Date of Award

8-2026

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Electrical Engineering and Computer Science

First Advisor

Thomas C. Eskridge

Second Advisor

Moti Mizrahi

Third Advisor

Khaled Ali Slhoub

Fourth Advisor

Brian A. Lail

Abstract

Trust in artificial intelligence is commonly assessed through self-reported scales or behavioral reliance, yet behavioral reliance is retrospective and can only be observed after a decision has already been made. This thesis examines whether prediction accuracy — a user's ability to predict what an AI system will recommend before its output is revealed — can serve as a prospective correlate of trust in the same empirical sense as behavioral reliance. The study was conducted in two phases using scenario-based AI decision tasks across disaster response, healthcare, and infrastructure restoration contexts, employing a between-group design in which participants either predicted AI outputs before seeing them, evaluated AI outputs directly, or received structured information about the AI's decision logic prior to prediction. Results showed that prediction accuracy did not significantly correlate with self-reported trust or behavioral reliance, though all accuracy-related correlations were consistently positive across both phases. The most notable finding was that prediction confidence — not accuracy — emerged as the dominant correlate of self-reported trust, indicating that users trusted the AI because they felt certain about their predictions, not because those predictions were correct. A structured intervention improved prediction accuracy slightly but did not reduce the gap between confidence and accuracy, suggesting that providing information about AI decision logic alone is insufficient to ground confidence in genuine understanding. These findings establish that prediction accuracy and prediction confidence are empirically distinguishable correlates of trust that require separate measurement and intervention strategies, and they contribute to understanding trust formation in human-AI interaction by demonstrating that felt certainty can drive trust independently of actual understanding.

Download

Share

COinS