This research project focuses especially on (1) the reconstruction of common non-monotonic logics (adaptive logics as the logical component, combined with a preferential component); (2) spelling out adaptive logics that are useful in an A.I. context and demonstrate their advantages over the usual machinery (decision methods, etc.); and (3) study of the typical properties of argumentation that may be reconstructed within a formal framework in view of the dynamic proofs of adaptive logics.