Explaining the Gap: Visualizing One's Predictions Improves Recall and Comprehension of Data
summary
Controlled experiments use two-by-three line graphs and crowd workers with 42 participants per condition to test three elicitation techniques: predicting (draw/type), self-explaining (text), and feedback on the gap, across visual and text modalities. Measuring absolute recall error and slope-based trend error after a 3-minute distractor shows that prompting prediction and self-explanation improves point and trend recall for low-to-moderate familiarity datasets, with visual prediction+feedback producing the largest, replicable gains while text prediction shows little effect.
AI-written summary · about
chapters (11)
- 0:00Introduction to study
- 1:53Research question and hypotheses
- 3:39Designing elicitation mechanisms
- 5:24Study design and conditions
- 7:12Data selection and visualization
- 8:57Procedure and recall tasks
- 10:47Results on recall accuracy
- 12:34Trend recall error findings
- 13:18Replication studies and contributions
- 14:25Future directions and implications
- 15:18Q&A session begins
related talks
01h 10mStorytelling through data
10mThe Future of Data Visualization
21mTurn Data into Action: How visualization aids decision intelligence
11mA Comparison of Cooperative and Competitive Visualizations for Co-located Collaboration
20mReport Design & Psychology (Quick Tips)
20mearn to make the point: data visualisation strategy
11mData Visualization & Storytelling
22mWhen Visualization Best Practices Fall On Deaf Ears