Expert Systems in Decision Support

R.B.Jones, 82/11/1


2.l Academic usage

The term 'expert system', in its recent usage, was coined by the academic community engaged in research into Artificial Intelligence(AI) to describe their increasingly successful attempts to exploit the techniques of AI in practically useful systems. Expert systems developed in these environments are many and varied. The most successful systems typically provide an expert problem-solving or advice-giving capability at a level which requires considerable knowledge of and experience in the problem domain, and which matches or exceeds the capability of human experts. The interfaces to these expert systems are usually suitable for use by people with no computing experience, though since they are often intended to assist experts in the field they may assume an understanding of specialist terminology in the problem domain.

It is an important characteristic many of these of these systems that they incorporate generalised problem solving capabilities and there is at least a conceptual distinction between the problem solving capability and the knowledge base on which this operates. This knowledge base supplies (domain specific and general knowledge essential to expertise, but without the inference system which operates upon it the solution of problems with this information would not be possible. Thus the expert system can be decomposed into interfacing and inference capabilities which are relatively domain independent, and a knowlelge base containing the necessary domain specific 'knowledge'. This decomposability of expert systems is exploited in knowledge engineering systems such as SAGE and ICLX, which are intended to allow a user with expertise in the problem domain but limited knowledge of the techniques of AI to construct an expert system by suppplying the necessary information to build up the knowledge base.

Some of the important problem areas in expert systems are:

These indeed form central themes in the Japanese fifth generation project, in which special hardware is poroposed for each of these problem areas. For ICL, however, a problem area of greater importance is that of integrating knowledge engineering systems and knowledge engineering techniques into the existing product range, in such a way that full access to and optimal utilisation of system facilities and resources is poss-ible through expert subsystems without impacting upon the efficiency of more traditional methods.

2.2 The inevitable future dilution

Once a market for 'expert systems' is established, it will inevitably be the case that manufacturers and software houses will use the labe I as widely as they credibly can. Since customers will not be familiar with academic usage, they will learn to understand by the term 'expert system' the sort of product which is typically dished up under that label, which will often be no more than a cosmetic adjustynent to last years packages, and will normally be much less 'intelligent' than the better products of university research projects. This trend is quite normal and largely harmless, The most unfortunate consequence is that it makes it more difficult to understand what we mean by 'expert system', do we mean an even more user friendly package, or do we mean the real thing?

2.3 Japanese usage

There can be no doubt that the aim of the Japanese fifth generation project is intended to implement oroducts which radically exceed the capabilities of existing bona-fide expert systems. The architecture of the proposed fifth generation computers is entirely built around the features of academic expert systems, to the extent that language executed by the processors will be based on a language hitherto confined to academic circles (PROLOG) and the targets for performance of those processors are expressed in 'Lips' (Logical Inferences per second).

Though we, may be sceptical about whether the Japanese will achieve all their targets, it would be unrealistic to suppose that they will fail to improve upon systems which are already operational on existing hardware and so we may be sure that some suppliers will be offering genuine expert system.

2.4 ICL usage

If our competitors are generally marketing a wide range of products as expert system it would be foolish, and indeed misleading of ICL not to follow popular usage. Broadening of the application of the term should not however be permitted to result in a lowering of our objectives. It is important that we do not allow ourselves to fall behind in this important area. It may prove necessary to invent a new terminology for bona-fide expert systems. Good expert systems will however be able to sell themselves, whatever the terminology.

up home © RBJ dated 1982/11/1 HTML 1996/6/6 edited 1997/4/18