Social psychology does not yet have a strong cognitive modeling tradition. This is not for lack of cognitive modeling tools that are relevant and useful for modeling social psychological phenomena. For instance, several researchers have successfully demonstrated how connectionist modeling techniques can be used to build computational explanations of key phenomena of interest to social psychologists, such as stereotyping, prejudice and priming (Kunda & Thagard, 1996; Schröder & Thagard, 2014). In this project we contribute to this important development by addressing a major obstacle to the progression of connectionist modeling in social psychology: That is, how can we reconcile the intuitive concepts that figure in the verbal explanations that pervade social psychological theories with formal properties and processes in connectionist models? We illustrate a systematic way of addressing this question by considering the theoretical concept of ‘social categories’, which plays a central role in social psychological theories. Using computer simulation, we show that if social categories are defined as ‘excluders’ in connectionist models then key social psychological phenomena can be replicated, while maintaining a clear link with the intuitive concept of social categories. We discuss the broader implications of our simulation results for both social psychology and cognitive modeling.