Altered Statistical Learning and Decision-Making in Methamphetamine Dependence: Evidence from a Two-Armed Bandit Task

Area of Science:

• Cognitive Neuroscience
• Computational Psychiatry
• Neuroeconomics

Background:

• Understanding goal-directed behavior is crucial for addiction research, as substance users struggle to balance immediate gratification with long-term objectives like sobriety.
• Abnormalities in learning and decision-making are hallmarks of substance use disorders, impacting an individual's ability to pursue long-term goals.

Purpose of the Study:

• To computationally model and identify specific learning and decision-making deficits in methamphetamine-dependent individuals (MDI) compared to healthy control subjects (HCS).
• To investigate the neural correlates associated with these cognitive abnormalities.

Main Methods:

• A two-armed bandit task was employed to assess exploration vs. exploitation strategies in learning reward rates.
• A computational framework, including the Dynamic Belief Model for learning and five decision policies (WSLS, ε-Greedy, τ-Switch, Softmax, Knowledge Gradient), was used to analyze subject behavior.
• Voxel-based morphometry was used to assess gray matter volume differences.

Main Results:

• Healthy control subjects (HCS) demonstrated adaptive learning, prioritizing recent data, and were best modeled by the Softmax decision policy.
• Methamphetamine-dependent individuals (MDI) exhibited simpler, learning-independent decision-making (WSLS) and, when modeled by Softmax, showed more pessimistic reward beliefs and reduced selection of optimal choices.
• MDI's avoidance of the most rewarding option correlated with reduced gray matter volume in the thalamic dorsal lateral nucleus.

Conclusions:

• Significant differences exist in learning and decision-making processes between MDI and HCS, highlighting cognitive impairments in substance use.
• The computational framework effectively reveals subtle abnormalities in learning and decision-making relevant to substance dependence.
• Thalamic gray matter volume may be a neural substrate underlying altered decision-making in MDI.