Semantic Intuitions in Statistical Causal Reasoning

Abstract

In psychology, philosophy, and linguistics there has been a debate about two competing frameworks of causal reasoning. Dependency theories, especially causal Bayes nets, focus on causally motivated statistical or counterfactual dependencies between events (causes and effects). In contrast, force dynamic theories implement causation as arising (deterministically) from force interactions involving agents impinging on the prior tendencies of patients. To date force dynamic theories have primarily focused on the representation of different semantic causal concepts in scene descriptions. Our goal is to bring the two competing frameworks together. We will present a model that implements the interaction between agents and patients in terms of probabilistic forces. We have tested this new model in an experiment in which we tested how contingency information interacts with the assumptions about intrinsic tendencies of patients in people’s usage of semantic causal concepts (e.g., CAUSE, PREVENT, HINDER, HELP, ALLOW, and ENABLE).


Back to Table of Contents