The Interplay of Gender and Anthropomorphism in AI Avatar Design: An Empirical Study on User Experience in Financial Contexts

Abstract

The design of AI virtual assistant (AVA) avatars significantly shapes user perceptions and experience. However, the interplay between avatar gender, anthropomorphism, and contextual task demands remains underexplored, particularly in specific cultural settings. Grounded in the Computers Are Social Actors (CASA) paradigm and Social Role Theory, this study investigates how these two key design attributes influence user experience within the financial service context of Hong Kong. We conducted an exploratory mixed-methods experiment employing a 2 (avatar gender: male vs. female) × 4 (anthropomorphism level: none, low, medium, high) design. Fourteen participants interacted with a series of avatar conditions while completing standardized financial tasks, followed by in-depth semi-structured interviews. Quantitative results, interpreted with caution due to the small sample size, indicated no statistically significant effect of avatar gender on perceived trust, emotional engagement, or anthropomorphism. Qualitative thematic analysis revealed three core insights: (1) a strong context-dependency of user preferences, where professional settings prioritize functionality and moderated anthropomorphism, while entertainment contexts favor expressive and customizable avatars; (2) a complex relationship with anthropomorphism, where moderate levels foster trust and approachability, whereas high realism can induce discomfort or psychological distance; and (3) a prevailing neutral stance toward avatar gender in task-oriented financial interactions, though underlying cultural stereotypes still persist. The study contributes by providing nuanced and context-sensitive evidence that explores the applicability of gender in professional AVA design and clarifies the boundary conditions of anthropomorphism's benefits.