[PDF]

Relevance-focused methods for reasoning with Conditional Knowledge Bases


Mihai Craciun

08/09/2025

Supervised by Richard Booth; Moderated by Víctor Gutiérrez Basulto

A conditional knowledge base (KB) is a set of statements of the form "if A, then normally/usually B". A classic question in Knowledge Representation has been "given a conditional knowledge base KB, which other conditionals should be said to "follow from" KB. Several solutions have been proposed over the years, including the preferential closure, the rational closure and, more recently, the disjunctive rational closure (DRC). The DRC is especially applicable in situations where we want our resulting set of conditionals to satisfy the KLM properties, plus the rule of disjunctive rationality, but *without* necessarily the rule of rational monotony. However the DRC treats some examples in unintuitive ways, and fails to capture certain properties that one might expect. This project is about defining and possibly implementing an alternative to DRC which does not suffer from the same flaws.


Final Report (08/09/2025) [Zip Archive]

Publication Form