The MYCIN reasoning program used an approach called backward chaining; whenever a rule was being considered such that the system did not know whether the condition on the left-hand side of the rule (i.e., the premise) was true, MYCIN would look backward to see whether the knowledge base contained any other rules that, when evaluated, could conclude information that might inform the evaluation of the current rule’s premise (Nearly all contemporary rule-based systems, on the other hand, use an inference method known as forward chaining: Whenever a production rule “fires” and the conclusion of that rule is proven to be true, the system looks forward for other rules in the rule base that could also might be able to fire now that the new conclusion is known to be true).
- The rules often formed a coherent explanation of MYCIN’s reasoning—those that applied to the current decision were displayed in response to a user’s questions. Although rules were stored in a machine readable format, English translations could be