Visual Debugging for Data Workflows
Visualization opportunities for evidence-driven debugging in data-intensive programming.
Debugging in pandas notebooks and similar data workflows is often an evidence-driven reasoning process: practitioners assemble clues from code, outputs, and intermediate states to locate expected–observed discrepancies.
Through interviews with practitioners, this project characterizes cross-cutting debugging challenges and translates them into visualization requirements—cross-artifact evidence alignment, expectation-grounded comparison, and traceable state evolution.
Outcomes
- First-author short paper at IEEE VIS 2026
- Related full manuscript under review at ACM CHI 2027
- Ongoing prototype for side-by-side expectation–observation comparison
Advisor: Prof. Rebecca Faust