One of the two main methods of reasoning when using inference rules. The other is forward chaining. Backward chaining starts with a list of goals and works backwards to see if there is any data available that will support concluding any goal. An inference engine using backward chaining would search the inference rules until it finds a rule which has a Then clause that matches a desired goal. When the If clause of that inference rule is not known to be true, then it is added to the list of goals.

For example, suppose that the goal is to conclude when to terminate a misbehaving employee, given that he has been late three times in the past month and has been employed less than 6 months, and that the rule base contains the following rules:

If employee is late more than twice in a given month and the employee is a probationary employee, Then terminate employment

If employee has been employed less than six months, Then employee is a probationary employee

This rulebase would be searched and the first rule would be selected, because its conclusion (the Then clause) matches the goal (terminate employment). It is not yet known that the employee is on probation, so the If statement is added to the goal list (in order to terminate the employee must be on probation).