Defeasible Reasoning


Defeasible Logic was introduced by Donald Nute in a 1994 seminal paper. The basic idea behind defeasible reasoning is that logic rules may contradict each other without introducing inconsistency. Instead, priorities among rules can be defined with the idea that inferences made by higher-priority rules are more important and, under certain conditions, might "defeat" inferences made by lower-priority rules.

This seemingly simple idea can have surprisingly many formalizations and quite a number of different versions of defeasible logic were proposed since 1994 -- all differing in some subtle ways that are hidden in the details of their proof theories. The version used in Flora-2 is called Logic Programming with Defaults and Argumentation (or LPDA). The main advantage of LPDA is that it provides a practical solution to the problem that different application domains often require different defeasible logics. However, defining and implementing a new proof theory each time is prohibitively expensive and unrealistic. In LPDA, however, the user can specify defeasibility conditions as an argumentation theory using a small number of natural logical rules. Different argumentation theories can be used together in the same knowledge base by putting them in separate modules, and this is how LPDA is implemented in Flora-2, which supplies several pre-defined argumentation theories.