嘉宾简介：郑捷，清华大学经济学学士、硕士，美国华盛顿大学经济学硕士、博士，现为清华大学经济管理学院经济系副教授，博士生导师。国际学术期刊Journal of Economic Behavior and Organization副编，Research in Economics副编，曾担任多份SSCI期刊客座主编。研究领域包括信息经济学、实验经济学、行为经济学、产业经济学，研究主题涵盖机制设计、市场设计、信息设计等经典问题和参照依赖、自我控制、互惠合作等行为问题。主持过多项国家自然科学基金项目（结题获“特优”评价），并多次在学术会议上做主旨演讲与专题报告。研究工作发表于《经济研究》、American Economic Review (Papers and Proceedings)、Games and Economic Behavior、Management Science、Nature Communications等国内外经济学、管理学、自然科学各领域的知名期刊。
Abstract: When does the “wisdom of crowds” help cultivate factual information and when is it harmful? In a laboratory experiment, we ask subjects to answer 50 True/ False questions, with the opportunity to revise their answers after receiving different levels of information about other subjects’ answers, incentivized confidence levels and higher order beliefs. We verify that in the True/False setting, the “wisdom of crowds” is most beneficial for easy questions and for less knowledgeable subjects, while performance on difficult questions and by more knowledgeable subjects is harmed by it. Based on treatments of varying information provision levels, we find that subjects in the Moderate-Information treatment outperform those in either the Low-Information or Full-Information treatments, indicating an intermediate optimal level of social information for decision-making. We then examine three commonly discussed decision-heuristics in the context of our True/False task: the simple Majority Rule (majority of others’ choices), the Maximum Confidence rule (others’ incentivized self-confidence levels), and the Surprising Popularity rule (relative popularity compared to the expected popularity). Although in the context of our True/False task, the Maximum Confidence rule yields the best performance, subjects tend to neglect the information about other decision-makers’ confidence levels, instead opting to rely heavily on Majority Rule. Our findings also indicate that decision-makers more generally have difficulty making use of higher-order information about others’ choices and beliefs. Our study documents the tendency to over-rely on majority opinion and suggests a possibility that confidence-weighted rating systems can potentially offer improvements when trying to cultivate factual knowledge.