Jean-Claude Dreher has invited Caroline Charpentier(Caltech USA) to come and give a talk on:
From information seeking to observational learning: behavioral and neural computations
Gathering information and observing others are crucial steps to learn about the world around us, yet little is known about the computational mechanisms that allow us to do so. In this talk, I will present two projects that sought to tackle this question. In the first project, we provide empirical evidence that decisions to seek information are not only driven by outcome uncertainty, but are also strongly valence-dependent. When a future outcome is desirable people are more likely to want to know, but when a future outcome is undesirable, they often choose to remain ignorant. Strikingly, participants are also willing to pay for knowledge and ignorance as a function of valence. Computationally, mesolimbic reward regions were found encode valence-dependent information prediction errors, suggesting a mechanism for why knowledge is not always valued. In the second project, we examined two strategies that people can deploy when learning from observing others: imitation (learning about and copying another agent’s actions) and emulation (learning by inferring the beliefs, intentions and goals of another agent). Depending on uncertainty and volatility in the environment, people adaptively arbitrate between the two strategies. Computational models suggest that arbitration is driven by the relative reliability of each strategy, with corresponding model-based regressors tracked in the brain’s valuation and mentalizing networks. Together, these findings provide novel insights on how people sample information and learn from others in an uncertain world. They pave the way for future investigations of the role of social and emotional factors in these processes.
ISC Council room, from 2pm