Abstract
The use of computers to handle corrosion knowledge is beginning to attract much attention. Recent experience with a theoretically based mathematical model and an expert system has shown problems in using existing information that had not been anticipated. Both involve representing knowledge in a computer, but in different ways. The attempt to represent knowledge in a computer imposes special requirements, and information which is valid in its original context may not be adequate for the new purpose. Problems of a similar nature have been encountered in developing expert systems in other fields. These problems are discussed using a model of the relation between ‘information’, in the usual sense, and the ‘knowledge’ needed to address a problem, which helps to clarify the reasons for the problems encountered. Two factors are particularly important, the context of the type of problem being addressed and the nature of the software used.