Backward chaining is a reasoning method that starts from a specific goal or hypothesis and works backward through a set of rules to determine whether the known facts actually support it. Rather than exploring everything that can be inferred from available data, the system asks a targeted question, "does the evidence support this particular conclusion?" and only follows the chains of rules relevant to answering it.
This goal-directed approach makes backward chaining efficient for verification tasks, such as medical diagnosis systems that test a specific suspected condition against a patient's symptoms, or troubleshooting tools that check whether a hypothesized fault explains an observed failure. It's the mirror image of forward chaining, which starts from facts and works toward whatever conclusions they support; the choice between the two generally comes down to whether the system already has a hypothesis to test or needs to discover what the data implies.
