Analyzing dynamic decision-making models using Chapman-Kolmogorov equations.

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Decision Theory

Background:

• Dynamic decision-making requires adaptive evidence accumulation, prioritizing recent information over older data.
• Experimental studies involve tasks where the correct choice changes unpredictably within a trial.
• Ideal observer models in these dynamic scenarios face doubly stochastic processes (observation and environmental changes).

Purpose of the Study:

• To represent the probability density of an ideal observer's belief in dynamic environments using differential Chapman-Kolmogorov equations.
• To enable efficient computation of ensemble statistics for normative and near-normative decision models.
• To provide metrics for comparing model performance, including decision accuracy and belief distribution divergence.

Main Methods:

• Utilized differential Chapman-Kolmogorov equations to model belief evolution in doubly stochastic environments.
• Employed decision response accuracy and Kullback-Leibler divergence as performance metrics for model comparison.
• Investigated the impact of internal noise on response accuracy and optimal integration timescales.

Main Results:

• Demonstrated that differential Chapman-Kolmogorov equations efficiently compute ensemble statistics for belief distributions.
• Showed that increased internal noise necessitates longer integration timescales for optimal decision-making.
• Validated the applicability of the method to tasks with discrete, pulsatile evidence arrival.

Conclusions:

• The proposed method offers an efficient way to analyze adaptive evidence accumulation in dynamic decision tasks.
• Belief distributions derived from this model can be empirically measured through decision confidence reports.
• The framework is versatile, applicable to various evidence-gathering scenarios, including discrete and continuous data streams.