Defining Salience in Case-Based Arguments

A largely unrealized goal of AI has been to design systems that can tailor descriptions of knowledge base objects to suit alternative points of view and contexts. This problem is examined here with respect to case based argu merits, arguments in which past cases are cited as justifications for a decision. We define the features of a case that are salient in describing it from various viewpoints and in a variety of argument contexts employed in HYPO, an adversarial case based reasoning program. We demonstrate HYPO's ability to slant case descriptions from opposing viewpoints and to tailor the descriptions to fit contextual roles in an argument, including filing the case, distinguishing it, posing it as a counterexample or as the target of a hypothetical.