Interval-Valued Hidden Markov Models for Recognizing Personality Traits in Social Exchanges in Open Multiagent Systems

Abstract

This paper presents an application of Interval-valued Hidden Markov Models to the modelling of agent personality traits in multiagent systems. The agents’ behaviors are modeled as probabilistic transitions functions, where interval-valued probabilities are used to express the uncertainty in determining those probabilities. The model of regulation of social exchanges is based on the concept of equilibrium supervisor, which is able to recommend the best exchanges for the agents to perform in order to achieve the equilibrium of the system.