论算法推荐的版权中立性  

On the Copyright Neutrality of Algorithmic Recommendation

作　　者：崔国斌[1] 

机构地区：[1]清华大学法学院

出　　处：《当代法学》2024年第3期55-69,共15页Contemporary Law Review

基　　金：国家社科基金重大项目“互联网交易制度与民事权利保护研究”(20&ZD192)的阶段性研究成果。

摘　　要：利用程序算法自动向用户推荐作品内容,已经成为网络服务商的普遍做法。这大大提升了用户扩散著作权侵权内容的能力,也扩大了服务商的收益。著作权法是否应当因此而强化算法推荐服务商的注意义务,存在很大争论。在决策者修正现有法律之前,在帮助侵权框架下分析算法推荐行为的属性,依然是合理的选择。在算法推荐前,虽然普遍存在内容安全审查,但是,其中的人工审查通常并不能导致服务商获知多数用户上传内容的版权属性。算法推荐技术本身也没有帮助服务商更有效地识别或预防用户侵权行为。如果单纯因为服务商采用算法推荐技术而要求服务商在预防用户侵权方面承担更高的注意义务,则会损害社会的整体福利。当然,内容过滤等网络技术的进步和网络商业模式的演进可能导致服务商承担更多的注意义务,但算法推荐技术本身并非推动这一制度改革的原动力。算法推荐过程的确存在一定的黑箱风险,强制披露难以有效解决这一问题,通知-删除(过滤)措施依然是更为有效的应对措施。Using algorithmic systems to automatically recommend content to users has become a common practice among internet service providers.It greatly enhances the ability of users to spread copyright-infringing content,while also expanding the revenue of service providers.There is considerable debate over whether copyright law should therefore strengthen the duty of care of algorithmic recommendation service providers.When the existing copyright liability rules for Internet service providers were originally designed,decision-makers had maintained a very delicate balance of interests among copyright owners,users and service providers.Before decision-makers amend existing laws,it is still a reasonable choice to analyze algorithmic recommendation under the legal framework of contributory infringement.In cases content storage and publishing service providers do not actually control their users'behavior or directly obtain substantial benefit from users'disseminating of copyright-infringing content,they are viewed"conduits"rather than"publishers"(content providers),and only bear contributory liability for their users'infringing activities.Algorithmic recommendation technology itself does not help service providers to effectively identify or prevent their users'infringing behavior.Although"content security"censorship is prevalent before algorithmic recommendation,it usually does not result in service providers being informed of the copyright-infringing nature of the user-generated content.Therefore,most algorithmic recommendation systems are"copyright-neutral,"and service providers generally should not bear additional liability for expanding users infringing activities through algorithmic recommendations.If service providers are required to assume more duty of care in preventing users'infringement simply because of their use of algorithmic recommendation technology,they would be forced to conduct more manual reviews,which would not only substantially increase their operating costs,but also restrict the users'freedom in transmi

关 键 词：算法推荐 著作权 帮助侵权 算法黑箱 安全港 

分 类 号：D923.41[政治法律—民商法学]