Model-free decision making is prioritized when learning to avoid harming others.

Area of Science:

• Neuroscience
• Cognitive Science
• Moral Psychology

Background:

• Moral behavior necessitates learning the consequences of actions on others.
• Theoretical learning frameworks distinguish between model-free (action-value caching) and model-based (action-outcome mapping) algorithms.

Purpose of the Study:

• To investigate the engagement and neural underpinnings of model-free versus model-based learning mechanisms in moral decision-making.
• To examine how learning to avoid personal harm differs from learning to avoid harming others.

Main Methods:

• Participants learned to avoid painful electric shocks for themselves and a stranger.
• Behavioral choices and neural activity (fMRI) were recorded during the learning and decision phases.
• Individual differences in moral judgment were assessed.

Main Results:

• Model-free decision-making was prioritized when learning to avoid harming others compared to oneself.
• Neural correlates of model-free prediction errors for others relative to self were identified in the thalamus/caudate.
• Subgenual anterior cingulate cortex (sgACC) showed activity related to model-free moral learning during choice, with enhanced sgACC-dlPFC connectivity after harming others.

Conclusions:

• Moral learning appears to favor efficiency (model-free) over flexibility (model-based) when avoiding harm to others.
• Specific neural circuits, including the thalamus/caudate and sgACC, support model-free moral learning.
• Individual moral judgment styles influence the model-free moral learning process.