Maintenance of expert systems: life-cycle validity

It is well known in the business computing community that program and data base maintenance are the largest component of the software system life cycle (Boehm, 1981). Little concern has been raised about the maintenance of knowledge bases of expert systems. The majority of interest in maintenance (which might be classified as being validity over the life cycle of the expert system (Mostow. 1985), and in the maintenance of consistency of a data base (Balzer et al. 1983). One commercial selling strategy of expert systems is that they do not change with time like human knowledge. A person who doesn't change with time however, is not necessarily an intelligent person. An expert system without maintenance is "an old fogey," unable to learn new things. In a preliminary study on now to implement a data base for critical evaluation of atomic and molecular data (Cann and Nicholls, 1980), it was decided to incorporate names of researchers, dates of data added as well as critical notes on methods and recommendations on results. The reasons for this were to be able to compare and update experimental data with time and with new experimental results. But, new results are not always better. Often, older values are obtained through more thorough study then some newer results. There are several different types of errors that can exist in data or knowledge bases.