Learning lists and gestural signs: dyadic brain models of non-human primates

Neuroscientific computational modeling has been successful in characterizing neural circuits for vision, motor control and decision making, but have not often been applied to behaviors exemplary of the primates, namely social learning, social cognition and communication. There are two main efforts we describe here: the modeling, via computer simulation, of (1) brain mechanisms of serial learning in monkeys and their transfer through observation of others’ performances, and (2) brain mechanisms of social learning and interaction that support the development of gestural repertoires in apes. The first main effort is focused on analyzing previously published behavioral and neurophysiological data in monkeys to construct novel computational models to explain these data, challenge existing interpretations, and generate testable hypotheses. We show how monkeys may learn sequences of items through trial-and-error mechanisms that manage multiple, concurrent learning processes, including of temporal order and reward-predictive value, and can yield behavioral patterns that qualitatively match those in the literature, while simulated neural responses predict response profiles for a variety of cortical and sub-cortical areas in macaques. Further data from the literature suggest how learning a list may be aided by observing another monkey generate that list. A crucial innovation in our work involves going beyond simulating the mechanisms coordinating an individual’s actions or decisions, by having multiple simulated agents able to learn from others’ behavior. We show how we can situate action-recognition and feedback processing elements to process others' performances and yield facilitated performance when simulated later in isolation. ❧ In our second main effort, we go further by simulating: (i) ape brains, (ii) interaction and gestural communication, and (iii) dynamic exchange of information, back-and-forth and not just one-way, between modeled apes. This ‘dyadic brain modeling’ can show how the ‘mutual shaping of behavior’ between apes may lead to novel gestural forms that serve as communicative signals between individuals, as hypothesized in the literature. Further, we show how alternative hypotheses of gestural acquisition—the pruning on an innate gestural repertoire—can be handled by our integrative model, and how we can explain multiple learning pathways leading to varied gestural repertoires in a single unified and computationally-specific model. ❧ These efforts help to clarify brain mechanisms managing social learning in primates—mechanisms likely highly conserved in humans—and mechanisms managing social cognition and interaction. Additionally, we show how competing hypotheses of gestural learning and usage in apes can be explained by one model, and offer hypotheses towards which varying usage patterns may be understood, with additional suggestions for brain mechanisms important in the evolution in apes and humans that yield the flexibility in gestural communication observed in the wild. Finally, the methodological innovation of dyadic brain modeling—simulating brain models in interaction—attempts to move the field of computational neuroscience towards modeling more complex behaviors.