Candidate Name: Milad Rogha
Program: Computer Science​​​​​​​
Committee Chairs: Dr. Wenwen Dou​​​​​​​
Committee Members: Dr. Depeng Xu, Dr. Doug Markant, Dr. George Banks
Abstract:

The growing importance of machine learning and large language models in decision-making across various domains, including healthcare and law, underscores the necessity of effectively and responsibly communicating their performance to avoid misinterpretation and overreliance. This dissertation explores how visualization design and contextual framing impact human understanding, confidence, and decision-making in AI-assisted environments through three empirical studies.

Three user experiments collectively investigate this challenge. The first explores how eliciting prior beliefs and presenting contrasting narratives affect engagement, recall, and attitude change toward data-driven news articles. The second investigates how the source and presence of AI-generated recommendations shape decision-making in content moderation, revealing a trade-off between decision speed and confidence. The third evaluates how alternative confusion matrix visualizations, such as frequency-framed and Sankey designs, affect lay users’ comprehension and their confidence in calibrating AI model performance.

Together, these studies contribute to a deeper understanding of how visualization design and contextual framing shape user cognition and behavior in AI-supported environments. These studies show that well-designed visualizations can meaningfully enhance non-experts’ ability to understand and appropriately use AI systems. The findings offer practical recommendations for creating more accessible, transparent, and responsible human-AI interactions by supporting informed and calibrated decision-making among lay users.