A Study About Abductive Reasoning Based On Argumentation Theory

Abductive reasoning is not only commonly adopted in daily life, but also plays an indispensable role in scientific discovery and some other areas. It is a kind of probable reasoning with the feature of non-monotonicity. The conclusion of abductive reasoning is defeasible. However, classical abductive models fail to capture this characteristic. To remedy this defect, the idea of this essay is to study abductive reasoning based on a method of non-monotonic reasoning: argumentation theory. The argumentation theory used in this essay is introduced from three parts: the presentation of argumentation systems (based on defeasible logic), argumentation semantics (based on extension) and the argumentation system semantics solving (based on argument game).This essay takes a simple commonsense reasoning and a study of an evidence in a legal case for examples. Based on argumentation theory, it attempts to depict the inner abductive mechanism, and preform reasonings. In this way, the essay tries to realize the aim of compensate the shortcomings of classical abductive models by the advantages of argumentation system.There are three steps to analyse the cases in this essay. First, according to the features of abductive reasoning, the analysis starts from the observations, combines it with related rules, then constructs arguments. Second, compare the conflict arguments according to the superiority relation and decide which one to be accepted. Arguments that cannot defend themselves should be rejected. Third, evaluate the arguments be accepted on step two based on the three criteria of abductive reasoning, and pick out the best explanation. There are still many respects requiring further investigated, for example, how to embed the criteria of abductive reasoning into the superiority relation of arguments, so that these two methods have a closer combination. This essay is so preliminary that it lacks in-depth research in many complicated problems yet. The essay points out some of the defects in the last chapter, as well as offers some proposals to future research.