CCNCS Seminar Details
A Probabilistic Approach to the Psychology of Conditional Reasoning [Joint with Reasoning Group]
|Speaker:||Prof Mike Oaksford|
|Date/Time:||Saturday 19 February 2011, 3.30pm|
Experimental research on human conditional reasoning has questioned whether people are rational compared to a particular logical standard, the material conditional. In this talk, I argue that this data can be better understood at the computational level from a probabilistic point of view, i.e., from the perspective of inductive logic. Some old and some new data are reviewed and a revised model presented to deal with the lack of fit highlighted by Oberauer (2006) which incorporates possible failures of the rigidity condition on Bayesian conditionalization (Oaksford & Chater, 2008). The remainder of the talk focuses on the algorithmic level. First, a preliminary dual process implementation of the probabilistic approach is presented (Oaksford & Chater, 2010), which identifies possible roles for other algorithmic theories such as mental models. Second, the results of recent experiments are reported contrasting the predictions of a Causal Bayes Net instantiation of the probabilistic approach and mental models theory for discounting and augmentation inferences in causal conditional reasoning (Ali, Chater, & Oaksford, 2011). The talk concludes with a tri-partite proposal for the cognitive architecture of reasoning, in which local models are constructed in WM with reference to global world knowledge in LTM which are then interrogated by control processes in the central executive.