Séminaire ISCMJ par Lei Zhang
Mercredi, 16 Mai, 2018 - 11:30

Jean-Claude Dreher a invité Lei Zhang du centre médical de l'Université d'Hambourg pour donner une conférence sur:

Les multiples facettes de l'influence social dans l'apprentissage ciblé

One of the main challenges in decision neuroscience originates from the fact that humans do not make decisions alone, but rather, are influenced by their social environment. In social situations in addition to making choices according to the action-outcome associations, humans tend to align themselves with others, even without any direct social interaction. Despite the rich literature on social influence, few studies have inspected the underlying neuro-cognitive and computational processes. In a fMRI study, I examine how social influence affects individuals’ choice and confidence in goal-directed learning, first by proposing a novel social influence paradigm that allows multiple players to interact with each other while engaging in a probabilistic reversal learning task. Using behavioral measurement and computational modeling, I demonstrate that social influence alters both choice and confidence and enhances individuals’ performance in goal- directed learning. Furthermore, an integrated value signal updated through both direct learning and observational learning is employed to guide future decisions. A model-based functional neuroimaging approach is applied to further identify that these dissociable value signals are encoded in the ventromedial prefrontal cortex and the anterior cingulate cortex gyrus, respectively. These analyses also show brain regions that represent dissenting social information and behavioral adjustment. Moreover, this work establishes the pattern of functional connectivity between the brain’s reward circuits and the social circuits. In a follow-up behavioral study, I assess how the expectation of the other agents’ expertise biases the social influence effect in goal-directed learning when playing with (un)intelligent computer agents in a modified social influence task. I show that such expectation only shifts choice preference but not confidence, and that social information is not integrated to facilitate learning. Together, this work provides a comprehensive behavioral and neurocomputational mechanisms of social influence in goal-directed learning. More broadly, these results shed light on how neurocomputational approach could be translated into psychiatry to accelerate precision medicine and personalized mental health.

Salle de conférence de l'ISC, le 16 mai à partir de 11h30