Degree

Doctor of Philosophy (PhD)

Department

Textiles, Apparel Design, and Merchandising

Document Type

Dissertation

Abstract

The rapid diffusion of artificial intelligence (AI) and generative AI (GenAI) is transforming fashion retail by enabling personalized, interactive, and data-driven styling services. Despite their growing market presence, empirical understanding of how consumers evaluate and adopt these services remains limited. This dissertation examines consumer adoption mechanisms across key AI-powered styling applications, including virtual try-ons (VTO), wardrobe management, new item recommendations, and both human-like and chatbot-based stylists.

A multi-method research design was employed. First, 23 semi-structured in-depth interviews explore how consumers interpret AI-driven styling services. Findings show that consumers rely on familiar schemas, such as digital tools or human stylists, while weighing benefits (e.g., efficiency, personalization) against concerns (e.g., privacy, authenticity), resulting in conditional openness toward adoption. Second, a scenario-based survey combined with a quasi-experiment (n = 558) was employed to examine the effects of AI agentic capabilities and VTO richness on new item recommendation services. PLS-SEM analysis results indicate that agentic capabilities enhance perceived personalization and competence, while perceived AI threat increases privacy concern and reduces perceived value. VTO richness improves playfulness, fit confidence, and adoption, while unexpectedly lowering privacy concern.

Third, this study investigates consumer adoption of AI-driven wardrobe management services. Using a scenario-based survey (n = 461) analyzed via PLS-SEM, results show AI Agenticity enhances personalization and competence, AI threat perception increases privacy concern, and service attitude drives adoption, moderated by technology innovativeness. Finally, this study examines how interface human likeness shapes consumer adoption of AI-driven styling services using a scenario-based randomized online experiment (n = 1,019) via PLS-SEM. Results show that, compared to chatbot interfaces, highly human-like AI stylists reduce anticipated quality, disclosure willingness, and service adoption. However, they indirectly enhance these outcomes through increased perceived uncanniness, signaling novelty and technological sophistication.

Theoretically, this research advances consumer–AI interaction by integrating multiple theories (e.g., Schema Theory, Behavioral Reasoning Theory, Digital Agenticity Theory, Privacy Calculus Theory, and Uncanny Valley Theory) within fashion contexts. Methodologically, it demonstrates the value of combining qualitative, experimental, and survey approaches. Managerially, the findings guide the design of AI styling systems that balance personalization, transparency, and psychological comfort to foster sustainable consumer adoption.

Date

7-6-2026

Committee Chair

Liu, Chuanlan

LSU Acknowledgement

1

LSU Accessibility Acknowledgment

1

Download

Available for download on Tuesday, June 14, 2033

Share

COinS