Towards the Self-regulation of Personality-Based Social Exchange Processes in Multiagent Systems

摘要

This paper introduces an approach for the self-regulation of personality-based social exchange processes in multiagent systems, where a mechanism of social equilibrium supervision is internalized in the agents with the goal of achieving the equilibrium of the exchanges, guaranteeing the continuity of the interactions. The decision process concerning the best exchanges that an agent should propose to its partner is modeled as one Partially Observable Markov Decision Processs (POMDPs) for each agent personality trait. Each POMDP is decomposed into sub-POMDPs, according to the current balance of exchange values. Based on the relationship that may be established between POMDPs and Belief-Desire-Intention (BDI) architectures, the paper introduces an algorithm to map the policy graphs of the sub-POMDPs to BDI plans, allowing for a library of BDI plans. Simulations were carried out considering different degrees of supervision.