应兰州大学应急管理研究中心、管理学院危机信息管理研究所邀请，浙江大学管理学院曹仔科副教授将于我院举办学术讲座。欢迎感兴趣的师生积极参加！

讲座题目：自主算法审核对用户内容贡献的影响：来自在线知识共享社区的自然实验

Effects of Autonomous Algorithmic Moderation on User Content Contribution: A Natural Experiment in an Online Knowledge-Sharing Community

主 讲 人：曹仔科 副教授

主 持 人：王洪鹏 副教授

时     间：2024年8月6日（星期二） 9:30-11:30

地     点：兰州大学城关校区（西区）齐云楼204教室

讲座简介：

在线数字平台越来越依赖算法进行平台治理。本讲座研究了一个日益增长的算法应用领域：自主算法内容审核。用户生成内容（UGC）平台通常会部署复杂的算法来筛选用户上传的内容，以过滤掉低质量内容。通过分析一个大型在线问答（Q&A）社区中发生的自然实验，发现引入自主算法内容审核系统后，尽管用户贡献答案的数量显著减少，而但贡献答案的平均质量有所提高，这体现在收到的点赞数、认知语言的使用和文本可读性上。研究结果表明，实施自主算法审核系统对平台治理而言是一把双刃剑：一方面，算法内容审核的威慑力抑制了用户生成新内容的动力（即威慑效应）；另一方面，用户会被激励投入更多精力来创作更高质量的内容（即替代效应）。还记录了用户反应的显著差异：在确实受到审核算法惩罚的用户中，只观察到了威慑效应，没有观察到替代效应，而在未受到审核算法惩罚的用户中，既观察到了威慑效应，也观察到了替代效应；此外，威慑效应并未显现在活跃度较低的用户群组内，在部署审核算法后，他们贡献内容的数量和质量均呈现出小幅增加。为进一步验证实证研究结果并为潜在机制提供直接证据，还进行了一项额外的在线实验。研究结果为使用算法进行自主平台治理的在线平台提供了重要启示。

Online digital platforms are increasingly relying on algorithms to perform platform governance. We study a growing application of algorithms in platform governance: autonomous algorithmic content moderation. Platforms thriving on user-generated content (UGC) commonly deploy sophisticated algorithms to screen content uploaded by users for the purpose of filtering out low-quality content. By exploiting a natural experiment occurred in a large online question-and-answer (Q&A) community, we find that the introduction of an autonomous algorithmic content moderation system significantly decreased the volumes of answers contributed by users, while the average quality of the contributed answers improved after the algorithm’s introduction, as reflected in the number of vote-ups received, usage of cognitive language and text readability. Our findings demonstrate that the implementation of autonomous algorithmic moderation system could be a double-edged sword for platform governance: on the one hand, the deterrence power of the algorithmic content moderation inhibits users’ incentives to generate new content (i.e., a deterrence effect); on the other hand, users are motivated to spend more efforts curating higher-quality content (i.e., a displacement effect). We also document significant heterogeneity in users’ responses: among users who were indeed punished by the moderation algorithm only the deterrence effect is observed but not the displacement effect, while among users who were not punished by the moderation algorithm both the deterrence and displacement effects are observed; moreover, less active users were not subject to the deterrence effect and both the volumes and quality of their content contributions showed a small but significant increase after the deployment of the moderation algorithm. An additional online experiment is conducted to corroborate the empirical findings and provide direct evidence of the underlying mechanisms. Our findings provide important implications for online platforms that use, or consider using, algorithms to perform autonomous platform governance.

主讲人简介：

曹仔科，浙江大学管理学院长聘副教授，博士生导师，获香港科技大学资讯系统 (Information Systems) 博士学位，浙江大学信息管理与信息系统学士学位。研究方向主要为在线用户行为决策分析和在线社区平台治理策略。多篇论文发表于Information Systems Research (UTD24期刊), MIS Quarterly (UTD24期刊)，Journal of Management Information Systems (FT50期刊) 等顶级期刊。博士论文获得2017年度INFORMS信息系统协会Nunamaker-Chen博士论文奖并列第二名 (second runner-up)。研究论文获得2021年度INFORMS电子商务领域 (eBusiness Section) 最佳论文奖第一名 (winner)。主持国家自然科学基金青年项目和优秀青年项目。