Blame on them, shame on us: combining affective and normative behaviour in intelligent virtual agents

Virtual environments often try to simulate situations where agents should display socially acceptable behaviour, as described by social norms. Actions that fulfil or violate norms can be potential sources for emotions, not just for the agent responsible, but for all agents that witness the event. We claim that the type and intensity of those emotions depends, not only on the salience (importance) of the norm, but also on the social relations between agents. We present a model to generate emotions triggered by a norm-related event, based on a subjective evaluation (appraisal) of this event made by every agent that witnessed it. This appraisal takes into account the salience of the norms and the social relations in order to create more believable agents.