The role of argument comparison in dialectical argumentation

There are a lot argumentation models thal have been developed inside Artificial Intelligence. Among these models, differents formal systems of defeasible argumentation are defined, where arguments for and against a proposition are produced and evaluated to verify the acceptabilily of that proposition. In this manner, defeasible argumentation allows reasoning with incomplele and uncertain information. The development of this kind of systems has grown in the last years [SIM92, BART, KOWA96, AG97, DUNG93, DUNGLP] but no consensus has been reached yet on some issues, such as the representation of arguments, the way they interact, and the output of that interaction. Even then, the main idea in these systems is that any proposition will be accepted as true if there exist an argument that supports it, and this argument is acceptable according to an analysis between it and its counterarguments. Therefore, in the set of arguments of the system, some of them will be "acceptable" or "justified" arguments, while others not. But this bi-valued classification" arguments is not enough, due to some situations that can be found in argumentation systems. The reasons of non-justification can be analyzed in more detail, so we can make a more specific classification the non-justified arguments. An argument of this kind can not be justified because, for instance, it has a justified defeater, it is involved in circular argumentation. In the former, we can think that the argument has been effectively defeated. In the lalter, the juslification of the argument falls in an "inconclusive" state. This is the starting point lo distinguish a third kind of arguments: those which left the dispute without any conclusion. There exist various names for this argulments, like defendibles, undecided, ambiguous and undetermined. In the rest the paper, we will call this arguments undecided. There is another reason to classify an argument as undecided. This reason is not so obvious as the one specified above, and is related to the comparison of arguments. Eje: Aspectos teóricos de inteligencia artificial Red de Universidades con Carreras en Informática (RedUNCI)