EUROCORES_LogiCCC LcpR_Modeling human inference within the framework of probability

Till the end of the twentieth century the psychology of reasoning and human inference took classical logic as a standard frame of reference. Classical logic determined the tasks that were investigated experimentally, the evaluation of the performance of the subjects in the experiments, and the way in which the results were interpreted and discussed. In the last years, however, the importance of non-classical approaches, such as probabilistic (Bayesian) approaches, grows rapidly. The research proposal will join this theoretical shift of paradigms. The theoretical work will be based on non-classical approaches to logic, especially on probability logic and nonmonotonic reasoning. The empirical part will include a series of psychological experiments (1) on (if-then) conditionals, (2) on different argument forms, (3) on rules of system P, (4) on the retraction of conclusions in the context of nonmonotonic inference, and (5) on the conjugacy principle. They will try to clarify whether rather general principles of deductive reasoning (like truth functionality or monotonicity) do hold for human reasoning. We will develop a model of the cognitive representations and processes involved in the solution of these tasks. The project will continue my work on Bayesian networks and extend the already existing cooperation with Radim Jirousek. Recently, Bayesian networks gained considerabe importance in the psychology of causal reasoning. The project will profit from the expertise of Angelo Gilio, especially on problems of coherence, on imprecise probabilities, and on system P. Our own experience with Bayesian networks will be of help to Max Kistler, Josef Perner and Sarah Beck. Our game theoretic simulations (together with Gerhard Schurz) will compare the efficiency of argument forms in different logical systems. They will contribute to the recent discussions on the role of evolutionary adaptiveness in human cognition.

Psychological investigations of human reasoning have a specially close relationship to logic. For more than one hundred years psychologists used classical logic as a framework for experiments and theories on human reasoning. Psychologists studied, for example, how good we are using rules like \If A then B; A is true", \therefor B is true" or, more tricky, \If A then B; B is not true", \therefor A is not true".Everyday reasoning, however, involves uncertainty. The project introduced probabilities to describe human reasoning. It contributed substantially to the present shift in psychology from using classical logic to using probability logic as a benchmark of the rationality of human reasoning.In a series of experiments we have shown that human reasoning is based on probabilistic inference rules. A special role plays the human understanding of if-then sentences in such rules. We have shown that the human understanding of if-then sentences follows the same principles as conditional probabilities.This holds not only for adults but also for children. Moreover, the understanding of if-then sentences is related to the capacity of the short term memory.Probabilistic reasoning is essential when we seek new information. In a number of theoretical contributions we have shown that more information may lead to less precise conclusions {a typical example of \less is more". This is of practical importance in optimizing information seeking in a medical decision making.