Deductive and probabilistic reasoning

Princeton University, Princeton NJ

Investigators

Abstract

Almost all humans are capable of deductive, inductive, and probabilistic reasoning, even if they have had little or no training in logic or the probability calculus. This research project aims to discover the mental processes underlying their reasoning and to develop computer software that implements the same processes. There are two reasons for the project. First, the development of an integrated theory of reasoning elucidates a key component of human mentality. Second, a program that simulates reasoning is a major step towards fulfilling a long-standing dream ? going back to the seventeenth century ? of an automatic system for evaluating inferences and for identifying fallacies. The computer program would therefore be a tool for improving reasoning. The theory underlying the project is that reasoning depends on simulating events in mental models. The project begins with series of psychological experiments to test the theory?s new applications to several crucial domains, which include how the meaning of premises and general knowledge can yield ?kinematic? mental models that unfold in time, how certain descriptions lead to systematic but compelling fallacies, and how the subjective probabilities of unique events derive from evidence that is not numerical. The experiments contrast the model theory?s predictions with rival accounts. Kinematic models predict a trend in the time to make inferences, because earlier events in a temporal simulation should be easier to access than later ones. The research then uses the experimental results and prior data to formulate a model theory that unifies the different components of everyday reasoning. The studies of subjective probabilities fill a major gap in the model theory and in the theory of subjective probability: how reasons yield numerical probabilities. The unified theory will be implemented progressively in a computer program, which will allow users to determine what follows from what. It will embody a mental architecture for higher cognition organized around operations on mental models. All the software will be in the public domain.