Prosocial learning: Model-based or model-free?

Area of Science:

• Cognitive Neuroscience
• Behavioral Economics
• Social Psychology

Background:

• Prosocial learning is crucial for societal cooperation, involving acquiring skills for decisions benefiting others.
• Understanding how learning strategies differ for self versus other is key to prosocial behavior.

Purpose of the Study:

• To investigate differences in learning strategies during value-based decision-making for oneself compared to another individual.
• To examine the influence of self-other considerations on model-based and model-free reinforcement learning.

Main Methods:

• A two-step reinforcement learning paradigm was employed.
• Participants completed separate learning blocks for self-decisions and decisions for a confederate.
• Behavioral data were analyzed using computational modeling to distinguish model-based (MB) and model-free (MF) learning.

Main Results:

• Reinforcement learning exhibited canonical model-based and model-free features.
• Participant behavior was best explained by a mixture of MB and MF control, with a stronger reliance on MB control enhancing learning success.
• No significant differences in behavioral performance or model-based learning parameters were observed between self and other conditions.

Conclusions:

• Learning strategies for value-based decisions do not significantly differ when individuals act for themselves versus for others.
• The findings suggest that prosocial decision-making may rely on similar underlying learning mechanisms as self-interested decision-making.