A flexible framework for simulating and fitting generalized drift-diffusion models.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Mathematical Psychology

Background:

• The drift-diffusion model (DDM) is a cornerstone for understanding decision-making processes.
• Methodological hurdles have historically constrained the development of novel DDM extensions.
• Existing DDM frameworks lack the flexibility to incorporate complex, user-defined functions.

Purpose of the Study:

• To introduce the generalized drift-diffusion model (GDDM) framework, enabling the creation and fitting of advanced DDM variants.
• To provide a software package that implements the GDDM framework for practical application.
• To facilitate innovation in decision-making models and experimental designs.

Main Methods:

• The GDDM framework extends traditional DDM parameters using arbitrary user-defined functions.
• Numerical solutions are obtained by directly solving the Fokker-Planck equation with efficient numerical methods.
• A significant speedup (≥100x) over standard methods is achieved, enabling maximum likelihood fitting on the complete response time (RT) distribution.

Main Results:

• The GDDM framework was successfully applied to fit both animal and human datasets from perceptual decision-making tasks.
• GDDMs demonstrated superior accuracy and required fewer parameters compared to existing state-of-the-art DDMs.
• The framework facilitates testing of complex decision-making mechanisms.

Conclusions:

• The GDDM framework significantly advances the flexibility and efficiency of decision-making modeling in cognitive neuroscience.
• This approach overcomes previous methodological limitations, paving the way for novel theoretical and experimental investigations.
• The provided software package democratizes the use of advanced DDM extensions for researchers.