Author

Dylan Wright

Date of Award

2025

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Science

Committee Chair

Vineetha Menon

Committee Member

Jacob Hauenstein

Committee Member

Bryan Mesmer

Research Advisor

Vineetha Menon

Subject(s)

Artificial intelligence, Augmented reality, Human-computer interaction, Explainable AI (XAI)

Abstract

Artificial Intelligence (AI) is increasingly deployed in high-stakes environments where success depends on more than computational accuracy; it also requires alignment with human perception and attention. Traditional explainable AI (XAI) methods provide static, post-hoc justifications that are poorly suited to time-critical decision cycles. This thesis investigates whether perceptually aligned, real-time explanations in Augmented Reality (AR) can enhance performance, workload balance, and trust within the Observe–Orient–Decide–Act (OODA) loop. A custom AR search-and-rescue simulation tested three conditions: no assistance, static overlays, and perceptually aligned explanations using adaptive cues such as occlusion-aware tethers and urgency-based visuals. The results show that any explanation improved results compared to no support, with perceptually aligned explanations producing the strongest gains: accelerating rescue performance, improving attentional alignment, reducing cognitive inefficiencies and fostering calibrated trust without distraction. These findings advance an OODA-centric framework for XAI and demonstrate that explanations can serve as active perceptual supports rather than passive rationalizations. The work contributes design principles for AR-enabled XAI systems that integrate objective, subjective, and physiological evidence to improve human performance in mission-critical contexts.

Available for download on Monday, June 15, 2026

Share

COinS