Handling Probability and Inconsistency in Answer Set Programming

Answer Set Programming (ASP) is a powerful declarative computing paradigm that is especially suitable for modeling commonsense reasoning problems. However, the crisp nature of the underlying semantics, the stable model semantics, makes it difficult to handle reasoning domains involving probability and inconsistency. To address this issue, we present an extension of logic programs under the stable model semantics, where rules are associated with weights. Under our semantics, probabilistic commonsense domains where inconsistency might be involved can be represented in an intuitive and elaboration tolerant way. Our semantics extends MLN and logic programming under stable model semantics. We have shown that probabilistic action domains and Pearl’s probabilisitic causal models can be represented, and various existing probabilistic logic programming frameworks can be embedded in our language. Future work includes further investigating the property of this language, devising algorithms for inference and learning in our language, and exploring various possible extensions of our language.