Learning from Experience to Generate New Regulations

Both human and multi-agent societies are prone to best function with the inclusion of regulations. Human societies have developed jurisprudence as the theory and philosophy of law. Within it, utilitarianism has the view that laws should be crafted so as to produce the best consequences. Following this same objective, we propose an approach to enhance a multi-agent system with a regulatory authority that generates new regulations –norms– based on the outcome of previous experiences. These regulations are learned by applying a machine learning technique (based on Case-Based Reasoning) that uses previous experiences to solve new problems. As a scenario to evaluate this innovative proposal, we use a simplified version of a traffic simulation scenario, where agents move within a road junction. Gathered experiences can then be easily mapped into regular traffic rules that, if followed, happen to be effective in avoiding undesired situations —and promoting desired ones. Thus, we can conclude that our approach can be successfully used to create new regulations for those multi-agent systems that accomplish two general conditions: to be able to continuously gather and evaluate experiences from its regular functioning; and to be characterized in such a way that similar social situations require similar regulations.