Explaining Defeasible Reasoning

Defeasible reasoning is a form of non-monotonic reasoning that allows conclusions to be revised when new or more specific information is introduced, making it possible to represent exceptional information. Although it has become increasingly important in AI, the algorithms used to compute defeasible entailment are complex, and understanding how a conclusion follows from a defeasible knowledge base is challenging for newcomers. Existing pedagogical tools focus on classical monotonic reasoning, while existing defeasible reasoners are designed for research use and typically report only the final entailment result. This project presents PDR-Tool, a web-based pedagogical tool that works like a debugger for defeasible reasoning within the Kraus, Lehmann and Magidor framework. Given a knowledge base and a query, users can step through the entailment process for four algorithms: Rational Closure, Lexicographical Closure, Basic Relevant Closure and Minimal Relevant Closure. Each step exposes the intermediate computations alongside the formal algorithm definition and a plain-English explanation. The system's correctness was validated using worked examples from the literature and comparison against existing implementations, and its faithfulness to the algorithm definitions was assessed through expert evaluation. PDR-Tool aims to bridge the gap between classical logic education and defeasible reasoning, making defeasible entailment accessible without research-level expertise.