Reliability of Decision-Making and Reinforcement Learning Computational Parameters.

Area of Science:

• Cognitive Neuroscience
• Computational Psychiatry

Background:

• Computational models offer mechanistic insights into cognition, crucial for psychiatric disorder research.
• Reliable computational measures are essential for successful translational psychiatric research.

Purpose of the Study:

• To assess the reliability of reinforcement learning and economic models in capturing individual characteristics.
• To evaluate the translational potential of computational parameters for precision psychiatry.

Main Methods:

• Fifty healthy individuals completed a restless four-armed bandit and a calibrated gambling task twice, with a two-week interval.
• Reinforcement learning and prospect theory models were derived from task performance data.
• Reliability of model parameters (learning rates, sensitivity, risk/loss aversion) was assessed.

Main Results:

• Reinforcement learning parameters showed good to fair reliability.
• Prospect theory parameters (risk and loss aversion) exhibited good to excellent reliability.
• Both models predicted future individual behavior, with personalized parameters yielding better predictions.

Conclusions:

• Reinforcement learning and prospect theory parameters derived from these tasks are reliably measurable.
• These reliable computational parameters can assess learning and decision-making mechanisms.
• Findings support the translational potential of computational parameters for precision psychiatry.