Dr. Yonatan Loewenstein
Hebrew University of Jerusalem
Reinforcement learning: from the laboratory to the field
Humans and animals change their behavior in response to the consequences of their past actions. This type of learning is extremely powerful and enables the learning of complex behaviors. We used reinforcement learning algorithms to characterize this learning in controlled laboratory settings. In particular, we showed how risk aversion emerges from this learning. To extend our results to natural conditions, we characterized learning behavior of professional basketball players and showed that the reinforcement learning algorithms utilized by these players are not optimal, and are detrimental to their performance.