Neuro-computational mechanisms and individual biases in action-outcome learning under moral conflict

This study investigated how individuals learn to make choices in morally conflicting situations where actions benefit the self but harm others. Using a combination of behavioral data and fMRI, the researchers found that choices were best described by a reinforcement learning model that tracks the expected values of self-benefit and other-harm separately, rather than combining them. Individual differences in preferences were reflected in a valuation parameter, which also predicted choices in an independent costly helping task. fMRI revealed that the ventromedial prefrontal cortex (vmPFC) reflected the bias in expected values toward the favored outcome, while the pain-observation network represented pain prediction errors independently of individual preferences.

Laura Fornari, Kalliopi Loumpa, Alessandra D. Nostro, Nathan J. Evans, Lorenzo De Angelis, Sebastian P. H. Speer, Riccardo Paracampo, Selene Gallo, Michael Spezio, Christian Keysers, Valeria Gazzola