State of the Art

Abstract

Many areas of artificial intelligence (AI) have had to struggle in various ways in which contextual dependencies arise, e.g. knowledge representation, natural language processing or expert systems to name a few. As the importance of context is frequently glossed over in the literature, researchers noted already in the early 80s that the denotation of the term has become murkier throughout its extensions in different fields of AI, calling it a conceptual garbage can [Clark and Carlson, 1981]. A classic AI example of contextual computing showed how the medicinal expert system MYCIN [Wallis and Shortliffe, 1982] can benefit from contextual considerations when prescribing treatments, resulting in fewer fatal intoxications as a result from the prescription [McCarthy, 1984]. This case constitutes a classic example as it establishes a blueprint for the so-called representational approaches to contextual computing in AI [Dourish, 2001, Dey, 2001]. Initially, we find an expert system that prescribes treatments based solely on the diagnosed disease.