作为一种运用人工智能让计算机“教会”自己不断进阶并寻找新数据以做出判断的“聪明技术”，算法推荐技术内蕴着摆脱人类控制的倾向和“从人指向物”的特定关系模式，具有高度伦理敏感性，并带来认知性与规范性两个方面的伦理风险。应对技术伦理挑战，展开算法推荐的伦理治理有三方面要求：一是以伦理先行的自觉，明确算法推荐的伦理规制原则，发挥伦理原则的路标作用；二是确立分布式道德责任，重塑算法推荐的责任主体，促进算法的负责任创新；三是优化算法推荐的伦理治理结构，展开敏捷治理，实现技术向善与技术赋能的契合。

算法推荐, 算法伦理, 伦理治理, 负责任创新

As a “smart technology” that uses artificial intelligence to let computers “teach” themselves to constantly advance and look for new data to make judgments, algorithm recommendation technology contains a tendency to get rid of human control and a specific relationship model of “from people to things”, which is highly ethical sensitive and brings ethical risks in both cognitive and normative aspects. In response to ethical challenges, there are three requirements for the ethical governance of algorithm recommendation. First, the awareness of ethics first, clear the ethical regulatory principles of algorithm recommendation, and play the role of ethical principles as a roadmap. Second, to establish the distributed moral responsibility, reshape the responsibility subject of algorithm recommendation, and promote the responsible innovation of algorithm. Third, to optimize the ethical governance structure recommended by the algorithm, carry out agile governance, and achieve the combination of kindness and empowerment.