Source-linked AI summary
Beyond the Mirror: Personal Analytics through Visual Juxtaposition with Other People's Data
Sungbok Shin, Sunghyo Chung, Hyeon Jeon, Hyunwook Lee, Minje Choi, Taehun Kim, Jaehoon Choi, Sungahn Ko, Jaegul Choo
TL;DR
Personal analytics often lacks comparative context because it centers on an individual’s own data. The paper introduces CALTREND, which juxtaposes anonymized online calendar logs through visual analytics and probes the approach with two domain experts. The study finds that comparative views support diverse interpretations shaped by domain-specific mental models.
Problem
Personal data interpretations may be limited when analytics focuses only on one individual’s records and lacks comparative social context.
Method
CALTREND combines anonymized calendar logs, t-SNE, topic modeling, and interactive visualizations to compare individual users with groups and other users.
Results
Two domain experts used CALTREND to interpret calendar behavior, producing diverse readings of personal data shaped by their professional mental models.
Takeaways & Limitations
Juxtaposing personal data with others’ can enrich interpretation, while useful meanings and visualizations depend on domain-specific perspectives.
Takeaways & Limitations
The exploration used one dataset and two domain experts, so the authors call for multiple datasets and experts to generalize the findings.
Abstract
from arXiv · showhide
An individual's data can reveal facets of behavior and identity, but its interpretation is context dependent. We can easily identify various self-tracking applications that help people reflect on their lives. However, self-tracking confined to one person's data source may fall short in terms of objectiveness, and insights coming from various perspectives. To address this, we examine how those interpretations about a person's data can be augmented when the data are juxtaposed with that of others using anonymized online calendar logs from a schedule management app. We develop CALTREND, a visual analytics system that compares an individuals anonymized online schedule logs with using those from other people. Using CALTREND as a probe, we conduct a study with two domain experts, one in information technology and one in Korean herbal medicine. We report our observations on how comparative views help enrich the characterization of an individual based on the experts' comments. We find that juxtaposing personal data with others' can potentially lead to diverse interpretations of one dataset shaped by domain-specific mental models.
1 INTRODUCTION
Personal analytics can support self-reflection, but interpreting individual data often requires comparison with broader social context. CALTREND addresses this gap by juxtaposing anonymized calendar records from multiple users and examining interpretations from domain experts.
- Comparative context is essential for judging whether personal patterns are substantial or socially typical.Examples include evaluating salary changes against peers’ earnings and sleep duration against wider population norms.
- A review of 18 calendar-visualization studies found that existing work mostly emphasizes self-comparison rather than comparisons with other individuals.
- CALTREND combines t-SNE and topic modeling to analyze temporal and textual patterns for individual and multiple calendar users.The system uses anonymized online schedule logs and integrates the resulting analyses into visual analytics.
- Two domain experts used CALTREND to interpret individual user behavior and propose actionable insights from the dataset.The experts represented Korean herbal medicine and human resources management in an IT company.
2 DESIGN REQUIREMENTS FOR CALTREND
CALTREND is designed as an analytical probe for interpreting individual calendar data in relation to others’ data. Its requirements emphasize contrastive, multi-faceted, temporal, and contextual analysis.
- CALTREND is intended to reveal deeper and more nuanced insights by examining an individual’s data alongside others’ datasets.
- DR1 Contrastive: The system must highlight differences between analyst-selected groups or individuals.
- DR2 Multi-faceted data analysis: The system should analyze datasets from multiple angles rather than relying on a single perspective.
- DR3 Catching different temporal patterns: The system must present temporal patterns at multiple granularities, including weekly and daily levels.
- DR4 Effective summary of contexts: The system must summarize extensive textual content to support effective analysis of schedule contexts.
3 CALTREND IMPLEMENTATION AND DESIGN
CALTREND processes anonymized online calendar logs into life-mode and scheduling features, then combines machine-learning analyses with interactive visualizations. Its views support user selection, temporal analysis, and textual context comparison.
- Implementation workflow: CALTREND’s workflow deidentifies logs, labels life modes, quantifies user features, extracts insights, and integrates them into interactive visualizations.
- Online schedule logs: The dataset contains 1,652,071 schedule entries from 1,025 primarily English-speaking users, with personally identifiable information removed.
- Preprocessing: Schedules are labeled as work or home using keyword concepts, with 63.2% labeled under one or both modes and 11.3% receiving multiple labels.
- Preprocessing: Eleven features represent schedule modifications, monthly volume, weekday-weekend ratios, hourly distributions, and work-home content rates.
- Selection Interfaces: A t-SNE projection lets analysts detect clusters, identify anomalous users, select groups, and alter layouts through feature weighting.
- Visual analytics interface: CALTREND combines user glyphs, hour-by-day and hour-by-week heatmaps, cumulative-frequency views, and topic-based wordclouds to expose temporal and textual patterns.Glyphs encode schedule totals, life-mode distributions, and hourly frequencies; the weekly heatmap also displays recurring time-slot keywords.
4 STUDY DESIGN
The study used CALTREND as a probe with two domain experts who analyzed calendar-user distributions, identified anomalous users, and considered domain-specific ways to characterize behavior.
- The study involved a marketing expert from LG Electronics and a Korean herbal medicine doctor under agreements protecting schedule-log information.
- Participants selected users, examined distributions, identified anomalous patterns, and proposed additional characterization methods for online calendar users.
- Each study lasted about 70 minutes, including system instruction, hands-on testing, and discussion of domain applications and limitations.
5 TAKEAWAYS
CALTREND shows that juxtaposing personal schedules with others’ data supports diverse, domain-specific interpretations and characterizations. Experts used these comparative views to identify specialized groups, anomalies, lifestyle segments, and potential applications across marketing and health contexts.
- 5 TAKEAWAYS: Experts interpreted the same calendar data through domain-specific mental models, producing diverse characterizations of users.The marketing expert emphasized lifestyles and interests, while the herbal medicine doctor focused on effort, stress, and anomalous schedules.
- 5 TAKEAWAYS: The medical expert viewed schedules as indicators of effort and stress, while recommending sleep data to improve clinical relevance.The proposed applications included investigating undisclosed patient information and potential depression or burnout indicators.
- 5 TAKEAWAYS: Comparative calendar views helped experts identify specialized user groups and additional characteristic dimensions.Experts adjusted cluster-map parameters and feature weights to find lifestyle segments and anomalous schedules.
- 5 TAKEAWAYS: Marketing analysis used calendar-derived lifestyle patterns to brainstorm product ideas and identify target groups within the STP framework.The process linked segmentation, targeting, and positioning to groups sharing calendar-based characteristics.
- 5 TAKEAWAYS: The findings suggest that tailored visualizations should account for domain-specific mental models when supporting self-understanding.The authors identify this adaptation as a promising direction for future work.
6 LIMITATIONS
The study’s conclusions are constrained by privacy risks in calendar data and by its limited empirical scope. Personal events may enable identification, while the use of one dataset and two experts limits generalization.
- 6 LIMITATIONS: Personal calendar events may create privacy risks because supposedly anonymous schedules can still identify users in small groups.The authors distinguish this risk from work-only events that are often visible within organizations.
- 6 LIMITATIONS: The study used one dataset and two domain experts, so broader generalization requires multiple datasets and experts from varied domains.The authors frame this expansion as the next step for testing whether juxtaposition enriches personal understanding more generally.
7 CONCLUSION
The paper investigates how juxtaposing other people’s data can enrich personal analytics. It proposes CALTREND, a visual system for comparing temporal and contextual differences in online schedule logs, and evaluates its potential through interviews with two domain experts.
- 7 CONCLUSION: CALTREND compares temporal and contextual differences between groups in online schedule logs and was examined through interviews with two domain experts.The experts represented Korean herbal medicine and marketing.