Human online adaptation to changes in prior probability.

Area of Science:

• Cognitive Neuroscience
• Decision Science
• Computational Neuroscience

Background:

• Optimal decision-making integrates uncertain sensory data with prior expectations.
• Prior probability shifts influence decision criteria, but tracking rapid probability changes remains unclear.

Purpose of the Study:

• Investigate observers' ability to track sudden changes in category probability.
• Determine the cognitive mechanisms underlying belief updating in changing environments.

Main Methods:

• Utilized a change-point detection paradigm with orientation-categorization tasks.
• Implemented a sample-and-hold procedure to update category probabilities.
• Developed and compared ideal Bayesian and heuristic models of observer behavior.

Main Results:

• A model combining exponential averaging with a bias towards equal priors best explained the data.
• This suggests a conservative bias and a stable, long-term equal-probability prior.
• A flexible Bayesian change-point detection model with incorrect beliefs also provided a good fit.

Conclusions:

• Observer decision criteria adapt to changing probabilities through a combination of on-line estimation and stable priors.
• This mechanism operates across different timescales, reflecting both rapid adaptation and long-term stability.
• Findings offer a biologically plausible explanation for how decision criteria are updated in dynamic environments.