Expert System - Disadvantages

Disadvantages

This section needs additional citations for verification.

The expert system has a major flaw, which explains its low success despite the principle having existed for 70 years: knowledge collection and its interpretation into rules, or knowledge engineering. Most developers have no automated method to perform this task; instead they work manually, increasing the likelihood of errors. Expert knowledge is generally not well understood; for example, rules may not exist, be contradictory, or be poorly written and unusable. Worse still, most expert systems use an engine incapable of reasoning. As a result, an expert system will often work poorly, and the project abandoned. Correct development methodology can mitigate these problems. There exists software capable of interviewing a true expert on a subject and automatically writing the rule base, or knowledge base, from the answers. The expert system can then be simultaneously run before the true expert's eyes, performing a consistency of rules check. Experts and users can check the quality of the software before it is finished.

Many expert systems are also penalized by the logic used. Most formal systems of logic operate on variable facts, i.e. facts the value of which changes several times during one reasoning. This is considered a property belonging to more powerful logic. This is the case of the Mycin and Dendral expert systems, and of, for example, fuzzy logic, predicate logic (Prolog), symbolic logic and mathematical logic. Propositional logic uses only invariant facts. In the human mind, the facts used must remain invariable as long as the brain reasons with them. This makes possible two ways of controlling the consistency of the knowledge: detection of contradictions and production of explanations. That is why expert systems using variable facts, which are more understandable to developers creating such systems and hence more common, are less easy to develop, less clear to users, less reliable, and why they don't produce explanations of their reasoning, or contradiction detection.

Read more about this topic:  Expert System