Working Backward

Origin. Pólya's problem-solving heuristics; Amazon's "Working Backwards" product process; goal-regression in AI planning.

Mechanism. Starts from the desired end state and asks what must be true immediately before it, then what must be true before that, recursing until reaching a state that is achievable from the present. Forward search can explore irrelevant branches; backward search is constrained by the goal.

Procedure. Describe the desired outcome in concrete detail. Ask: what is the last thing that must happen to achieve this? Then: what must be true for that to happen? Continue recursively. Connect the chain to current reality. The sequence from present to future is the plan; the sequence from future to present is the derivation.

Applies to. Planning, proof construction, debugging, any problem where the goal is clearer than the path.

Limitations. Works when the goal is well-defined and the steps are reversible in reasoning. For goals that are vague or for domains where what enables X is context-dependent, backward chaining generates an explosion of possibilities.