English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Humans and animals frequently need to make choices among options with imperfectly known reward outcomes. In neuroscience, this is often studied using the multiarmed bandit task, in which subjects repeatedly choose among bandit arms with fixed but unknown reward rates, thus negotiating a tension between exploitation and exploration. Here, using a modified version of the bandit task in which we query subjects reward expectations of unchosen arms, we investigate how general reward availability in the environment affects human prior beliefs. Based on self-report data and computational modeling of behavioral data, we obtain converging evidence that human subjects systematically under-estimate reward availability. Additional computational analyses reveal that this under-estimation compensates for two other apparent suboptimalities in human behavior, namely a default assumption of environmental non-stationarity, and the use of a simplistic decision policy. This result represents a concrete instance in which multiple sub-optimalities in brain computations synergistically interact to achieve much better-than-expected behavioral outcome. This work raises the intriguing possibility that many apparently isolated limitations in brain computation and representation may actually work together to achieve highly intelligent behavior in a broader context, and also sheds light on computationally efficient algorithms that could be adopted by artificial intelligence systems.

Human learning and decision-making in the bandit task: Three wrongs make a right