International Joint Conference on Artificial Intelligence · 1989 · 246 citations · 8 references
ComputingComputer ScienceCausal InferenceFormal MethodsKnowledge RepresentationAutomated ReasoningDeductive ReasoningEvent RepresentationExplanationTemporal Knowledge RepresentationTemporal LogicAbductive InferenceAbductive ReasoningCausal LawsEvent CalculusPersistenceTemporal Causal ReasoningTemporal Logic ReasoningTemporal Reasoning
Causal laws are argued to be best expressed in the natural form “effect if cause.” The paper proposes a temporal reasoning framework where prediction is deduction and explanation is abduction, aiming to achieve default persistence that resolves issues in prior models. The framework employs a shortened Event Calculus, assumes that known properties require event‑based explanations, and extends SLD resolution to implement abductive explanation alongside deductive prediction. Applying abduction under this assumption yields default persistence that successfully addresses problems that have troubled other formulations.
This paper presents an approach to temporal reasoning in which prediction is deduction but explanation is abduction. It is argued that all causal laws should be expressed in the natural form effect if cause. Any given set of laws expressed in this way can be used for both forwards projection (prediction) and backwards projection (explanation), but abduction must be used for explanation whilst deduction is used for prediction. The approach described uses a shortened form of Kowalski and Sergot's Event Calculus and incorporates the assumption that properties known to hold must have explanations in terms of events. Using abduction to implement this assumption results in a form of default persistence which correctly handles problems which have troubled other formulations. A straightforward extension to SLD resolution is described which implements the abductive approach to explanation, and which complements the well-understood deductive methods for prediction.
8
A logical framework for default reasoning
David Poole · Artificial Intelligence · 1988 · 729 citations
Nonmonotonic logic and temporal projection
Steve Hanks, Drew McDermott · Artificial Intelligence · 1987 · 441 citations
Henry Kautz · 1986 · 119 citations