Expressions for Bayesian confidence of drift diffusion observers in fluctuating stimuli tasks

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Decision Science

Background:

• Decision confidence modeling is crucial for understanding cognitive processes.
• Existing models face challenges with trial-by-trial variability in fluctuating stimuli.
• Computational efficiency is key for practical application in complex tasks.

Purpose of the Study:

• To develop a computationally cheap approach for modeling decision confidence.
• To incorporate trial-by-trial variability in stochastically fluctuating stimuli.
• To derive practical expressions for confidence distributions.

Main Methods:

• Utilized the drift diffusion model framework with time-dependent thresholds.
• Incorporated a Bayesian confidence readout and pipeline evidence accumulation.
• Derived expressions for confidence distributions with normally-distributed stimulus fluctuations.

Main Results:

• Developed novel expressions for confidence probability distributions.
• Accounted for drift rate variability and metacognitive noise.
• Validated approximations through simulations, showing feasibility for trial-by-trial analysis.

Conclusions:

• The derived expressions offer a computationally feasible method for modeling decision confidence.
• This approach enhances understanding of confidence in tasks with fluctuating stimuli.
• Provides insights into optimal observer confidence and empirical patterns.