The declarative-procedural battle was an important one in Artificial Intelligence (AI). It dissolved rather than being resolved; the result was a much greater respect for the importance of knowledge representation in current AI work.
This paper states some good reasons for choosing a declarative knowledge representation. An intelligent interpreter of declarative knowledge is similar to the behavior of an intelligent and competent human being. The interpreter will find at each step the constraint which is the most difficult to satisfy, adapting itself to all the details of the problem.
The principles of such an interpreter are discussed by taking as examples two implemented interpreters: one interpreter, for the language ALICE [1], solves various combinatorial problems; another interpreter [2] receives syntactic, semantic, and pragmatic information in a declarative form for understanding natural language.
Although the paper reasonably argues the preference for declarative representation, I cannot agree that man is a “general interpreter of declarative knowledge.” It seems more natural to find a hybrid declarative-procedural interpretation for this case.