CBR

Case-based reasoning (CBR) is a problem-solving paradigm within artificial intelligence and cognitive science. It solves new problems by adapting solutions that previously worked for similar cases, rather than deriving solutions from general rules or models.

The typical CBR cycle is described as the four Rs: retrieve, reuse, revise, and retain. A relevant past case is retrieved from a case base, its solution is reused or adapted to fit the current problem, the outcome is revised as needed, and the new experience is retained to update the case base.

Case bases are stored collections of past situations, actions, and outcomes. Retrieval relies on similarity judgments, indexing, and domain-specific representations to locate the most relevant cases. Adaptation strategies modify the retrieved solution to account for differences in context, constraints, or goals.

CBR has roots in cognitive science and engineering research from the 1980s and 1990s, with notable work by Janet Kolodner and others. It has been applied across disciplines, including medicine, legal reasoning, engineering, and customer support.

Advantages of CBR include rapid problem solving when suitable cases are available, interpretability of the rationale by showing past cases, and continual improvement as the case base grows. Limitations include dependence on a comprehensive and well-annotated case base, the challenge of designing effective adaptation methods, and reduced effectiveness on problems with little prior experience.

Related concepts include analogical reasoning and retrieval-based AI. CBR frameworks often integrate with other AI methods, knowledge management systems, and decision-support tools.