Avoiding drill-down fallacies with VisPilot: assisted exploration of data subsets
summary
VisPilot is an interactive recommendation system that automatically picks k=10 visualizations using a proven approximation algorithm to maximize a min-max objective: choose views that deviate most from their closest parent to avoid the drill-down fallacy. The prototype runs on multi-attribute tables (demo on Connecticut police-stop data), supports expanding a chosen node, and flags small sample sizes. A lab study with 18 data-experienced participants shows better insight discovery, prediction of unseen visuals, and attribute-importance rankings than breadth-first and k-means baselines.
AI-written summary · about
chapters (9)
- 0:00Start
- 0:36Drill Downs
- 2:48Challenges of Manual Drill Down
- 4:29Drill Down Fallacy
- 6:54Demo
- 9:10Breadth-First Search
- 9:34K-Means Clustering
- 12:05Conclusion
- 15:00Minimum Sample Size of each Graph
related talks
42mInteractive Data Analysis: Visualization and Beyond
20mVoyager 2: Augmenting Visual Analysis with Partial View Specifications
32mData Viz for Engineers: Optimizing Insight & Decision Making Through Visualization
10mThe Future of Data Visualization
21mA Visual Interaction Framework for Dimensionality Reduction Based Data Exploration
50mDVUX: Data Visualization and User Experience
17mFlight of Insights: Navigating Data Turbulence with SQL Modelling and Visualisation
22mResViz: Politics and Design Issues in Visualizing Academic Metrics