). Additionally, in the providing feedback information Note: * p < .1, ** p < .05, *** p < .01. 27.

(36) condition, there were no significant differences in the group provided with motivational messages (p > .05) (Table 5.5). Similarly, there were no significant differences in the providing and not providing feedback information conditions in each personality group as well (p > .05 for “Thinking” type users and p > .05 for “Feeling” type users) (Table 5.6). According to these results, the “Feedback” factor had no effect on completion rate overall. Table 5.4 Summary of independent sample t-test of users’ completion rate (in Percent) based on different “Feedback” factors N. Mean. w/o. 42. 75.75. w/. 45. 立. Std. Deviation. df. 治 85 政 32.21 大 77.11 31.18. t. Sig.(1-tailed). -0.20. .421. 學. ‧ 國. Feedback. w/, w/o. Table 5.5 Summary of independent sample t-test of users’ completion rate (in Percent) based on whether they received motivational messages and different. 23.68. w/. y. 83.37. Sig.(1-. t. tailed). sit. 25. Deviation. df. er. w/o. Std.. a l 29 81.66 24.83 52i v n Ch U engchi. n. Messages. Mean. io. Motivational. N. Nat. With. Feedback. ‧. “Feedback” factors. 0.26. .399. Table 5.6 Summary of independent sample t-test of users’ completion rate (in Percent) based on personality types and different “Feedback” factors Std.. Feedback. N. Mean. Thinking. w/o. 32. 70.79. 34.80. Type. w/. 32. 72.52. 34.21. Feeling. w/o. 10. 91.61. 13.73. Type. w/. 13. 88.40. 18.60. Deviation. 28. df. t. Sig.(1-tailed). 62. -0.20. .421. 21. 0.46. .326.

(37) 5.2 Summary of Results and Discussion Both experiments in the first and second design cycles showed promising results that partially support our design rationale to develop an iOS application using motivational messages and repetitiveness as design factors to achieve higher completion rates. We summarize the results of all experiments in Table 5.7.. Table 5.7 Summary of the hypotheses and refinement testing results Test Group Presenting motivational messages. H.1. H.2(a). H.2(b). Support. -. -. Messages matched. -. personalities. 立-. Thinking-type. Support. Not Not 政Support治Support大 -. -. ‧ 國. Feeling-type. -. -. -. Support. Support. Nat. Not Support Not. Support. Support. sit. y. The findings are summarized below.. -. Not. ‧. personality. Not. -. 學. personality. Repetitiveness Feedback. io. er. Finding 1. Presenting motivational messages can improve completion rate. When presenting motivational messages to users of Active Track, they are more. n. al. i n U. v. motivated than those who do not receive any type of message. The difference in. Ch. engchi. completion rate in the two groups was significant with a p-value less than 0.05 and was approximately 15 percent on average; these results supported H.1. Finding 2. Whether motivational messages match users’ personalities has no impact on completion rate. The results in our first design cycle showed that people presented with matched messages did not perform better than those presented with mismatched messages. We further analyzed the two personality groups to compare differences in the matched and mismatched conditions; however, the results also showed no difference in each group. Thus, H.2 (a) and H.2 (b) were not supported.. 29.

(38) Active Track’s negative result for H.2 could have several reasons. First, the total sample size was 166, including 79 removed samples and 87 valid samples. The effective response rate was 52.4%. Therefore, the matching messages condition had 87 valid samples and the analysis on specific personality types from the selected samples became increasingly smaller, which may have influenced the accuracy of the findings. Second, whether the difference between the two types of our designed messages is significant enough to make users distinguish messages that are logical or emotional may affect the results as well. Lastly, another possible reason is that the experimental period was too short to obtain influential results or that the manner in which Active Track presented messages was attractive to specific personality types. Additional studies to discuss the above factors could be conducted in the future.. 治政大 receiving repetitive Finding 3. Thinking-type people are more motivated when 立 messages. ‧ 國. 學. In our second design cycle, the significant statistical results illustrated that adding the “Repetitiveness” factor for people who received motivational messages can enhance. ‧. their completion rate. The two findings in this cycle were:. y. Nat. sit. (1) People are more motivated when presented with repetitive motivational messages.. al. er. io. The strong evidence (p < .01) showed that after adding the “Repetitiveness” factor. n. in the experimental group, the completion rate was 23% higher on average than in. Ch. i n U. the control groups without the “Repetitiveness” factor.. engchi. v. (2) People with “Thinking” type personalities are more motivated when presented with repetitive motivational messages. There was also a statistical difference among Thinking-type participants when presenting different degrees of repetitiveness of messages. The mean completion rate of Thinking-type participants with repetitive messages was approximately 85% of the completion rate (p < .01); however, those who received no repetitive messages only had 57% on average. The interesting results of the “Repetitiveness” factor in the Thinking and Feeling groups may be attributable to different personality characteristics. Emotional people tend to make decisions based on others’ opinions and Thinkingtype people are more willing to trust facts or make decisions based on rules (Haghshenas et al., 2014). Repetitive notifications from Active Track, which was released by an “academic institution” may strengthen the level of reliability for 30.

(39) Thinking-type people based on the “mere exposure effect.” However, a possible reason for causing insignificant differences in the Feeling group is that emotional people may affected by others’ suggestions (a possible “social” factor), which may reduce the impact of the “mere exposure effect” on them.. Finding 4. The feedback factor had no impact on completion rate in this study. According to the results in the second design cycle, all the results related to the “Feedback” factor were not significantly different. As a result, the feedback design in Active Track had no effect on completion rate in our study.. However, Active Track’s failure to enhance completion rate by the “Feedback” factor. 政治大 both the groups with or without 立 feedback information can view their completion rate. has several implications. According to the design (see Figure A9. in Appendix A),. ‧ 國. 學. on the main page, and the difference between the two groups is the content of their completed results (white box in Figure A9.). Reminding participants of their completion rate through notifications and the completed results on the main page did. ‧. not affect their intention to exercise in this study. Because our research focused on. y. Nat. knowledge of progress (KP) interventions for the feedback factor, we should remove. sit. the percentage value from the main page in the “without feedback” experimental. er. io. group to better differentiate the “Feedback” factor, which may improve the influence. al. n. v i n C h or redesignedUin a future study. feedback information could be modified engchi. of motivational effects between the two groups. Thus, appropriate content design of. 31.

(40) Chapter 6 Conclusion 6.1 Summary This research applied the “Tailoring” persuasive principle to identify personalities (Thinking- and Feeling-type) using MBTI assessment and the “Reminder” persuasive rule to develop corresponding text messages (logical and emotional) to enhance motivational level to exercise adherence. We used the design science methodology to develop a mHealth app and identified that presenting motivational messages to users can improve their adherence to activity schedules regardless of whether the messages matched their personalities or not. Further, the “repetitive notifications” had a. 政治大. combined effect on completion rate for people with “Thinking” rather than “Feeling” type personalities. Lastly, there was no evidence to validate the impact of the. 立. “Feedback” intervention on adherence to exercise plans. The refinement possibilities. ‧. ‧ 國. 學. and limitations will be discussed in the next section.. 6.2 Limitation and future work. y. Nat. sit. The proposed research project adopted design science approach and Apple. er. io. ResearchKit framework. Our aim was to make design artifacts accord with the needs. al. n. v i n plans. Our experiment results C revealed motivational messages and repeated h e nthat hi U c g notifications can enhance motivation levels. Nonetheless, this study was limited by of users and identify design components capable of increasing adherence to exercise. sampling bias, design flaws in Active Track and the messaging and feedback system, which may have influenced the accuracy of the results.. First, during the two months data collecting period, the effective sample size was small (87 active users). The total sample size from the firebase database was 166; however, many of the users did not finish the survey and many others did not continue using Active Track throughout the extended experiment period. A new recruitment plan could be formulated to ensure that most of the participants continue using the application to the end of the experiment. Such a recruitment plan could be. 32.