Examining AI Prompt Design for Influencing User Beliefs
This project uncovers how subtle changes in AI-generated responses—like added detail or confident tone—can significantly steer user beliefs, based on a large-scale study we designed
RelEYEance: Real-Time Gaze-Based AI Reliance Detection
This project introduces the RelEYEance model for detecting user reliance on AI in real-time. Leveraging eye-tracking data, we developed a clustering pipeline to classify user reliance into over-reliance, under-reliance, and appropriate reliance during a time-sensitive monitoring task. The model enables adaptive interventions for recalibrating reliance.
Highlight-Informed Saliency Model (HISM) for GUI Design
This project explores how dynamic visual highlights influence user attention in graphical user interfaces (GUIs). Using eye-tracking data, we analyzed temporal changes in gaze behavior across free-viewing and task-driven scenarios. Our novel Highlight-Informed Saliency Model (HISM) integrates spatial and temporal data to predict attention shifts over time, outperforming existing models. These insights enable the design of more effective and user-focused GUIs.
Influence of Work Motivation and Task Difficulty on Human Reliability.
📑 PublishedProceedings of the 2nd International Conference on Reliability Systems Engineering (ICRSE 2017), Beijing, 2017, 2017