Reasoning, Learning, and Classifying with Uncertain Causal Models

Abstract

The last 20 years have seen a growing interest in the role of causal knowledge in numerous areas of cognition. Many studies have investigated how causal relations are learned from observed correlations. Others have tested their impact on various forms of reasoning, including inference, interventions, decision making, analogy, and classification. The original goal of this proposal (intended to cover 3 years of research but funded for 18 months) was to study three aspects of human causal reasoning. The first aspect is how people reason causally under uncertainty, that is, when beliefs are held with less than complete confidence. We address a particular application in which a representation of uncertainty resolves how inconsistencies among beliefs are resolved. The second aspect is how individuals reason with "conjunctive causes," that is, when a cause only operates when it is accompanied by one or more other causes (e.g., a spark only yields fire when there is also fuel and oxygen). The third aspect concerns how people reason with inhibitory causes, that is, factors that disable or deactivate a causal mechanism. Research on the first two of these topics was conducted and is near completion, and is described in this paper.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Nov 19, 2012
Accession Number
ADA578263

Entities

People

  • Bob Rehder

Organizations

  • New York University

Tags

Communities of Interest

  • C4I

DTIC Thesaurus Topics

  • Abstracts
  • Bayesian Networks
  • Causal Reasoning
  • Cognition
  • Cognitive Science
  • Contracts
  • Judgment
  • Models
  • New York
  • Observation
  • Probability
  • Probability Distributions
  • Psychology
  • Reasoning
  • Social Psychology

Readers

  • Aerospace Test and Evaluation
  • Artificial Intelligence

Technology Areas

  • AI & ML