算法推荐中的隐私保护与“信息茧房”治理研究

Governing Algorithmic Recommendation: Privacy Protection and the “Information Cocoon” Effect

摘要: 以淘宝、美团、抖音等电子商务及互联网平台为代表的算法推荐系统在降低搜寻成本、提升交易效率的同时，诱发了个人信息过度收集、决策不透明与差别定价等风险，并通过“同质化推送 + 偏好强化”的正反馈机制造成了“信息茧房”。本文以民法路径为方法论主轴，在人格权保护框架下，重构“隐私—个人信息—信息自主”三层次权利；于《消费者权益保护法》之下，评估差别定价与算法歧视之可责性；以《个人信息保护法》与《互联网信息服务算法推荐管理规定》为治理中枢，检讨“一键关闭个性化”等制度之效力边界，并引入欧盟GDPR第22条与DSA中“非画像推荐选项”的比较法启示。同时，在民法上精细化平台的告知义务与安全保障义务，构建可操作的举证责任分配与损害推定规则，以实现创新与权利保障的均衡。

Abstract: Algorithmic recommendation system exemplified by platforms such as Taobao, Meituan, and Douyin have enhanced transactional efficiency in the digital economy but also raise pressing legal challenges, including excessive personal data collection, opaque decision-making, and discriminatory pricing. These systems further exacerbate the information cocoon effect through preference amplification and content homogenization. Anchored in a civil law framework, this paper examines the rights to privacy, personal information, and informational self-determination, explores platform accountability under the “Consumer Rights Protection Law”, and analyzes regulatory mechanisms under the “Personal Information Protection Law” and the “Administrative Provisions on Algorithmic Recommendation”. Drawing on Article 22 of the GDPR and the DSA’s provisions on non-profiling recommendations, the study reflects on adaptive rulemaking and platform obligations, aiming to reconcile technological efficiency with rights protection.