Abstract

A new framework for modeling reasoning about complex systems is described by Pearl (2000) and Spirtes, Glymour, and Scheines (1993). The fundamental idea is that people represent the world by decomposing it into autonomous causal mechanisms such that human reasoning is understood as a tool to support human action. I describe experimental tests of the framework that examine one of its key assumptions - its representation of actual and counterfactual intervention - in order to evaluate its plausibility as a source of cognitive models. I will discuss how people make counterfactual inferences in the context of both deterministic and probabilistic causal arguments and contrast such reasoning with reasoning about conditional ("if-then") statements. I will also touch on how people learn about causal relations, focusing on the relative value of merely observing causal events as opposed to acting, intervening to modify the operation of the causal system.

Reference
Sloman, S.A. & Lagnado, D.  (submitted). Do We “do”? [doc pdf]

Steven A. Sloman is Associate Professor of Cognitive & Linguistic Sciences, Brown University