Belief Systems and Plans For Communication

Communication between rational agents can be seen as an exchange of internal concepts expressed using a shared language. Most of the communication acts between humans are performed through Natural Language (NL), and the meaning of the expressions of the language adopted is assumed to be known and properly understood by all of the participants. This suggests that the process of communication using NL requires the source-agent (‘speaker’) to hypothesise a model of its audience’s (‘hearer’) knowledge and beliefs. Moreover, the initial reasons which originate the intention (or need) to communicate (and which may or may not be included in the message) might consist of more abstract perlocutionary goals, such as the induction of a specific reasoning process (inference) within the hearer’s mind producing the change of one or more of his/her beliefs. Since these goals play an important role in the process of determining the content and the structure of a discourse, it becomes necessary for the speaker not only to hypothesise and maintain a model of the hearer’s beliefs, but also to plan the structure of the message according to the effects that the utterances will have upon them. Whilst research in modal logics has produced many theoretical results on the subject of belief systems, the applications of these theories to other domains (like beliefs communication in NL) and the issues related to their actual implementation, such as belief grounding, have been, in a sense, neglected. In this paper I introduce a natural language oriented model of beliefs which uses the concept of ‘event’ as fundamental unit for the knowledge representation. The basic features of the model defined include a qualitative degrees-of-belief approach to the representation of uncertain beliefs, two-level nested mutual belief expressions, and grounding of each belief through rationality and experience. Secondly, taking this model as a starting point, I formalise a general Theoretical Framework for belief systems representation which embodies an inference-rules Boolean-functions based reasoning mechanism for beliefs extension and proving. This general model represents the knowledge on two different levels of abstraction, separating between implicit and explicit information. Each of these two levels contains propositions expressed using a distinct language of beliefs: the ‘outer’ language, adopted to model the implicit knowledge, and the ‘inner’ language, used to specify explicit beliefs and constituting a proper subset of the former. Finally, I illustrate the integration of the specific model implemented which meets the requirements of the above mentioned theoretical paradigm within a planning system, in which the set of beliefs held by the system represents the current mental state of the speaker. Accordingly, any solution found by the planner will be composed of a sequence of actions (operator instances) that transforms the initial state into a final mental state in which the assigned perlocutionary goal is achieved. Hence, the system resulting from this integration constitutes an automatic generator of plans for NL communication of beliefs among agents which may be driven by perlocutionary intentions. 1 E-mail: [email protected].