Conference on Cognitive Computational Neuroscience

People often have to switch back and forth between different environments that come with different statistics. While some environments require fast learning (i.e., high learning rates), other environments call for lower learning rates. Previous studies have shown that people adapt their learning rates to their environment when differences in reward volatilities are clustered in time (Behrens et al., 2007). However, differences in learning rates observed in these studies could reflect emergent properties of participants’ responses to locally experienced prediction errors. As such, it remains unclear whether people can actually learn about environment-specific learning rates, and instantaneously retrieve them when revisiting environments (i.e., meta-learn). Here, using Bayesian hierarchical modelling of participants’ behavior in a continuous estimation task while alternating between multiple environments, we show that people can meta-learn. We conclude that humans can learn to associate different learning rates to different environments, offering important insights for developing theories of meta-learning (Silvetti et al., 2018) and context-specific control (Abrahamse et al., 2016).