The role of the generative theory of understanding language in artificial intelligence research

An important area of Artificial Intelligence research is in Natural Language Processing--the attempt to construct computer programs that understand languages that humans speak. This paper examines the generative theory of understanding and two issues that are fundamental to it: 'tacit' knowledge and rule-following. Rule-following as a normative activity is seen as central to understanding a language. D'Agostino's dichotomy of Linguistic Mentalism and Linguistic Realism in Chomsky's work is discussed in terms of various conceptions of rule-following. The results of this discussion are applied to the question as to whether a computer could follow a rule. Since rule-following is seen as being incompatible with a description of rules in mechanical terms, it is suggested that a computer could not follow a rule, and hence, understand a language. Given this result, NLP projects are conceptually unable to complete their tasks as set out for themselves. A suggestion for natural language processing research is to concentrate on the project of constructing better human-computer interfaces rather than building a machine that could understand language is made. Finally, the possibility of a shift from a computational paradigm to a more 'social' paradigm in cognitive science is discussed.