Electronic Theses and Dissertations Archive

Date

2026

Document Type

Dissertation

Degree Name

Doctor of Philosophy

Department

Business Administration

Committee Chair

Huigang Liang

Committee Member

Nirmalee Raddatz

Committee Member

Yafang Li

Committee Member

Yuan Zhang

Abstract

This dissertation presents three empirical essays examining how deepfakes affect how people evaluate technology, trust political candidates, and respond to the discovery of synthetic media. Across all three studies, effects prove to be deeply context-dependent — shaped by moral framing, reputational anchors, audience identity, and the timing of discovery rather than by the content of the deepfake itself. Essay 1 examines how individuals evaluate deepfake applications in terms of perceived risk and perceived benefit. Using Rational Choice Theory and the Hunt-Vitell model, a 2×2 factorial vignette experiment (N = 275) reveals that malevolent deepfakes reduce perceived benefit and increase perceived risk through Ethical Concern and Social Approval as parallel mediators. Ethical Concern mediates both outcomes; Social Approval mediates perceived benefit only. Purpose framing selectively attenuates the Ethical Concern pathway under hedonic conditions while leaving Social Approval intact, revising the Hunt-Vitell model's parallel-pathway assumption. Essay 2 examines how political deepfakes asymmetrically shape voter trust. Integrating Expectancy Violation Theory, anchoring-adjustment, and voter gender, two between-subjects experiments conducted during the 2024 U.S. presidential election (N = 223; N = 339) demonstrate that negative deepfakes sharply erode trustworthiness for norm-conforming candidates while leaving norm-defying candidates unaffected, with male voters actually awarding higher trust to norm-defying candidates after negative deepfake exposure. Positive deepfakes produce the mirror-image pattern, confirming the anchoring account across valence conditions. Effects are jointly moderated by candidate identity and voter gender. Essay 3 shifts the frame from exposure to discovery, examining what happens psychologically when a viewer learns that content they believed to be authentic was AI-generated. Drawing on Cognitive Dissonance Theory, a 2×2 between-subjects experiment (N = 380) finds that disclosure produces a significant full indirect effect on media distrust through anger and disgust, emotions selectively activated by the deception event. Stimulus authenticity moderates the disclosure-to-emotions pathway monotonically, confirming CDT's magnitude principle in a perceptual domain. Together, the three essays establish that deepfake effects are contingent rather than content determined. The dissertation contributes theory-driven frameworks to a field that has operated through outcome-testing, derives testable revisions to three theoretical frameworks, and offers governance implications for platform design, campaign strategy, and AI disclosure policy.

Comments

Data is provided by the student.

Library Comment

Dissertation or thesis originally submitted to ProQuest/Clarivate.

Notes

Embargoed until 07-21-2028

Download

Available for download on Friday, July 21, 2028

Share

COinS