Mental States Recognition from Speech Acts through Abduction

In a Multi-Agent Environment, high degrees of interaction require the agents' being able to evchange information using high level languages and communication protocols. Referring to the classical Speech Acts Theory, it seems to us that the idea of predicting the effects of un utterance over the mental state of the addressee is rather optimistic. These effects are not really completely "a-priori" predictable by the speaker. Along with some predicatable main effects there are other unforeseeable effects. These can be seen as the result of some kind of plausible inference, particularly abduction, performed by the hearer himself over the received communication, its own actual mental state and its previous mental image of the speaker's mental state. In this paper we propose and explore the idea that, having heard an utterance, the hearer can use abduction to create or update his mental image of the speaker's mental state. This idea can be seen as a generalization of the concept of "Plan Recognition" to that of "Mental State Recognition" and is useful in man-machine or machine-machine communication.