Speech Acts and Multiple Environments

Abstract

This brief paper outlines joint work that implies a view of the explication of Speech Acts different from those current in Al. It is exploratory at present, but we intend that it should lead to programs as soon as possible. We propose a system that engages in dialogue with a human user so as to discuss others and in doing so, interprets the appropriate communicative force of the user's utterances. The three key features of the explication of co mmuni cative force presented are: I. multiple envi_ronments: the ability to interpret an expression or a des c ription relative to some particular knowledge, e.g. relative to the beliefs of one person about another. It is a ge neralization of the computer science approach to expression evaluation, using the notion of an environment to fix the interpretation of names occurring in the expression. 2. dual knowledge representation: an "experiential" representation is used to mem orize sentences or other episodes (exemplified by pseudo-tests, or PTs), while a frame-type repres e ntation is used to memorize knowledge needed more generally for reasoning. 3. a least effort principle of understanding: considered to be a language universal, which requires (a) that the system maintain as much representation as possible in the frame-like forms, rather than in PTs, and (b) that the system maintain a highly redundant representation in the PTs, achieved by a process we call percolation of belief.