FYI - defeasible reasoning with knowledge graphs

The Semantic Web is closely associated with description logics, e.g. for the interpretation of OWL schemas.  Everyday knowledge is however often uncertain, imprecise, incomplete, inconsistent and changing, i.e. imperfect knowledge. That is a challenge for logic and deductive proof which essentially relies on perfect knowledge.

Defeasible reasoning, by contrast, deals with plausible arguments, exploiting prior knowledge in regard to the likelihood of inferences. You can have arguments in support of, or counter to the supposition in question. Conclusions may need to be withdraw in the light of new information.  Argumentation has been studied by a long line of philosophers since the days of Ancient Greece, and is ubiquitous in human society, e.g. in courtrooms, ethics, politics and everyday discussion.

You may be interested in the talk and paper I recently presented at the Knowledge Graph and Semantic Web Conference, which this year took place in Zaragoza, Spain.  I introduce the plausible knowledge notation (PKN) along with a web-based demonstrator. PKN covers properties, relationships, contextual scopes, implications, fuzzy ranges, fuzzy modifiers, fuzzy quantifiers, structural analogies, metadata denoting gut feelings in lieu of statistics, and statements about statements.

Slides: https://www.w3.org/2023/11/12-Raggett-defeasible-reasoning.pdf

Paper: http://arxiv.org/abs/2309.12731

Comments welcomed!

Dave Raggett <dsr@w3.org>