Reward in statistics-based human decisions
Humans and other animals need to learn to compress their sensory input efficiently - guided by both the statistical regularities in, and the reward available from, the environment. Recent psychological and neuroscience studies show that reward signals duly modulate the representations learned. However, these representations determine how we can recognise objects and concepts, and reward's indirect influence over these remains obscure. Hypothesising that humans use limited computational resources to steer learned representations towards their utility for gaining reward, RELEARN proposes a mathematical model that integrates reinforcement learning and probabilistic representational learning. We will test behavioural and neural predictions in experiments where humans learn to gain reward based on the feature statistics of a simulated environment.
Fields of science
Call for proposalSee other projects for this call
Funding SchemeMSCA-IF-EF-ST - Standard EF
See on map