An Analysis of Artificial Intelligence Based Models for Describing Cornmunicative Choice
Charles Pavitt · Communication Theory · 1991
Workers in artificial intelligence (Al) have proposed various models for behavioral choice that can be adapted for use in descriptions of the knowledge and behavior of communicators. The use of AI-based models in this context assumes that the communicator can be conceived of as choosing an utterance from a set of relevant options. This view of “communicative choice” is consistent with “purposive” explanations for human action, but advocates of A1 models are not satisfied with the models, such as the practical syllogism, employed by “action” theorists. Instead, A1 modelers base their accounts of activity on models that are technically equivalent to high-level computer programs. Three types of AI-based models —plans, transition networks, and production systems —are reviewed, and the extent to which each can represent a set of “commonsense” notions about communication is analyzed. It apperirs that plans and transition networks, in their “pure” form, have the same advantap and disadvantages when used as models for communicative choice and that these strengths and weaknesses are complementary to those of production systems. While a hybrid model containing features of each may be ideal, it is not clear that such a model is feasible, and it may be best to apply each type of model to the situation it best represents.