Empirical Calibration of Adaptive Bounded-Confidence Dynamics in Refugee–Host Interaction Networks

This paper develops an empirically calibrated agent-based model of behavioral and network dynamics between Egyptian host-community members and Syrian refugees. Behavioral data from the Dictator Game, Ultimatum Game, and Prisoner’s Dilemma were used at two stages of the calibration. Multidimensional Item Response Theory (MIRT) was used to estimate agents’ initial behavioral orientations, while group-pair response frequencies were used to derive confidence and tolerance thresholds. Across 50 simulation runs, the model reached behavioral consensus while maintaining a cross-group edge fraction above the random-mixing benchmark. Behavioral assortativity remained positive in some runs, indicating that local behavioral sorting can persist despite cross-group connectivity. Jointly replacing the empirical behavioral initialization and empirically weighted network with random initialization produced a similar final outcome. Local sensitivity analysis showed that moderate parameter changes affected assortativity more than the final behavioral gap or cross-group edge fraction. These results show that behavioral convergence, cross-group connectivity, and local behavioral sorting can coexist within the calibrated model.