Liang (Jackie) Tang's Dissertation Defense

This dissertation examines how different forms of proximity visual, cognitive, and spatial—shape human trust and collaboration with embodied agents. As AI systems increasingly operate as social partners in virtual and mixed reality contexts, understanding the mechanisms that foster or erode trust becomes essential for designing effective and reliable human–AI relationships. Building upon theories of social cognition, embodiment, and proxemics, this research proposes that proximity—beyond physical distance—functions as a multidimensional construct that governs how humans perceive, interpret, and calibrate trust toward AI agents.

Across three studies, this dissertation systematically investigates these dimensions. Study 1 explored visual proximity through self–avatar similarity, examining how users’ embodied representations influence perceived alignment and initial trust toward AI partners. Results showed that greater avatar similarity increased perceived identification, social presence, and baseline trust, demonstrating that visual embodiment shapes the psychological foundations of human–AI rapport. Study 2 examined cognitive proximity by manipulating agents’ communication framing and reasoning transparency. Findings indicated that when agents conveyed human-like reasoning styles and goal alignment, participants exhibited higher cognitive resonance, improved interpretability, and more stable trust trajectories. These results extend trust-in-automation models by highlighting that cognitive congruence—rather than competence alone—drives sustainable trust. Study 3 investigated spatial proximity in virtual navigation tasks, varying the distance (personal vs social zone) in a collaborative maze environment. Participants interacting with closer agents demonstrated stronger trust development, faster learning, and more fluid communication, while those with distant agents displayed improved trust calibration—showing reduced overcompliance and greater critical evaluation of AI guidance. Together, these findings reveal that proximity modulates both emotional engagement and analytical control in human–AI interaction.

Integrating across studies, this dissertation demonstrates that proximity operates as a fundamental organizing principle in human–AI trust formation. Visual and cognitive proximity foster identification and understanding, while spatial proximity dynamically shapes the affective and behavioral calibration of trust. These multidimensional insights extend Hall’s proxemics theory to intelligent systems, showing that human–AI relationships are governed by social distance cues analogous to human–human interaction. Practically, the findings inform the design of embodied AI and virtual agents by emphasizing that optimal proximity—visual, cognitive, and spatial that supports balanced trust: strong enough to enable cooperation, yet calibrated enough to prevent overreliance. This framework contributes to the theoretical foundation of trustworthy AI, advancing the design of interactive systems that are not only intelligent but socially attuned, adaptive, and human-centered.

Questions? Contact Liang Tang.