Decision Threshold Learning in the Basal Ganglia for Multiple Alternatives.

Area of Science:

• Computational Neuroscience
• Cognitive Science
• Decision Science

Background:

• Reinforcement learning (RL) and evidence accumulation models like the diffusion decision model (DDM) are historically separate but increasingly integrated.
• The RL-DDM combines value learning with DDM but struggles with multi-option decisions and environmental flexibility.
• Current models lack clear mapping to neurophysical processes, particularly in basal ganglia go/no-go decision-making.

Purpose of the Study:

• To propose a novel decision model addressing limitations of existing RL-DDM approaches.
• To extend decision modeling to handle multiple choices and context-dependent environmental changes.
• To link computational models of decision-making to neurobiological mechanisms in the basal ganglia.

Main Methods:

• Combined a multichoice sequential probability ratio test (MSPRT) decision model with a dual-pathway basal ganglia threshold learning model.
• Developed a model that learns decision thresholds to balance time cost and error cost.
• Incorporated context-dependency for flexible adaptation to speed-accuracy trade-off (SAT) changes.

Main Results:

• The proposed model effectively scales to decisions with multiple options.
• It demonstrates flexibility in adapting to rapid, context-cued changes in reward environments.
• The model reproduces the magnitude effect observed in value-based decisions and is applicable to various evidence types.

Conclusions:

• The novel model successfully integrates RL and DDM concepts for multi-option decisions.
• It provides a neurophysiologically plausible account of decision threshold learning in the basal ganglia.
• This work bridges separate research areas, linking RL-DDM to dopaminergic motivation and risk-taking models.