4. Inference engine

This is the part of the expert system that works out a reasonable solution when given a set of answers from the user.

The simplest inference engine is a straight forward decision tree. Such as the one below for diagnosing a faulty broadband connection

You may have experienced this kind of question and answer when phoning a help desk. You may feel infuriated at being asked such basic questions, but the assistant is simply running through their diagnostic tree.

Quite often, the user cannot answer with a Yes-No. To cope with this, either the decision tree becomes more complicated or the inference engine uses techniques such as 'heuristics'.

Heuristic systems are designed to work with uncertainty or provide answers based on experience. Humans call them 'rules-of-thumb', 'common sense', 'educated guess'. For expert systems, this includes making use of artificial intelligence, fuzzy logic and backward chaining algorithms.