Agent-based persuasion model with concessions dependent on emotion and time beliefs

Abstract

Automated negotiation agents require human-like adaptability in emotionally charged and time-constrained settings. This study introduces an Emotion-Time Dual-Process Framework that integrates the Appraisal Tendency Framework with dynamic temporal modeling. Emotions are decomposed into pleasantness and certainty dimensions and mapped to six emotional persuasion strategies. A variable-rate time function is designed to capture the perceptions of dynamic time pressure. Emotion and time pressure jointly drive a state-dependent concession updating model. The proposed framework was validated through a series of simulation experiments based on different scenarios. The results demonstrate that the proposed framework has significant advantages in improving negotiation success rates, joint utility, and outcome fairness against baseline models. In particular, incorporating emotional factors reduces utility disparity between parties by 28.55%, while the proposed time function improves negotiation efficiency by 12.99% without sacrificing fairness or the success rate. This study provides a theorical basis for developing highly more human-like and adaptive intelligent negotiation systems.