Title Page
Abstract
Contents
Chapter 1. Introduction 18
1.1. Background and Motivation 18
1.2. Research Questions and Overview of Approaches 22
1.3. Contribution 24
1.4. Structure of Dissertation 25
Chapter 2. Related Work 26
2.1. Personal Ambient Information Visualization and Its Challenges 27
2.1.1. Background and Practice in the Personal Context 27
2.1.2. Necessity of Personalization Support 28
2.1.3. Necessity of Tight Accordance with User's Attention 29
2.2. Personalization of Notification Method 30
2.2.1. Notification Delivery Mechanisms to Minimize Distraction 30
2.2.2. Supporting Longitudinal Interactions with Notifications 32
2.3. Integration of BCI into Attention-Adaptive Ambience 34
2.3.1. Background of EEG and BCI 34
2.3.2. Prediction of Attention State based on EEG data 35
Chapter 3. DataHalo: Customizable Smartphone Notification Visualization System to Embody PAAIV 37
3.1. Design Rationale 38
3.2. System Design and User Interfaces 41
3.2.1. System Components 41
3.2.2. Halo Configuration 41
3.2.3. Halo Visualization 44
3.2.4. Use case 45
3.3. Usability Study 48
3.3.1. Study Procedure 48
3.3.2. Participants 49
3.3.3. Analysis 49
3.3.4. Result 50
3.4. Implications and Design Improvements 52
Chapter 4. Understanding User's Design Strategies toward PAAIV in Daily Context 55
4.1. Deployment study 56
4.1.1. Study Design 56
4.1.2. Participants 57
4.1.3. Technical Specifications 58
4.2. Data Analysis 60
4.2.1. Data Preprocessing 60
4.2.2. App-Level Customization Analysis 61
4.2.3. Virtual Category-Level Design Analysis 62
4.2.4. Halo Notification Analysis 62
4.3. Result 63
4.3.1. Notification Management and Attendance before Using DataHalo 63
4.3.2. App-level Usage of DataHalo 64
4.3.3. Participants' Design Motivations and Choices 71
4.3.4. Virtual Category-Level Design Strategies 75
4.3.5. Halo Notification Analysis 79
4.3.6. Usability and User Experience 81
4.4. Discussion 82
4.4.1. Customization Support towards the Personalized Ambience 82
4.4.2. The Ease of Adaptation to Trace the Moving Sweet Spot of User Experience 84
4.4.3. Generalizability beyond Culture and Population 85
4.4.4. Implications for Better Personalization Support 86
Chapter 5. Exploring Design Space of Attention-Adaptive Ambient Information Visualization 88
5.1. Expected Technological Challenges of A3InfoVis 90
5.2. EEG Collection & Statistical Feature Analysis 93
5.2.1. Study Design 93
5.2.2. Participants 95
5.2.3. Data Preprocessing 95
5.2.4. Statistical Analysis 96
5.2.5. Result 97
5.2.6. Analytic Insights 102
5.3. Analysis on Technological Challenges 103
5.3.1. Method 104
5.3.2. Result 107
5.3.3. Challenges Revisited: Possibilities and Limitations 111
5.4. Analysis on Design Opportunities 114
5.4.1. Participants 114
5.4.2. Study Design 115
5.4.3. Analysis 115
5.4.4. Result 116
5.5. Design Implications 119
5.5.1. Visualization of Predicted Attention State with Credibility 120
5.5.2. Adequate Application Scenarios of A3InfoVis 121
5.5.3. End-user Support for Design of Personalized A3InfoVis 123
Chapter 6. Discussion 125
6.1. Lessons Learned 125
6.2. Design Implications for PAAIV 127
6.2.1. Personalization According to User's Design Perspectives 129
6.2.2. Personalization According to User's Real-time Attentional Demands 130
6.2.3. Integration of Two Personalization Approaches 131
6.3. Future Research Opportunities for PAAIV 133
6.3.1. Personalized Visualization Design Empowered by Image Generation Models 133
6.3.2. Extending Expressiveness of Halo Visualization 134
6.3.3. Design and Evaluation of Attention-Adaptive Ambience Visualization for Smartphone Notifications 134
6.3.4. Integration of Other Contextual Information into PAAIV 135
6.4. Limitations 135
6.4.1. DataHalo 136
6.4.2. A3InfoVis 136
Chapter 7. Conclusion 138
7.1. Summary of Approaches 138
7.2. Summary of Contributions 140
7.3. Summary of Future Research Opportunities 140
Bibliography 143
APPENDICES 155
APPENDIX A. Halo Visualization 155
A.1. Connection with Constructive Visualization 155
APPENDIX B. Usability Study of DataHalo 156
B.1. Virtual Categories during the Task 2 156
APPENDIX C. Deployment Study of DataHalo 158
C.1. Post-study questionnaire 158
C.2. Popular Design Strategies 158
국문 초록 160
Table 5.1. Two-way RM ANOVA Results of Relative Power of Frequency Band 99
Table 5.2. Two-way RM ANOVA Result of Task Engagement Index 101
Table 5.3. Average Performance Metrics of Personal Models by Selected Features 110
Table 5.4. Comparison of Performance Metric between Personal Models and General Model. 111
Table B.1. Smartphone applications that the usability study participants chose for the second halo design task and virtual categories they created. 157
Table C.1. Questions for inquiring the user experience during the DataHalo usage. 158
Table C.2. Frequent and widely used design strategies among participants. #P stands for the number of participants who shared the strategy and Freq. stands for the total frequency of the strategy. 159
Figure 1.1. Three dimensions of Personalized Ambience. Ambience is a unique characteristic shaped by information, longitudinal delivery method, and representation of a given... 19
Figure 3.1. Visual interface of DataHalo: (a) DataHalo integrated with the smartphone home screen. Each app has its own halo visualization that extends a conventional app... 39
Figure 3.2. The main tabs of the halo configuration page (top) and the underlying pipeline of the visualization construction (bottom). 42
Figure 3.3. The halo visualization Han created. 45
Figure 3.4. The change of graphical mark based on visual encoding. Elapsed time and Han's attendance triggered the change. 46
Figure 3.5. Examples of app halos created by participants during the usability study. In the end of each halo design task, we captured the preview from the halo configuration page. 51
Figure 3.6. Customization flow of example-based importance modeling. 53
Figure 3.7. Customization flow of example-based visual encoding. 54
Figure 4.1. Halo creation and edit events by app category.We aggregated the events according to app category provided by Google Play Store. The color encodes the week the... 64
Figure 4.2. Halo creation and edit events by participant. The color encodes the week the event occurred. 65
Figure 4.3. Personal difference in usage of DataHalo. (a) P02 pursued pragmatism while (b) P06 valued aesthetic experience. Upon P06's request of not disclosing the photo of... 70
Figure 4.4. An app-level design motivation (M1) and four virtual category-level design motivations (M2-5)with corresponding choices. For each design motivation, the grey-colored design choice implies that user had the minimal motivation to customize their interaction... 72
Figure 4.5. Distribution of design choices for the user-created virtual category and the Remainder category. 76
Figure 4.6. Relationship between design motivations, choices, virtual category-level design strategy, and graphical mark of P02's Loop Habit Tracker. 77
Figure 4.7. The result of halo notification analysis 79
Figure 4.8. The result of 7-point Likert scale questionnaire about the UX of DataHalo. 83
Figure 5.1. List of commercial EEG devices available for purchase as of 2023. 92
Figure 5.2. Correspondence between electrodes of the 10-20 system and the brain region variable. 96
Figure 5.3. Results of the post-hoc analysis on the brain regional effect. The row corresponds to the brain regions: (i) prefrontal, (ii) frontal, (iii) central, (iv) parietal, (v) tempo-... 98
Figure 5.4. Pipeline of the real-time classification of the user's attention state. The classifier receives the brain wave feature matrix (19 channels * 24 features) every one second... 104
Figure 5.5. Average performance of personal models trained with channels of the selected regions. For classifiers, XGBoost (solid blue) always outperformed C-SVC (light... 108
Figure 6.1. PAAIV Framework and its two personalization approaches. 128
Figure A.1. The components of constructive visualization and corresponding design constraints of halo visualization. Basic Token refers to a discrete graphical mark representing... 155