Learning Interpersonal Complementarity: A Computational Process Model of the Interpersonal Situation

This project presents a computational process model of interpersonal interaction in which complementarity (e.g., dominance–submission, warmth–warmth) emerges through learning rather than being assumed. Using an active inference framework, agents begin with naïve beliefs about social contingencies and learn the predictability of interpersonal responses via prediction error. Simulations show that agents interacting with realistic social partners shift toward complementary behavior and that variation in learning dynamics produces distinct, clinically relevant interpersonal patterns.