When one brain needs to learn from another: the case of observational facilitation of list learning in macaques

Abstract

Given a neural system that equips an agent to attempt to carry out some learning task based on its own interaction with the “non-social” world, what extra neural machinery is required to enable learning to be facilitated by (repeated) observation of successful completion of that task by another agent? We provide one answer by exploiting an understanding of data and models on mirror neurons to extend a prior neurocomputational model of list learning by macaques, SCP1, which addressed results of the simultaneous chaining paradigm (SCP) to yield a new model, SCP2, that addresses social facilitation (observational learning) effects based on the SCP. SCP2 extends SCP1 by adding action-recognition elements and (vicarious) reward-processing elements to facilitate performance following observation of a demonstrator. Our simulations suggest prior experience is important for the observed facilitation and serves to bridge the (separately collected) neurophysiological and behavioral data. Crucially, the inner workings of SCP1, as distinct from its successful performance on the SCP dataset, are irrelevant. What is crucial is the “wrapping” of the “do it alone” model to support social facilitation. This study provides an example of dyadic brain modeling, simulating brain models of interacting agents, in the case in which the behavior of only one member of the dyad is affected by the behavior of the other.