Uncertainty plus prior equals rational bias: an intuitive Bayesian probability weighting function.

Area of Science:

• Cognitive Psychology
• Decision Theory
• Behavioral Economics

Background:

• Empirical studies reveal systematic distortions in how people evaluate probabilistic choices.
• These distortions include overestimating low probabilities, underestimating high probabilities, and differential treatment of positive/negative outcomes.
• Existing models like prospect theory use nonlinear probability weighting functions to capture these effects.

Purpose of the Study:

• To propose a novel Bayesian framework for understanding the psychological basis of probability weighting.
• To demonstrate how Bayesian inference can account for observed decision-making biases.
• To develop a probability weighting function that is both robust and efficient.

Main Methods:

• Developed a Bayesian model integrating prior information with probability statements via Bayes' rule.
• Analyzed the effects of informative (experience-based) and uninformative (ignorance-based) priors on probability weighting.
• Employed Bayesian model comparison to adaptively combine different prior types.
• Used Internet blogs to estimate priors for generic positive and negative outcomes.

Main Results:

• Any reasonable prior distribution on probabilities inherently leads to overweighting low probabilities and underweighting high probabilities.
• Combining informative and uninformative priors using Bayesian model comparison yields adaptive strategies.
• The proposed Bayesian probability weighting function accurately replicates key empirical findings on decision-making distortions.

Conclusions:

• A Bayesian approach provides a coherent psychological rationale for probability weighting.
• The adaptive combination of prior knowledge and uncertainty offers a robust and efficient decision-making mechanism.
• This framework successfully explains major characteristics of human judgment under risk.