A Quality-Aware Calendar Dashboard for Personal Data Reflection: Design, Implementation, and Empirical Evaluation

Personal informatics systems are increasingly collecting multimodal data from smartphones and wearables, including sensor streams, self-reports, and behavioural logs. However, two limitations constrain their usefulness for reflection. First, data quality remains opaque: systems treat all observations as equally reliable despite missing data, delayed responses, or sensor noise, preventing users from distinguishing genuine behavioural patterns from artefacts of data collection. Second, planning and reflection are structurally separated: calendars represent planned activities, while dashboards summarise past behaviour, making it difficult to relate what was planned to what actually occurred. As a result, users lack the ability to reason about their behaviour in a way that is both temporally integrated and grounded in the reliability of the underlying data. This thesis explores how the calendar can be reconfigured to support reflection over heterogeneous personal data. Rather than treating it solely as a scheduling tool, a calendar-based interface was developed that brings together planned activities, sensor-derived behaviour, and self-reports within a shared temporal view. This allows users to see not only what was intended, but also what actually occurred and when. To support interpretation, the system also exposes missing entries, delayed responses, and sensor gaps through visual indicators, making the reliability of the data visible at the moment of interpretation. This work makes three contributions. First, we introduce a temporal integration approach that aligns heterogeneous data streams within a shared calendar structure, enabling direct comparison between plans and observed behaviour. Second, we formalise a computational data quality framework for personal informatics, operationalising metrics such as question delivery success, response latency, answer duration, and sensor completeness. Third, we report results from two evaluations: (a) a controlled study (N=130) showing that quality-aware visualisation improves the accuracy of behavioural reflection, and (b) a six-week field deployment (N=73) demonstrating that real-time monitoring of data quality supports timely intervention, which was associated with improved data completeness. In the field deployment, researchers were able to detect data gaps and intervene within 24 hours, with response rates increasing following reminders, helping sustain participant engagement over time.