Catarina Dutilh Novaes (Groningen) gives a talk at the MCMP Colloquium (24 January, 2013) titled "Reasoning biases and non-monotonic logics". Abstract: Stenning and van Lambalgen (2008) have argued that much of what is described in the psychology of reasoning literature as `reasoning biases' can more accurately be accounted for by means of the concept of defeasible, non-monotonic reasoning. They rely on the AI framework of closed-world reasoning as the formal background for their investigations. In my talk, I give continuation to the project of reassessing reasoning biases from a non-monotonic point of view, but use instead the semantic approach to non-monotonic logics presented in Shoham (1987), known as preferential semantics. I focus in particular on the so-called belief-bias effect and the Modus Ponens-Modus Tollens asymmetry. The ease with which these reasoning patterns are accounted for from a defeasible reasoning point of view lends support to the claim that (untrained) human reasoning has a strong component of defeasibility. I conclude with some remarks on Marr’s ‘three levels of analysis’ and the role of formal frameworks for the empirical investigation of human reasoning.