Propositional Defeasible

UCT Honours Project in Knowledge Representation and Reasoning

Knowledge Representation and Reasoning is a field with Artificial Intelligence where knowledge and reasoning is modeled using formal logic. We focus particularly on defeasible reasoning, which is a form of reasoning that, in contrast to classical methods, allows us to model information that typically holds. For example, we can model statements such as ‘birds typically fly’. We use the KLM approach to defeasible reasoning which is based on a series of properties expressing expected characteristics of defeasible reasoning. The framework enables us to use rules to infer new information from our modeled information.

It is often useful to know why particular information is being inferred. For example, in our comic the one penguin believes (infers) it can fly because ‘birds typically fly’ and ‘penguins are birds’. The formal concept associated with this idea for defeasible reasoning is referred to as defeasible explanation. Currently, little work has been done on defining and computing defeasible explanations.

Research Aims

Our research aims to improve our theoretical understanding of defeasible explanation and propose algorithms for evaluating defeasible explanations within the KLM framework.