The drift diffusion model as the choice rule in reinforcement learning.

Area of Science:

• Cognitive Neuroscience
• Computational Psychiatry
• Decision Science

Background:

• Traditional reinforcement-learning models oversimplify decision-making processes.
• Sequential-sampling models capture choice accuracy and response time but assume static decision values.
• A gap exists in modeling dynamic choice processes within reinforcement learning.

Purpose of the Study:

• To integrate reinforcement-learning and drift diffusion models for a more comprehensive understanding of decision-making.
• To capture both within- and across-trial dynamics in choice processes.
• To investigate the utility of this integrated model in explaining stimulant medication effects in adult ADHD.

Main Methods:

• Implemented reinforcement-learning models where the drift diffusion model describes the choice process.
• Utilized hierarchical Bayesian parameter estimation to fit data from a reinforcement-learning paradigm.
• Compared model variants to assess their ability to capture effects in adult ADHD patients.

Main Results:

• The best-fitting integrated model successfully described learning, choices, and response times.
• Hierarchical Bayesian modeling demonstrated accurate estimation of model parameters.
• The model captured effects of stimulant medication in adult ADHD patients.

Conclusions:

• The combined reinforcement-learning and drift diffusion model offers a powerful framework for studying learning and decision-making.
• This approach provides new insights into cognitive and neural mechanisms underlying these processes.
• The model shows promise for understanding alterations in learning and decision-making in clinical populations like ADHD.