How can we find some reasoning that people do, for which a particular logic is the appropriate model?

Keith Stenning (Edinburgh) gives a talk at the MCMP Colloquium titled "How can we find some reasoning that people do, for which a particular logic is the appropriate model?". Abstract:The psychology of deductive reasoning has extensively investigated tasks which it has supposed classical logic to be the appropriate standard of reasoning for, and the goal that subjects adopt in its laboratories. \cite{stvl08:book} presents evidence for a range of these tasks that a substantial proportion of subjects do not adopt this goal, and that defeasible logics (specifically Logic Programming) provide better models of their understanding of what they are supposed to do. Apart from issues about the `fit' of defeasible logics to the data of existing tasks, this argument points to the question whether untrained people do ever adopt classical logic as a model, and if so how to find tasks where this adoption is maximised and most clearly contrasted with other possible logical models. This talk will briefly outline the narrowness of the tasks that have been used, and propose a method of designing better. Results from a pilot experiment will illustrate some initial progress, and perhaps throw some light on general issues about how to match empirical evidence to logical model.