文/张传文 超级平台用户规模扩大，商业属性不断强化，同时作为社会基础设施的角色和功能更加明显，算法治理 正在面临新的挑战。有一些专家甚至断言，未来平台将在其内部发挥着"准立法权"、"准司法权"和"准 行政权"，政府、研究机构或媒体难以全面掌握平台企业的用户状况和实际运行情况，其相关行动需要 平台企业给予积极配合才可能完成，实现"穿透式监管"或实质性监督方面存在现实困难。 信息茧房、观点极化……如何解决"困在算法里"，做到"算法向善"？ "清朗·网络平台算法典型问题治理"专项行动开展以来，针对网民反映强烈的算法推荐加热低俗信息、 加剧"信息茧房"、加重观点极化等问题风险，中央网信办督促指导重点平台对信息推荐算法功能进行持 续性和针对性优化，取得阶段性成果，如公开算法规则原理，开发"茧房评估""一键破茧"等功能。但算 法治理仍是一项长期性、系统性工程，需常态化巡查监督和持续优化，中央网信办正在努力探索多样 化"破茧"路径，做到切实维护网民合法权益。 技术在创新，监管创新步伐亦未停止。境内互联网信息服务算法备案信息公开发布，是监管部门在算法 治理领域迈出的重要一步，目前国家互联网信息办公室的公布数量已达数百条。平台 ...

Core Viewpoint - The article discusses the efforts of major online platforms in China to optimize their algorithm recommendation systems in response to government initiatives aimed at addressing issues such as the promotion of vulgar content, the exacerbation of "information cocoons," and the polarization of viewpoints [1][2]. Algorithm Transparency - Major platforms have taken steps to enhance algorithm transparency by publicly disclosing their operational rules and promoting user awareness [3]. - Douyin has established a "Safety and Trust Center" to explain its recommendation logic and governance outcomes [4]. - Weibo has improved the transparency of its trending search algorithms by publicizing ranking rules and data metrics [4]. - WeChat Video Account has utilized simple graphics and videos to inform users about its algorithm recommendations [4]. Breaking "Information Cocoons" - Platforms have introduced innovative features like "Cocoon Assessment" and "One-Click Break Cocoon" to help users mitigate the risks associated with "information cocoons" [5][6]. - Douyin has upgraded its "Usage Management Assistant" to visually present users' recent browsing content [6]. - Xiaohongshu has implemented features for content preference assessment and exploration of diverse recommendations [6]. - Kuaishou has enhanced its algorithm to promote positive and useful content [6]. Content Review Mechanism - Platforms are continuously improving their content review mechanisms to strengthen the promotion of positive content and prevent the recommendation of vulgar or harmful information [7][8]. - WeChat Video Account has refined its dual mechanism of "Friend Recommendations" and "Algorithm Recommendations" to better identify and eliminate inappropriate content [8]. - Douyin has introduced a verification mechanism for trending topics to prevent the spread of misleading or fabricated content [8]. User Empowerment - Platforms are optimizing features related to user interest management and content feedback to allow users to adjust their algorithmic recommendations [9]. - Kuaishou offers detailed interest preference management tools for users to customize content delivery [10]. - Weibo has provided various negative feedback options for users to express disinterest in certain content or creators [10]. - The government acknowledges that while progress has been made in algorithm governance, there are still gaps in functionality and content quality that need to be addressed [10].

资料图。本文来源： 中央网信办持续加强信息推荐算法治理 是一项长期性、系统性工程，网信部门将开展常态化巡查，督导信息推荐算法平台持续优化算法运 行机制和管理规则，不断提升推荐内容质量，探索多样化"破茧"路径，创新用户赋权功能，切实维 护网民合法权益。 ■ "清朗·网络平台算法典型问题治理"专项行动开展以来，针对网民反映强烈的算法推荐加热低俗信 息、加剧"信息茧房"、加重观点极化等问题风险，中央网信办督促指导重点平台针对性优化信息推 荐算法功能、调整信息推荐算法规则。重点平台积极响应，签署"算法向善"南宁宣言，完善算法推 荐内容审核，开设专门网站、频道或账号集中公开算法规则原理，开发上线"茧房评估""一键破 茧"等创新功能，完善用户兴趣偏好管理服务，提升算法推荐内容多样性。 近日，抖音、小红书、微博、快手、微信视频号、哔哩哔哩等平台围绕正能量内容加权推荐、用户 自主选择权保障、推荐内容多样性优化、提升算法透明度等核心环节，系统性优化完善多项功能。 在提升算法透明度方面。 公开算法运行规则，推进透明展示工作，保障用户知情权。抖音通过开 设"安全与信任中心"网站、举办开放日活动等方式，向公众阐释推荐逻辑、干预机制及 ...

中央网信办持续加强信息推荐算法治理。 "清朗·网络平台算法典型问题治理"专项行动开展以来，针对网民反映强烈的算法推荐加热低俗信息、加剧"信 息茧房"、加重观点极化等问题风险，中央网信办督促指导重点平台针对性优化信息推荐算法功能、调整信息 推荐算法规则。重点平台积极响应，签署"算法向善"南宁宣言，完善算法推荐内容审核，开设专门网站、频道 或账号集中公开算法规则原理，开发上线"茧房评估""一键破茧"等创新功能，完善用户兴趣偏好管理服务，提 升算法推荐内容多样性。 近日，抖音、小红书、微博、快手、微信视频号、哔哩哔哩等平台围绕正能量内容加权推荐、用户自主选择权 保障、推荐内容多样性优化、提升算法透明度等核心环节，系统性优化完善多项功能。 在提升算法透明度方面。公开算法运行规则，推进透明展示工作，保障用户知情权。抖音通过开设"安全与信 任中心"网站、举办开放日活动等方式，向公众阐释推荐逻辑、干预机制及治理成效。微博提升热搜算法透明 度，公示上榜规则、数据规则，上线热搜热度标签，标示热度背后的推动因素。微信视频号通过通俗易懂的图 文、视频方式向用户公示，发布《一图读懂微信视频号算法推荐》《算法破茧系列》。 在破除"信息 ...

今天（5月2 2日）1 0点，微信公众号"网信中国"发文称，中央网信办持续加强信息推荐算法治 理。 全文如下 ↓ ↓ "清朗·网络平台算法典型问题治理"专项行动开展以来，针对网民反映强烈的算法推荐加热低 俗信息、加剧"信息茧房"、加重观点极化等问题风险，中央网信办督促指导重点平台针对性 优化信息推荐算法功能、调整信息推荐算法规则。 重点平台积极响应，签署"算法向善"南宁宣言，完善算法推荐内容审核，开设专门网站、频 道或账号集中公开算法规则原理，开发上线"茧房评估" "一键破茧"等创新功能，完善用户兴 趣偏好管理服务，提升算法推荐内容多样性。 近 日 ， 抖 音 、 小 红 书 、 微 博 、 快 手 、 微 信 视 频 号 、 哔 哩 哔 哩 等 平 台 围 绕 正 能 量 内 容 加 权 推 荐、用户自主选择权保障、推荐内容多样性优化、提升算法透明度等核心环节，系统性优化 完善多项功能。 来 源 | 微信公众号"网信中国" 在完善推荐内容审核方面。不断健全推荐内容审核机制，加强正能量内容推送，防范算法推 荐低俗不良信息。微信视频号完善"好友推荐"和"算法推荐"双重机制，不断迭代升级识别打 击模型，严禁低俗恶 ...

Group 1 - The core viewpoint of the article emphasizes the dual nature of algorithms, highlighting their benefits in daily life while also addressing the issues arising from their misuse, such as "information cocoon," "big data discrimination," and "induced addiction" [1][2] - The report outlines the accelerated governance efforts in China regarding algorithm misuse, including the implementation of the "Internet Information Service Algorithm Recommendation Management Regulations" in 2022 and the 2024 joint action by the Central Cyberspace Administration and other departments to address typical algorithm issues [1][2] - The article points out the challenges faced in algorithm governance, such as companies evading supervision under the guise of "technical neutrality," the lack of transparency in algorithm logic, and users' over-reliance on single platforms [2] Group 2 - The article suggests a multi-faceted approach to return algorithms to their intended purpose, advocating for stronger legal frameworks, increased penalties for algorithm misuse, and innovative technical governance that allows for controlled testing of algorithms [2] - It emphasizes the importance of creating a "co-governance ecosystem" involving various stakeholders, including regulatory bodies, companies, and consumers, to enhance market transparency and promote rational consumer decision-making [2] - The article concludes that algorithms should be guided by legal frameworks and ethical values to ensure they serve as a catalyst for social progress, balancing efficiency with fairness and responsibility [2]

Core Viewpoint - Recent updates in apps like Xiaohongshu and Weibo allow users to visualize and adjust their content preferences, addressing concerns about algorithm transparency and user control over recommendations [1][4]. Group 1: User Control Features - Xiaohongshu introduced a "content preference adjustment" feature that displays a pie chart of users' browsing preferences over the past week, allowing them to adjust recommendation intensity across categories such as fashion, food, and travel [2][3]. - Users can choose from three levels of recommendation intensity: "reduce recommendation," "default intensity," and "increase intensity," and can also delete or add content tags [2][3]. - Similar features have been implemented in Weibo and Bilibili, where users can manage their content preferences through a pie chart and adjust recommendation settings [2][3]. Group 2: Algorithm Governance and Concerns - There are growing concerns about algorithmic issues such as "killing familiarity" and "information cocooning," prompting users to call for more control over algorithmic recommendations [4]. - The implementation of the "Internet Information Service Algorithm Recommendation Management Regulations" in 2022 mandates that platforms provide options for users to opt-out of personalized recommendations and manage their user tags [4]. - In response to these concerns, major platforms have introduced options for users to toggle personalized recommendations, aligning with regulatory requirements [5]. Group 3: Regulatory Actions - A joint initiative by four ministries launched in November aims to address issues like "information cocooning" and algorithmic manipulation, with a focus on user rights and algorithm transparency [5]. - The initiative requires platforms to conduct self-assessments and report on their compliance with the regulations, which has led to various governance measures being announced by platforms like Douyin, Weibo, and Xiaohongshu [5].

Core Insights - The article discusses the recent trend of algorithm transparency among major Chinese internet platforms, highlighting the shift from "mysterious technology" to a more open approach in algorithm governance [1][2][5] - It emphasizes that while algorithm transparency can enhance user trust and improve platform governance, it also poses challenges to existing business models [1][4][8] Group 1: Algorithm Transparency - Douyin has publicly disclosed its algorithm for the first time, marking a significant move in the industry [2] - Other platforms like Meituan and Xiaohongshu have also initiated similar transparency measures, aiming to give users more control over algorithmic recommendations [2][3] - The trend is seen as a necessary step for internet companies to adapt to regulatory expectations and improve user engagement [1][5] Group 2: Challenges and Opportunities - The shift towards algorithm transparency is viewed as a double-edged sword, potentially disrupting established business models while also fostering a more ethical digital environment [1][3][8] - Concerns about issues like "information cocoon" and "big data discrimination" have prompted regulatory scrutiny, leading to a new governance framework for platform algorithms [5][6] - The article suggests that platforms must balance algorithmic efficiency with ethical considerations to avoid negative societal impacts [7][8] Group 3: Industry Implications - The article notes that algorithm governance will be a critical focus for internet companies by 2025, as they navigate the complexities of user trust and regulatory compliance [1][5] - It highlights the need for platforms to rethink their underlying logic and user expectations in light of algorithmic changes [8] - The consensus among industry experts is that algorithm transparency should not only meet regulatory demands but also enhance the overall user experience and societal values [4][8]

在算法决策的时代，法律如何定义边界？当人机交互变得日益复杂，责任归属该如何界定？从自动驾驶 归责到AIGC内容风险，从个人隐私保护到算法公正透明，我们必须面对技术革命带来的法律困境。技 术与伦理的平衡点在哪里？未来法律的变革方向是什么？ 4月20日，第七季"三联·新知大会"的首场活动"未来契约：AI时代的法律刻度"在三联韬奋书店美术馆店 举办。中国社会科学院文化法制研究中心研究员周辉、北京大学法学院副教授戴昕、清华大学公共管理 学院副教授陈天昊和对外经济贸易大学法学院教授张欣等法学专家深入探讨了AI时代的法治问题，剖 析如何构建一个既能促进创新又能守护人类核心价值的法律框架。 法律怎样定义算法？ 活动现场，周辉从算法的概念谈起，"简单来说，算法就是为了让电脑来解决问题而设计出来的一套计 算方法，这是当前的数字时代或者是AI时代的一个场景的界定。但在早期时代，算法更多的是用来解 决具体的算术问题，从特定条件得出相应的结果。到了近代，尤其是讨论到AI早期的一些概念——涉 及图灵机以及我们现在信息通讯技术的发展，使得我们对算法的理解，可能相对感觉会更加复杂。到了 大数据时代，基于这种技术的复杂性，相应的展示过程和呈 ...