Argumentation In Flux (Modélisation du changement dans la théorie de l'argumentation)

Abstract argumentation frameworks are a widely used formalism in the field of artificial intelligence. They are used to represent conflicting information by means of a set of arguments and an attack relation. The main problem studied in the literature is their evaluation, i.e., the determination of the justified points of view on the status (accepted or not) of the arguments. The research in this thesis is motivated by the idea that this is not a static process, and that there are many real life examples in which external information plays a role. We address this issue from three points of view.First, we look at intervention and observation in argumentation. These are notions usually studied in the context of causal networks, which are structures used to encode causal connections between events. In these models, an intervention represents the active causation of an event in the interest of predicting the effects, while the passive observation of an event allows one to infer both the most likely causal explanation as well as the effects. In argumentation, intervention captures a hypothetical mode of arguing, where we hypothetically fix the status of an argument in the interest of determining the effects.An observation, on the other hand, captures a revision process: changing the status of an argument requires us to retrace our steps in the line of reasoning that led to the initial status and to accept the most likely hypothesis that explains the new status. We propose models for these two types of reasoning and analyze them using a postulate-based approach.Second, we develop a model of abduction in argumentation, where changes to an argumentation framework act as hypotheses to explain an observation. We present dialogical proof theories for the main decision problems (i.e., finding hypotheses that explain an observation) and show that this model can be instantiated on the basis of abductive logic programs.Third, we look at change in preference-based argumentation. Preferences have been introduced in argumentation to encode, for example, relative strength of arguments.An underexposed aspect in these models is change of preferences. We present a dynamic model of preferences in argumentation, based on what we call property-based argumentation frameworks. It is based on Dietrich and List's model of property-based preference and provides an account of how and why preferences in argumentation may change. The idea is that preferences over arguments are derived from preferences over properties of arguments and change as the result of moving to different motivational states. We also provide a dialogical proof theory that establishes whether there exists some motivational state in which an argument is accepted