Probabilistic Logic for Knowledge Representation and Automated Reasoning

Conventional statements of logic (e.g., simple statements of the form “if x then y”) allow individuals and machines to make quick and efficient determinations of the state of the world through the rules of deduction. This type of reasoning, however, does not naturally accommodate a fundamental and irreducible aspect of our knowledge about the world: we are more often than not uncertain about our knowledge to some degree or another. Dealing with uncertainty requires using a probabilistic representation of reasoning that allows one to express and draw inferences in cases when the facts are uncertain rather than just true or false. The Decisive Analytics Corporation/MIT (DAC/MIT) team proposes a powerful and elegant method which combines the desired expressive power of conventional logic with a sound and consistent treatment of uncertainty, resulting in an automated reasoning engine that integrates logical relations with probabilistic reasoning about complex, imprecise, and uncertain situations. The proposed hybrid inference engine will moreover be capable of hypothesizing new attributes, new relationships, and even new types of objects in its representation space and thus yield more expressive capability than other statistical relational formalisms.