Four general representations and processes for use in problem solving

It is argued that “knowledge representation” as normally understood is one of four very general constructs — two representations and two processes — which are commonly used in artificial intelligence (AI) research yet are largely unrecognised. These constructs mirror much of the “generate-and-test” strategy for problem solving in AI where competing solutions are generated and then tested to select the best one — except that one of the constructs, “coherence representation,” is not present in the strategy. Examples of various problems are given that show uses of the constructs. Some implications for artificial intelligence research of the constructs, especially coherence representation, are discussed.