| An open system should admit agents from many sources. These agents' goals might conflict. Therefore, some actions performed by one agent might be detrimental to another. Such actions can be either acceptable or unacceptable within a given system, as defined by social norms. Limiting actions of agents to enforce social norms would be advantageous. An existing multi-agent system could be modified to do this by observing communications between agents and then removing agents that perform unacceptable actions.; In this thesis, an existing multi-agent system is modified so that social commitments define social norms. An investigation into the usefulness of sanctioning anti-social behaviour is then conducted. Results show that sanctioning is useful in systems comprised of some "good" agents, but detrimental in systems where all agents are "good" (efficiency is decreased) or where all agents are "bad" (these systems become useless). |
