Neural Information Processing Systems http://nips.cc/

How humans make repeated choices among options with imperfectly known reward outcomes is an important problem in psychology and neuroscience. This is often studied using multi-armed bandits, which is also frequently studied in machine learning. We present data from a human stationary bandit experiment, in which we vary the average abundance and variability of reward availability (mean and variance of reward rate distributions). Surprisingly, we find subjects significantly underestimate prior mean of reward rates -- based on their self-report, at the end of a game, on their reward expectation of non-chosen arms. Previously, human learning in the bandit task was found to be well captured by a Bayesian ideal learning model, the Dynamic Belief Model (DBM), albeit under an incorrect generative assumption of the temporal structure - humans assume reward rates can change over time even though they are actually fixed. We find that the pessimism bias in the bandit task is well captured by the prior mean of DBM when fitted to human choices; but it is poorly captured by the prior mean of the Fixed Belief Model (FBM), an alternative Bayesian model that (correctly) assumes reward rates to be constants. This pessimism bias is also incompletely captured by a simple reinforcement learning model (RL) commonly used in neuroscience and psychology, in terms of fitted initial Q-values. While it seems sub-optimal, and thus mysterious, that humans have an underestimated prior reward expectation, our simulations show that an underestimated prior mean helps to maximize long-term gain, if the observer assumes volatility when reward rates are stable and utilizes a softmax decision policy instead of the optimal one (obtainable by dynamic programming). This raises the intriguing possibility that the brain underestimates reward rates to compensate for the incorrect non-stationarity assumption in the generative model and a simplified decision policy.

Neural Information Processing Systems http://nips.cc/

This paper presents an intriguing computational dissection of a particular form of reward rate underestimation in a bandit task (what the authors call as "pessimism bias"). Modeling suggests that this bias can be accounted for by a Bayesian model which assumes (erroneously) that reward rates are dynamic. The paper is well-written and the methods are sound. I think it could do a better job relating to previous literature, and there are some questions about the modeling and behavioral analysis which I detail below. Specific comments: I was surprised that there was no mention of Gershman & Niv (2015, Topics in Cognitive Science), which is one of the only papers I'm aware of that manipulates reward abundance.

How humans make repeated choices among options with imperfectly known reward outcomes is an important problem in psychology and neuroscience. This is often studied using multi-armed bandits, which is also frequently studied in machine learning. We present data from a human stationary bandit experiment, in which we vary the average abundance and variability of reward availability (mean and variance of reward rate distributions). Surprisingly, we find subjects significantly underestimate prior mean of reward rates -- based on their self-report, at the end of a game, on their reward expectation of non-chosen arms. Previously, human learning in the bandit task was found to be well captured by a Bayesian ideal learning model, the Dynamic Belief Model (DBM), albeit under an incorrect generative assumption of the temporal structure - humans assume reward rates can change over time even though they are actually fixed.