Embedding Social Perception Dimensions in a Semantic Space: Toward a Quantitative Synthesis

What is it about?

This paper introduces a novel computational method to quantitatively compare three foundational models of social perception: the Stereotype Content Model (SCM), the Dual Perspective Model (DPM), and the Semantic Differential (SD). The new method uses word embedding techniques to map the core dimensions of each model (e.g., Warmth, Competence, Communion, Agency) into a single, shared semantic space. The method automates the selection of over 500 contrasting word pairs (like "good-bad" or "strong-weak") to create objective, numerical representations of these abstract theoretical concepts. The results reveal the complex relationships between the models, confirming that SCM's “Warmth” and DPM‘s “Communion” are semantically very similar, as are SCM’s “Competence” and DPM‘s “Agency”. The analysis suggests that social perception is fundamentally structured around two core components: subjective evaluation (likability) and objective attributes (capability). To validate this new approach, the computationally derived dimensions were used to classify Rosenberg’s 64 personality traits, demonstrating significantly improved predictive accuracy over previous methods—by 19% for SD, 13% for DPM, and 4% for SCM dimensions.

Featured Image

Photo by Judy Been on Unsplash

Photo by Judy Been on Unsplash

Why is it important?

Traditional methods for comparing these theories, such as vignette-based experiments, are often inefficient, difficult to scale, and prone to bias. This research provides a powerful, objective, and scalable new tool for the social sciences, bridging the gap between abstract theory and large-scale computational analysis. By converting theoretical dimensions into quantifiable vectors, the method allows for direct and rigorous comparison of foundational models that were previously difficult to integrate, helping to resolve long-standing conceptual debates in the field of social perception. The practical implications are significant. This approach moves beyond the limitations of traditional self-report and experimental studies, offering a more reliable and generalizable way to analyze social constructs. It facilitates the development of a more unified and comprehensive theory of how we perceive others. Furthermore, by creating a mathematical framework for understanding social judgments, this work can support the integration of more nuanced, human-like social reasoning into artificial intelligence and robotics, advancing both social science and technological innovation.

Perspectives