Valence-Dependent Belief Updating: Computational Validation.

Area of Science:

• Cognitive Psychology
• Computational Neuroscience
• Behavioral Economics

Background:

• Belief updating is influenced by both motivational desires and cognitive factors.
• Previous research suggests people favor good news over bad news when updating beliefs.
• The precise mechanisms and extent of this valence-dependent updating require further investigation.

Purpose of the Study:

• To computationally model and test for valence-induced biases in belief updating.
• To control for cognitive variables like prior beliefs and knowledge in belief updating.
• To compare Bayesian and reinforcement learning models of belief updating.

Main Methods:

• Employed computational modeling, comparing biased and unbiased Bayesian models and reinforcement learning models.
• Conducted an experiment with 80 adverse life events, assessing participants' risk estimations before and after receiving base rate information.
• Quantified belief updates and analyzed valence-dependent updating by comparing good news and bad news scenarios.

Main Results:

• Participants showed smaller belief updates and greater deviation from rational predictions following bad news compared to good news.
• A biased (optimistic) Bayesian model provided a better fit to the data than an unbiased model.
• Reinforcement learning models indicated higher learning rates for good news and highlighted the role of personal knowledge.

Conclusions:

• The study provides computational validation for a genuine optimism bias in human belief formation.
• Motivational factors, specifically the desire for favorable conclusions, demonstrably bias human judgments.
• Reinforcement learning models, in this context, offered a superior explanation of belief updating compared to Bayesian models.