Causal Modeling, Causal Inference, and Causal Perception

Indiana University, Bloomington IN

Investigators

Abstract

SES 99-06565 - Charles R. Twardy (Indiana University) "Causal Modeling, Causal Inference, and Causal Perception" Under the auspices of this grant, the Principal Investigator will conduct a two-year training and research fellowship at Monash University (Clayton, Victoria, Australia) to learn the techniques of Bayesian network modeling so that he can extend his investigations into causality, causal inference, and scientific methodology. The Principal Investigator will employ Bayesian network models to represent human cognition in the context of argument generation and analysis, and to model how humans learn causal networks (including deviations from theoretically normative inference). He will then analyze the implications of this work for causal perception, causal judgment and causal reasoning in scientific inquiry. Bayesian network models are a unified and more powerful alternative to several mainstream analyses of causation: they naturally accommodate complex causal interactions and progressive causal inference. This proposal seeks to extend the domain of these models to provide a method for more thoroughly testing or analyzing human causal induction than has been available in the philosophical or psychological literature. The leading account of causality in the philosophy of science, CQ theory, may be able to provide an ontological solution to Hume's problem, but it does not deal with Hume's profound worries about human causal induction. Bayesian theory is one of the leading methods of dealing with inductive inference, and Bayesian networks are being employed specifically to represent causal networks. The Principal Investigator proposes to combine causal insights from Bayesian modeling theory, cognitive psychology, and the philosophy of science to provide a unified theory of human causal inference. This proposal involves training the Principal Investigator in the methods of Bayesian network modeling, and then applying Bayesian models in ways analogous to the work the Principal Investigator recently completed to combine psychological and philosophical accounts of causal perception and causal judgment. This will provide the Principal Investigator with the solid technical foundation for continuing philosophical inquiry in causation and methods of scientific discovery.