INTEGRATION OF MODAL AND FUZZY METHODS OF KNOWLEDGE REPRESENTATION IN ARTIFICIAL AGENTS

To enable artificial systems to meaningfully use a semantic language of communication is one of the long-term and key targets not only in the field of artificial cognitive agents, but also of AI research in general. Given existing solutions for grounding of modal statements of a language of communication and an idea to model internal concepts of the agent as zadehian fuzzy-linguistic concepts, this paper shows how to meaningfully combine the two within a single framework. An accomplished goal is a model for grounding of modal and non-modal statements of a language of communication based on concepts modelled internally as fuzzy sets spanned over the domain of observation. This paper describes a way in which fuzzy-linguistic concepts are activated by perceptual inputs and how an agents grounds respective non-modal statements. Further, an agent supposed to describe an unobserved part of the environment can use autoepistemic operators of possibility, belief, and knowledge to describe its cognitive attitude toward it. It is discussed how the modal extensions of statements with fuzzy-linguistic concepts should be grounded in order to preserve the common-sense. The resulting constraints put on the model of grounding are formally represented in a form of analytical restrictions put on the so-called relation of epistemic satisfaction.