A diffusion-based framework for modeling systematic, time-varying cognitive processes.

Area of Science:

• Cognitive Psychology
• Computational Neuroscience
• Decision Science

Background:

• Cognitive models typically use static parameters, failing to capture dynamic psychological changes during tasks.
• Practice, learning, and boredom systematically alter cognitive states over time.
• Existing models do not adequately integrate temporal dynamics into decision-making frameworks.

Purpose of the Study:

• Introduce and validate the Parameters Across Time Diffusion Decision Model (ParAcT-DDM) framework.
• Model time-varying changes in diffusion decision model parameters (drift rate and threshold).
• Assess the performance of ParAcT-DDM against standard diffusion models using empirical data.

Main Methods:

• Developed the ParAcT-DDM, allowing diffusion model parameters to vary across time (trial- or block-varying).
• Focused on modeling changes in drift rate (efficiency) and threshold (caution).
• Empirically tested ParAcT-DDM variants on four existing datasets, including data with pre-experimental practice.

Main Results:

• ParAcT-DDM variants significantly outperformed the standard diffusion model across all tested datasets.
• Evidence suggests time-varying cognitive processes occur even in experiments designed to minimize practice effects.
• The standard diffusion model yields biased parameter estimates when time-varying processes are present.

Conclusions:

• The ParAcT-DDM framework offers a more robust approach to modeling cognitive processes by incorporating temporal dynamics.
• Accounting for time-varying parameters is crucial for accurate inferences in cognitive and decision-making research.
• ParAcT-DDM enhances the reliability of cognitive modeling by addressing systematic biases caused by unmodeled temporal changes.