Catalogue of Artificial Intelligence Techniques
Aliases: Preferential Entailment, Probabilistic Entailment
Author(s): Judea Pearl
A Non-monotonic Inference system in which defaults are interpreted as statements of high conditional probabilities, infinitesimally close to zero or one. The valid consequences are those statements that would acquire arbitrarily high probabilities whenever the probabilities assigned to the premises are made sufficiently high. Syntactical derivations require the solution of a series of propositional satisfiability problems. The system is suitable for maintaining consistency and enforcing specificity ordering, but does not support chaining and contraposition.
- Pearl, J., Probabilistic Semantics for Non-monotonic Reasoning: A survey
Proceedings of First Intl. Conf. on Principles of Knowledge Representation and Reasoning (KR'89)