A Bayesian view of covariation assessment

Area of Science:

• Cognitive Psychology
• Decision Making
• Bayesian Inference

Background:

• Traditional models of judgment, like the phi coefficient, interpret deviations in assessing variable relationships as errors.
• Observed phenomena include joint presence having a larger impact than joint absence, and prior beliefs influencing judgment.
• These observed patterns deviate from normative models, leading to their classification as systematic errors.

Purpose of the Study:

• To re-evaluate common judgment patterns in assessing relationships between variables with presence/absence levels.
• To demonstrate the consistency of these patterns with a Bayesian approach rather than normative errors.
• To investigate the role of prior beliefs and variable rarity in judgment formation.

Main Methods:

• Participants assessed relationships between variables presented with presence and absence levels.
• Analysis focused on the impact of joint presence versus joint absence on judgments.
• Prior beliefs and the relative rarity of variable presence/absence were considered.

Main Results:

• Joint presence significantly impacted judgments more than joint absence, consistent with prior observations.
• However, empirical evidence showed joint absence was perceived as more informative when absences were rare.
• Participant judgments were found to be influenced by their pre-existing beliefs about variable relationships.

Conclusions:

• Observed judgment patterns are not necessarily errors but are consistent with Bayesian principles.
• The perceived informativeness of joint presence versus absence depends on their relative rarity.
• Incorporating prior beliefs is a normative aspect of Bayesian reasoning in judgment tasks.