A high-dimensional model of social impressions

People form social impressions from visual cues such as faces, which are argued by various models to arise from some limited set of fixed dimensions (e.g., trustworthiness and dominance). We argue that these dimensions, rather than reflecting intrinsic mechanisms, emerge from adaptive visuo-semantic processes in a high-dimensional neural-state space. Drawing on attractor neural-network models, we propose a framework treating social impressions as dynamic trajectories that stabilize over time, influenced not only by visual cues but also by conceptual associations and higher-order social cognition. Unlike low-dimensional models, this framework can account for cultural, individual, and situational factors that shape impressions. A high-dimensional framework makes several novel predictions and can offer a more accurate and complete understanding of the fluidity and complexity of social perception.