Backward reasoning is used in reasoning systems when rules are run'backwards' in order to ask "what would need to be true in order to show X". That is, given a goal to find sub-goals that need to be solved, themsleves using backward reasoning. It is particularly useful when the backward branching factor is smaller than the forwards branching factor, for example wgere there are huge numbers of fact known combing them to generate derived knowledge through forward reasoning would generate vast numbers of anticident facts in the hope that one of them iis useful. In contrast only a small number of them will be relevant to particular goal. In an expert system backward reasoning can be used to suggest questiosn to ask; for example, given "all men are mortal", and the goal is to prove "Socrates is mortal", a backward reasoning system might ask "Is Socrates a man?"
Used on pages 39, 40, 347, 427, 428, 429
Also known as backward chaining, antecedent-driven, backward, consequent-driven, goal-driven
See also forward reasoning