Diagnostic Interference: People's Use of Information in Incomplete Bayesian World Problems

Abstract

Probabilistic inference word problems require people to integrate three types of information concerning a hypothesized cause of an event--base rate information p(H) concerning the relative frequency of the cause in question, evidence E that the cause was responsible, and the reliability of that evidence p(E/H)--and to evaluate the probability p(H/E) that the cause in question was responsible for the event. Three classes of hypothesis are proposed to explain how people answer these word problems and why the answers often neglect the base rate information--normative probabilistic reasoning, heuristic strategies,and nonnormative information integration. In a questionnaire study, 265 students estimated the probability of the cause before and after each type of information was presented. Information was presented in 6 orders, so some subjects responded to each possible subset of the information. Findings include the following: (1) many subjects respond with numbers that are available in the problem presentation; (2) the more recent information has a greater impact; (3) there is no universally applied weighted averaging scheme that accounts for the average response in all conditions; and (4) the typical subject's responses are well described in terms of the use of strategies contingent on the kind of information that is available. (kr)

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jul 01, 1990
Accession Number
ADA227678

Entities

People

  • Robert M. Hamm

Organizations

  • University of Colorado Boulder

Tags

Communities of Interest

  • Biomedical
  • C4I
  • Ground and Sea Platforms

DTIC Thesaurus Topics

  • Accuracy
  • Causal Reasoning
  • Classification
  • Cognition
  • Cognitive Science
  • Information Processing
  • Probability
  • Psychology
  • Reasoning
  • Reliability
  • Social Sciences
  • Statistics
  • Students
  • Test And Evaluation
  • Theorems
  • Thinking
  • Training

Readers

  • Artificial Intelligence
  • Organizational Psychology.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference
  • AI & ML - Information Retrieval