Cue Utilization in a Numerical Prediction Task

Abstract

Forty subjects were trained to use scatter plots with regression lines to make numerical predictions of one variable (criterion) from another variable (cue). The subjects were trained on two separate cues, differing in validity. Later, the two cues were presented together, simultaneously for 20 subjects, successively for the rest. Subjects were asked to use both cues to predict the criterion. Instructions emphasized that the two cues were independent of one another, but did not specify how the two should be combined. Initial analyses indicated that a regression model provided an adequate fit to the data, that the subjects showed conservatism similar to the conservatism found in previous Bayesian inference studies. However, more molecular analyses indicated patterns of behavior which consistently deviated from the optimal model. The post hoc hypothesis that subjects were regressing each cue, then averaging the regressed values, was supported by the data for most subjects. Searching for heuristic strategies, rather than relying on the apparent fit of optimal models was advocated fur future research.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Feb 01, 1974
Accession Number
AD0775902

Entities

People

  • Paul Slovic
  • Sarah Lichtenstein
  • Timothy C. Earle

Organizations

  • Oregon Research Institute

Tags

Communities of Interest

  • Human Systems

DTIC Thesaurus Topics

  • Analysis Of Variance
  • Bayesian Inference
  • Bayesian Networks
  • Cognition
  • Conservatism
  • Consistency
  • Data Science
  • Factorial Design
  • Information Processing
  • Information Science
  • Judgment
  • Models
  • Motor Skills
  • Probability
  • Psychology
  • Regression Analysis
  • Surveys

Fields of Study

  • Psychology

Readers

  • Operations Research
  • Regression Analysis.
  • Speech Processing/Speech Recognition.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference