Metacognition as a bridge between social learning and exploration-exploitation

根据试验的结果，又进行了一次新的试验。我们发现，人类参与观察学习是为了避免选择带来的潜在负面回报，并在较小程度上受益于更有经验的同伴。尽管可能的负奖励和高度观察的情况会导致更好的选择（在双臂强盗任务中有更多的最佳选择），但人们对知道最佳选择是什么缺乏信心。两个后续实验表明，这种选择和信心之间的分离是由奖励偏见和自我中心偏见介导的，奖励越低，信心就越低，自我中心偏见则会增加信心。初步结果已在美国神经经济学学会发表。

Following the results of the pilot, a new experiment was conducted. We found that humans engage in observational learning to avoid potential negative reward from choices, and to a lesser extent to benefit from more experienced peers. Although situations of likely negative reward and high observation led to better choices (more choices of the best alternative in a two-armed bandit task), people were less confident in knowing what the best option was. Two follow-up experiments revealed that such dissociation between choices and confidence was mediated by a reward bias, whereby confidence is reduced when reward is lower, and an egocentric bias, whereby making more own choices increases confidence. Preliminary results were presented at the Society for Neuroeconomics in the US.