Econometrica · 2026 · Sergei Guriev、Emeric Henry、Théo Marquis、Ekaterina Zhuravskaya
中文摘要
我们构建了一个综合框架，用以评估旨在遏制社交媒体上虚假新闻传播的政策干预。利用2022年和2024年美国选举期间在Twitter和X上开展的一项随机实验，我们评估了提高对错误信息警觉性的启动干预、事实核查、确认点击和内容考量提示。结果表明，启动干预在减少虚假新闻分享方面最为有效，同时能够维持真实新闻的传播。我们构建了一个以党派说服、党派信号传递和声誉关切为分享动机的模型，并对其进行结构估计。我们识别出政策影响分享行为的三条渠道：（i）更新对内容真实性和党派属性的认知；（ii）提高声誉考量的凸显程度；（iii）增加参与成本。政策效果的差异可由凸显渠道和成本渠道解释。内容中立的启动干预能够以很低的成本，最有效地提高声誉考量的凸显程度，而且在更新真实性认知方面几乎与事实核查同样有效。当用户遇到真实性无争议的内容时，凸显渠道的作用更强。
Abstract
We develop a comprehensive framework to evaluate policy interventions aimed at curbing false news dissemination on social media. Using a randomized experiment on Twitter and X during the 2022 and 2024 U.S. elections, we assess priming for misinformation awareness, fact‐checking, confirmation clicks, and content consideration prompts. Priming proves most effective in reducing false‐news sharing while preserving true news dissemination. We build and structurally estimate a model of sharing, motivated by partisan persuasion, partisan signaling, and reputational concerns. We identify three channels through which policies influence sharing: (i) updating perceived veracity and partisanship of content, (ii) raising salience of reputation, and (iii) increasing engagement costs. Differences in the effects of policies are explained by the salience and cost channels. Content‐neutral priming is best at enhancing salience of reputation at minimal cost and almost as effective as fact‐checking in updating veracity. Salience channel is stronger when users encounter uncontroversially true content.