Democratizing diffusion decision models: A comprehensive tutorial on developing, validating, and fitting diffusion decision models in Python with FlexDDM

Diffusion decision models (DDMs) are frequently used tools in the cognitive sciences for modeling decision-making processes. However, conventional DDM software often falls short in flexibility and accessibility, limiting researchers' ability to explore new formulations of the model. To address barriers of entry, we present the Flexible Diffusion Decision Model (FlexDDM) Python package, a versatile and user-friendly software package designed to simplify the development, testing, and validation of DDMs. FlexDDM allows for extensions of the standard DDM that would otherwise not be possible with software reliant on traditional likelihood-based inference, while being simpler to implement programmatically than currently available likelihood-free options. In this tutorial, we provide an overview of FlexDDM’s unique features for efficient simulation, followed by a comprehensive guide on how to write and customize models in Python. We then demonstrate the package's fitting and model validation routines through an empirical analysis of conflict resolution tasks. Our analysis of six different datasets reveals that models incorporating motivation-based factors outperform contemporary DDMs in five of the six datasets, offering superior explanations of individual differences in cognitive control. These results underscore the utility of FlexDDM in advancing cognitive modeling by making it easier for researchers to experiment with and apply sophisticated modeling techniques, ultimately contributing to a deeper understanding of decision-making processes.