Replicating Human Social Perception in Generative AI: Evaluating the Valence-Dominance Model

TL;DR

The paper addresses whether multimodal generative AI can replicate the valence–dominance theory of facial social perception. It applies PCA to trait ratings produced by Claude 3.5 Sonnet, GPT-4 Turbo, and Gemini 1.5 Pro, comparing the latent dimensions to human judgments across world regions. Results indicate that the first two components align with trustworthiness (valence) and dominance, paralleling prior human data, while a third component is present but variably interpretable. These findings suggest AI systems may encode foundational social-perception frameworks, with important implications for bias, decision-making, and human–AI interactions; the work also raises questions about the third component and its cultural specificity.

Abstract

As artificial intelligence (AI) continues to advance--particularly in generative models--an open question is whether these systems can replicate foundational models of human social perception. A well-established framework in social cognition suggests that social judgments are organized along two primary dimensions: valence (e.g., trustworthiness, warmth) and dominance (e.g., power, assertiveness). This study examines whether multimodal generative AI systems can reproduce this valence-dominance structure when evaluating facial images and how their representations align with those observed across world regions. Through principal component analysis (PCA), we found that the extracted dimensions closely mirrored the theoretical structure of valence and dominance, with trait loadings aligning with established definitions. However, many world regions and generative AI models also exhibited a third component, the nature and significance of which warrant further investigation. These findings demonstrate that multimodal generative AI systems can replicate key aspects of human social perception, raising important questions about their implications for AI-driven decision-making and human-AI interactions.

Figures (1)

• Figure 1: PCA loading matrices for each world region and generative AI model. Positive loadings are shaded in red, while negative loadings are shaded in blue, with darker shades indicating stronger loadings. The proportion of variance explained by each component (Prop. Var) is displayed at the top of each table.