TrustmeWatcher: An Application for Workplace Micro-Sensing and Explainable Well-Being Feedback

Local AI processing protects privacy: The application runs its well-being prediction model locally on users' devices rather than sending data to external servers, allowing participants to maintain control over sensitive workplace activity information while still receiving personalized feedback.

Transparency through explainable AI: The system uses XAI techniques to help users understand why the AI made specific predictions about their well-being based on their computer activity, moving beyond "black box" predictions that users cannot interpret or trust.

Integration of multiple data types improves insights: By combining continuous computer activity monitoring with self-reported questionnaire responses and video highlights, the system creates a more complete picture of workplace well-being than any single data source alone, with the model trained on these aligned records from 17 participants.

User control over surveillance is built-in: The application includes explicit Privacy Control features allowing participants to pause or resume camera and eye-tracker data collection, acknowledging that workplace monitoring raises legitimate privacy concerns that should be addressed through user agency rather than ignored.

Dashboard presents predictions in understandable terms: Rather than showing raw AI scores, the interface displays well-being predictions using "semantic bands"—human-friendly categories—making the AI's assessment of six normalized state scores and overall well-being accessible to non-technical users.