Knowledge engineering problems during expert system development

Expert systems are artificial intelligence computer programs which emulate human expertise within some domain of interest. The recent financial successes of such expert systems as PROSPECTOR and XCON have created a demand in government and industry for the developmental skills necessary to build such systems. Expert system knowledge engineering and in particular knowledge acquisition continues to be the most difficult process encountered during expert system develpment. Reasons for knowledge acquisition difficulties find their basis in philosophical assumptions made about the cognitive process. The assumptions are that human cognition consists of discrete elements which may be characterized (by some introspection) to produce an expert system rule; secondly, that the processing requirements for expert domain heuristics is equivalent to the processing requirements for a deterministic rule set. Suggestions for improving the knowledge engineering process are offered.