Applications and Evolution of Neuroeconomics and Reinforcement Learning Models in Substance Use Disorders

This chapter examines the application of neuroeconomics and reinforcement learning (RL) models in understanding substance use disorders (SUDs). Neuroeconomics combines insights from neuroscience, psychology, and economics to investigate decision-making processes, particularly in contexts where maladaptive behaviors like addiction emerge. RL offers a computational framework to analyze how agents adapt their behaviors through interactions with the environment, focusing on reward processing, value computation, and prediction error signaling. The chapter provides a detailed overview of key RL models used to study SUDs, including temporal difference RL, model-based and model-free approaches, and Bayesian frameworks. These models elucidate how decision-making processes are impaired in addiction, highlighting transitions from goal-directed to habitual behaviors, dysregulated reward systems, and altered risk-reward tradeoffs. RL-inspired frameworks also offer explanations for relapse and compulsive substance-seeking, emphasizing their utility in predicting addiction trajectories and informing interventions. Looking ahead, future directions in RL modeling emphasize integrating advanced machine learning techniques, such as generative AI and hierarchical RL, to capture the complexity of addiction-related behaviors. Context-aware frameworks and multi-scale approaches combining RL with biophysical and theoretical models are discussed as avenues to bridge neural, behavioral, and environmental data. These innovations aim to refine predictive capabilities, personalize treatment strategies, and identify novel intervention points. By synthesizing diverse modeling approaches, this chapter underscores the transformative potential of RL in decoding the intricate mechanisms underlying SUDs and advancing addiction research.