Nathan Thierry (Grenoble INP) and Geronimo Mirano (MIT): “‘New Insights into the Mechanisms of Human Observational Learning’”
The observation of others’ choices is an important means by which we can learn about the world and communicate with others. We propose a new uncertainty-based observational learning model in which individuals use not only information about observed choices themselves, but also the time taken to make them, allowing them to make inferences about others uncertainty. We show that this model efficiently learns observed values and describes the behaviour of subjects in a novel observational learning task with a computer agent whose reaction times are manipulated. The model predicts that specific quantities, namely value differences, individual uncertainty, and observed uncertainty, should each be co-represented in the brain. We tested this in a human versus human fMRI hyper-scanning task, and show that they have a convergent representation in lateral orbitofrontal cortex, suggesting this region is central to the social learning network. The results suggest an integrated computational and neurobiological account of how convergent, stable values arise in human networks.