Sensorimotor learning biases choice behavior: a learning neural field model for decision making.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• Sensorimotor processing involves parallel selection and specification of actions, emerging from competition between movement plans.
• Frontoparietal sensorimotor areas are crucial for action selection with ambiguous sensory input, encoding multiple motor goals and exhibiting competitive selection.
• Existing models lack the ability to incorporate sensorimotor learning and changing reward contingencies into decision-making processes.

Purpose of the Study:

• To present a dynamic neural field model capable of learning arbitrary sensorimotor associations.
• To investigate how network plasticity and learning history influence action selection and decision-making.
• To provide an integrated account of sensorimotor integration, working memory, and action selection in ambiguous choice situations.

Main Methods:

• Development of a dynamic neural field model with a reward-driven Hebbian learning algorithm.
• Simulation of action selection dynamics under varying reward contingencies.
• Validation against monkey cortical recordings and prediction of choice errors in a control experiment.

Main Results:

• The model accurately simulates action selection dynamics observed in primate studies.
• The model successfully predicts choice errors, highlighting the role of learning history.
• Demonstration that network plasticity significantly influences choice behavior and adaptation to new reward contingencies.

Conclusions:

• Dynamic neural field models can integrate sensorimotor processing, learning, and decision-making.
• Adaptive models are essential for understanding how learning history and plasticity shape action selection.
• This work offers a unified framework for sensorimotor integration and decision-making in complex environments.