We propose using Bayesian Persuasion as a tool for social media platforms to combat the spread of online misinformation. As platforms can predict the popularity and misinformation features of to-be-shared posts, and users are motivated to only share popular content, platforms can strategically reveal this informational advantage to persuade users to not share misinformed content. Our work mathematically characterizes the optimal information design scheme and the resulting utility when observations are not perfectly observed but arise from an imperfect classifier. Framing the optimization problem as a linear program, we give sufficient and necessary conditions on the classifier accuracy to ensure platform utility under optimal signaling is monotonically increasing and continuous. We next consider this interaction under a performative model, wherein platform intervention through signaling affects the content distribution in the future. We fully characterize the convergence and stability of optimal signaling under this performative process. Lastly, the broader scope of using information design to combat misinformation is discussed throughout.

翻译：我们提出利用贝叶斯说服作为社交媒体平台对抗在线虚假信息传播的工具。由于平台能够预测待分享帖子的流行度和虚假信息特征，而用户受动机驱使仅分享热门内容，平台可战略性地揭示这一信息优势，以说服用户不分享含有虚假信息的内容。我们的工作从数学上刻画了最优信息设计方案，以及当观察结果并非完美观测而是源自不完美分类器时所产生的效用。通过将优化问题构建为线性规划，我们给出了分类器准确性的充分必要条件，以确保在最优信号策略下平台效用单调递增且连续。随后，我们考虑了在表现性模型下的这一交互过程，其中平台通过信号干预影响未来的内容分布。我们全面刻画了此表现性过程中最优信号策略的收敛性与稳定性。最后，本文始终探讨了利用信息设计对抗虚假信息的更广泛前景。