Centralized Regulation of Social Exchanges Between Personality-Based Agents

This paper presents a centralized mechanism for solving the coordination problem of personality-based multiagent systems from the point of view of social exchanges. The social exchanges are evaluated by the agents according to certain scales of exchange values. The agents may have different personality traits, which induce different attitudes towards both the regulation mechanism (blind obedience, eventual obedience etc.) and the possible profits of social exchanges (egoism, altruism etc.). A notion of exchange stability can be defined, and the connections between agents’ personalities and deviations of social exchanges from the stability point can be established. The model supports a decision procedure based on Qualitative Interval Markov Decision Processes, that can solve the problem of keeping the stability of social exchanges between agents, in spite of the different personality traits of the agents. The paper deals only with transparent agents (agents that allow the supervisor to access their balances of exchange values), but we hint on the case of nontransparent agents (agents that restrict such access). The model is analyzed theoretically with respect to the stability of the process, and a contextualized sample of its performance as a supervisor of social stability of a simulated multiagent system is presented.