Evaluation of Evidence in Causal Inference.
Abstract
In three experiments, we investigated what evidence people use in making inferences about causality in complex and uncertain situations. Given evidence consisting of multiple observations of some outcome, with each observation including information about the presence or absence of that outcome and of some of its possible causes, subjects estimated the strength of the causal relationship between the outcome and a predetermined possibly-causal event. Over problems and over experiments, the nature and strength of evidence supporting the causal role of the hypothesized cause varied along many dimensions. Using regression-modeling, we found a set of five evidence types that together gave a good account of subjects' judgments. Four of the independent variables in this model directly concern the relation between the hypothesized cause and the outcome (confirmation by Joint Presence and by Joint Absence of target and outcome, and disconfirmation by violation of sufficiency and of necessity of the target for the outcome), and the fifth represents the goodness of alternative causes as explanations for the outcome. Over the experiments, involving four groups of subjects and five sets of problems, this single linear model accounted for 84 to 90% of the variance in each problem-set. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1979
- Accession Number
- ADA079762
Entities
People
- Miriam W. Schustack
- Robert Sternberg
Organizations
- Yale University