A Bayesian approach to norm identification

When entering a system, an agent should be aware of the obligations and prohibitions (collectively norms) that will affect it. Several solutions to this norm identification problem have been proposed, which make use of observations of either other's norm compliant, or norm violating, behaviour. These solutions fail in situations where norms are typically violated, or complied with, respectively. In this paper we propose a Bayesian approach to norm identification which operates by learning from both norm compliant and norm violating behaviour. By utilising both types of behaviour, our work not only overcomes a major limitation of existing approaches, but also yields improved performance over the state-of-the-art. We evaluate its effectiveness empirically, showing, under certain conditions, high accuracy scores.