In my Independent Human-Centered AI Evaluation Project, I reviewed 2,834 messages across a multi-week, cross-tool workflow to evaluate human–AI collaboration reliability. I developed an evidence-based failure taxonomy covering context continuity, trust calibration, source binding, and agent handoffs.
Through my ongoing Human–AI Cognitive Partnership Research, I’m developing a framework that uses explicit evidence-status labels to trace how challenge, clarification, and feedback reshape working models while preserving uncertainty about causality. I also evaluated a multi-agent knowledge workflow spanning ChatGPT, Voice, Work, Google Sheets, GitHub, and Todoist, turning observed failures into a traceable feedback package and recurrence log.
I independently conceived and developed The Infinite Discussion, a speculative interaction experiment in which two AI personas were instructed to remain in permanent disagreement. I designed its behavioral constraints and dialogue framework, then translated the resulting machine dialogue into an editorial and exhibition experience examining machine vision and human perception.
