Catalogue of Artificial Intelligence Techniques
Categories: Inference and Reasoning , Problem Solving
Author(s): Robert Kowalski
Abduction was identified by the philosopher Charles Pierce as an especially important form of hypothesis formation. In the simplest case it has the form: from `A' and `B implies A' infer `B'. The abductive hypothesis `B' can be regarded as a possible explanation of `A'. To be a useful hypothesis, `B' should be consistent with other beliefs. Abduction is non-monotonic and non-deterministic. For example, from `A', and `B implies A', and `C implies A' we can infer `B' or we can infer `C' as possible alternative hypotheses. Adding `not B' as a new belief, non-monotonically withdraws `B' as a hypothesis. In comparison to this logical view, some researchers like Reggia have taken Peirce's original idea and interpreted it in terms of set coverings. Abduction has recently gained popularity in Artificial Intelligence for such applications as fault diagnosis, language understanding and image understanding. It has also been proposed as an alternative to Non-monotonic Logics for Default Reasoning.
- Paul, G., Approaches to abductive reasoning: an overview Artificial Intelligence Review 7 (1983), 109--152.