Neural Processing of Feedback Signals During Continuous Feedback Learning
Haesoo Park, Denis G. Sukhodolsky, Christopher Pittenger, Michelle Hampson · 2024
Reinforcement-driven learning involves processing feedback signals to optimize future actions and decision making. This feedback can be discrete, occurring as a separate event from action or anticipation, or continuous, occurring constantly and concurrently with ongoing decision-making. Despite the real-world relevance of continuous feedback, most research paradigms investigate feedback processing as a discrete process. To address this, we analyzed fMRI data from four neurofeedback studies in which participants engaged in continuous feedback learning. Neural responses to the most positive, most negative, and the weakest feedback signals were compared. Our results show that nucleus accumbens activity tracked with feedback valence, and specifically, was greater in response to success than to failure during neurofeedback. Medial prefrontal cortex (mPFC), posterior cingulate cortex (PCC), hippocampus, supplementary motor area (SMA) and superior temporal gyrus (STG) were more activated in response to weak feedback than to strong feedback. Our findings suggest that during continuous feedback learning, weak feedback (more than strong feedback) increases activity in brain regions previously implicated in self referential thinking, decision-making under uncertainty, and referencing past experiences. These data suggest the ambiguity inherent in weak feedback may result in engagement of these processes to support context-dependent interpretations of the feedback signals.