[Purpose/Significance] As a key driving technology of artificial intelligence, algorithms have become deeply embedded in public administration, platform services, financial credit, employment, education, health care and national security. While algorithmic systems improve efficiency and promote digital innovation, they also bring about risks such as opaque decision-making, algorithmic discrimination, unclear responsibility, insufficient remedies and regulatory lag. The United States, relying on its global technological influence and relatively rich governance practices, has gradually formed a policy framework for algorithmic governance at the federal level. A systematic analysis of U.S. algorithmic governance policies is therefore helpful for understanding the evolution of global algorithmic governance and can provide useful references for China to improve its localized algorithmic governance system. [Method/Process] This study takes 12 U.S. federal policy texts related to algorithmic governance from 2018 to 2024 as research samples, including enacted laws, congressional bills, policy blueprints, joint statements and governance frameworks. Based on qualitative content analysis, the study uses NVivo12 Plus to conduct coding and category extraction. The research process includes policy text collection, text preprocessing, coding framework construction, inter-coder reliability testing and two rounds of coding. In the first round, the study focuses on algorithmic reporting and conducts open coding from the dimensions of report type, content requirements and responsible subjects. In the second round, the remaining policy provisions are coded from the bottom up, covering such dimensions as algorithmic transparency, algorithmic impact assessment, algorithmic audit, algorithmic regulation and algorithmic accountability. Through this process, the study identifies the core elements, internal relations and practical paths of U.S. algorithmic governance. [Result/Conclusion] The results show that the U.S. policy framework for algorithmic governance consists of five interrelated components: algorithmic transparency, algorithmic impact assessment, algorithmic audit, algorithmic regulation and algorithmic reporting. Algorithmic transparency serves as the basic premise for governance by requiring disclosure of the existence, process and platform operation of algorithms. Algorithmic impact assessment supplements transparency by focusing on risks, negative impacts and affected groups before or during algorithm deployment. Algorithmic audit further strengthens accountability by examining fairness, compliance and possible harm from an independent or supervisory perspective. Algorithmic regulation provides institutional support through the participation of commissions, federal agencies, state authorities, enterprises and other stakeholders. Algorithmic reporting functions as an information support system and a documentary basis for accountability by recording key information throughout the governance process. Overall, U.S. algorithmic governance shows four main characteristics: the governance goal has shifted from simple responsibility attribution to broader social reflection; governance subjects are diverse and involve complex interests; governance ideas integrate individual empowerment and institutional supervision; and the reporting system supports full-process accountability. [Innovation/ Value] The innovation of this study lies in its systematic coding of U.S. federal algorithmic governance policy texts and its construction of an integrated policy framework from the perspective of accountability. Compared with studies that mainly focus on legal interpretation or single policy analysis, this study reveals the internal relationship among transparency, assessment, audit, regulation and reporting, and clarifies how these elements jointly support algorithmic accountability. It also distinguishes algorithmic governance from data governance by emphasizing the decision-making process, risk transmission and responsibility tracing of algorithmic systems. [Insufficient/Improvement] This study still has some limitations. First, the research samples are mainly federal-level policy texts, while state-level legislation, industry standards and local governance practices are not fully included. Second, some policy texts are bills or policy proposals that have not yet become binding law, so their actual implementation effects require further observation. Third, the study mainly adopts text analysis and does not fully combine empirical data from enterprises, regulators or affected users. Future research can expand the sample scope, compare federal and state governance practices, and further examine the implementation effects of algorithmic governance policies in specific application scenarios.
[1] ISOAHO K, GRITSENKO D, MAKELA E. Topic modeling and text analysis for qualitative policy research[J]. Policy studies journal, 2021, 49(1): 300-324.
[2] 张莉莉, 朱子升. 算法歧视的法律规制: 动因、路径和制度完善[J]. 科技与法律(中英文), 2021(2): 15-21. (ZHANG L L, ZHU Z S. Legal regulation of algorithm discrimination: motivation, path and system improvement[J]. Science technology and law (Chinese & English), 2021(2): 15-21.)
[3] WILLIAMS R, CLOETE R, COBBE J, et al. From transparency to accountability of intelligent systems: moving beyond aspirations[J]. Data & policy, 2022, 4: e7.
[4] 陈慧君. 算法辅助量刑的风险成因与规范进路—— 以美国实践争议为切入点[J]. 青少年犯罪问题, 2023(2): 45-59. (CHEN H J. Risk causes and regulatory approaches of algorithm-assisted sentencing: taking the U.S. practice disputes as the entry point[J]. Issues on juvenile crimes, 2023(2): 45-59.)
[5] GRITSENKO D, MARKHAM A, POTZSCH H, et al. Algorithms, contexts, governance: an introduction to the special issue[J]. New media & society, 2022, 24(4): 835-844.
[6] 贾开. 人工智能与算法治理研究[J]. 中国行政管理, 2019(1): 17-22. (JIA K. Research on artificial intelligence and algorithm governance[J]. Chinese public administration, 2019(1): 17-22.)
[7] AVNOON N, KOTLIAR D M, RIVNAI-BAHIR S. Contextualizing the ethics of algorithms: a socio-professional approach[J]. New media & society, 2023, 26(10): 5962-5982.
[8] DE SILES E L. Soft law for unbiased and nondiscriminatory artificial intelligence[J]. IEEE technology and society magazine, 2021, 40(4): 77-86.
[9] LI X J, DENG G S, WANG X Z, et al. A hybrid recommendation algorithm based on user comment sentiment and matrix decomposition[J]. Information systems, 2023, 117: 102244.
[10] KRAFFT T D, ZWEIG K A, KONIG P D. How to regulate algorithmic decision‐making: a framework of regulatory requirements for different applications[J]. Regulation & governance, 2022, 16(1): 119-136.
[11] LIU X, GLOCKER B, MCCRADDEN M M, et al. Correction to lancet digit health 2022; 4: e384–97[J]. The lancet digit health, 2022, 4(6): e405.
[12] KAZIM E, KOSHIYAMA A S, HILLIARD A, et al. Systematizing audit in algorithmic recruitment[J]. Journal of intelligence, 2021, 9(3): 46.
[13] HU L, HE S, HAN Z, et al. Monitoring housing rental prices based on social media: an integrated approach of machine-learning algorithms and hedonic modeling to inform equitable housing policies[J]. Land use policy, 2019, 82: 657-673.
[14] KAUN A. Suing the algorithm: the mundanization of automated decision-making in public services through litigation[J]. Information, communication & society, 2022, 25(14): 2046-2062.
[15] ETTER M, ALBU O B. Activists in the dark: social media algorithms and collective action in two social movement organizations[J]. Organization, 2021, 28(1): 68-91.
[16] BASUKIE J, WANG Y, LI S. Big data governance and algorithmic management in sharing economy platforms: a case of ridesharing in emerging markets[J]. Technological forecasting and social change, 2020, 161: 120310.
[17] 周文泓, 吴琼, 田欣, 等. 美国联邦政府数据治理的实践框架研究—— 基于政策的分析及启示[J]. 现代情报, 2022, 42(8): 127-135. (ZHOU W H, WU Q, TIAN X, et al. Research on the practice framework of U.S. Federal government data governance: policy-based analysis and enlightenment[J]. Modern information, 2022, 42(8): 127-135.)
[18] 丁玮. 数据主义视角下美国跨境数据政策演进及我国的应对[J]. 杭州师范大学学报(社会科学版), 2021, 43(1): 120-129. (DING W. Evolution of U.S. cross-border data policies from the perspective of dataism and China’s countermeasures[J]. Journal of Hangzhou Normal University (social sciences edition), 2021, 43(1): 120-129.)
[19] 余丽, 张涛. 美国数据有限性开放政策及其对全球网络安全的影响[J]. 郑州大学学报(哲学社会科学版), 2019, 52(5): 12-19. (YU L, ZHANG T. U.S. data limited openness policy and its impact on global cybersecurity[J]. Journal of Zhengzhou University (philosophy and social sciences edition), 2019, 52(5): 12- 19.)
[20] 马海群, 蔡庆平, 崔文波, 等. 美国数据与算法安全治理: 进路、特征与启示[J]. 信息资源管理学报, 2023, 13(1): 52-64. (MA H Q, CAI Q P, CUI W B, et al. Governance of data and algorithm security in the United States: approaches, characteristics and enlightenment[J]. Journal of information resources management, 2023, 13(1): 52-64.)
[21] 陆凯. 美国算法治理政策与实施进路[J]. 环球法律评论, 2020, 42(3): 5-26. (LU K. U.S. algorithm governance policies and implementation approaches[J]. Global law review, 2020, 42(3): 5-26.)
[22] 邹丽雪. 美国推动人工智能在生物医学领域应用的政策分析及其对中国的启示[J]. 科技管理研究, 2023, 43(16): 31- 38. (ZOU L X. Policy analysis of the U.S. promotion of artificial intelligence in biomedicine and its implications for China[J]. Science and technology management research, 2023, 43(16): 31- 38.)
[23] SAURWEIN F, JUST N, LATZER M. Governance of algorithms: options and limitations[J]. Digital Policy, Regulation and Governance, 2015, 17(6): 35-49.
[24] WIERINGA M. What to account for when accounting for algorithms: a systematic literature review on algorithmic accountability[C]//Proceedings of the 2020 conference on fairness, accountability, and transparency. Barcelona: ACM, 2020: 1-18.
[25] 姜天海, 贾萍萍, 张增一. 政策工具视角下欧美国家与国际组织开放科学政策文本分析及其启示[J]. 图书情报工作, 2022, 66(22): 119-133. (JIANG T H, JIA P P, ZHANG Z Y. Textual analysis and enlightenment of open science policies in European and American countries and international organizations from the perspective of policy tools[J]. Library and information service, 2022, 66(22): 119-133.)
[26] FRIEDMAN L M, HAYDEN G M. American law: an introduction[M]. New York: Oxford University Press, 2017.
[27] 杨慧, 杨建林. 融合LDA模型的政策文本量化分析—— 基于国际气候领域的实证[J]. 现代情报, 2016, 36(5): 71-81. (YANG H, YANG J L. Quantitative analysis of policy texts combining LDA model: an empirical study based on the international climate field[J]. Modern information, 2016, 36(5): 71-81.)
[28] GOVE R. Automatic narrative summarization for visualizing cyber security logs and incident reports[J]. IEEE transactions on visualization and computer graphics, 2021, 28(1): 1182-1190.
[29] 刘佳, 黄敏豪, 林玥, 等. 算法素养对推荐信息用户采纳行为的影响研究[J]. 图书情报工作, 2025, 69(5): 13-27. (LIU J, HUANG M H, LIN Y, et al. Research on the influence of algorithmic literacy on users' adoption behavior of recommended information[J]. Library and Information Service, 2025, 69(5): 13- 27.)
[30] 於春. 算法、权力与治理：智能时代美国新闻自由与民主政治的反思[J]. 当代世界与社会主义, 2022(3): 117-122. (YU C. Algorithm, power and governance: reflection on U.S. freedom of the press and democratic politics in the intelligent era[J]. Contemporary world & socialism, 2022(3):117-122.)
[31] JOHNON D G. Algorithmic accountability in the making[J]. Social philosophy and policy, 2021, 38(2): 111-127.
[32] 汪庆华. 算法透明的多重维度和算法问责[J]. 比较法研究, 2020(6): 163-173. (WANG Q H. Multiple dimensions of algorithm transparency and algorithm accountability[J]. Comparative law studies, 2020(6): 163-173.)
[33] 孟天广, 李珍珍. 治理算法：算法风险的伦理原则及其治理逻辑[J]. 学术论坛, 2022, 45(1): 9-20. (MENG T G, LI Z Z. Governing algorithms: ethical principles of algorithm risks and their governance logic[J]. Academic forum, 2022, 45(1): 9-20.)
[34] RAMCHURN S D, STEIN S, JENNINGS N R. Trustworthy human-AI partnerships[J]. Iscience, 2021, 24(8): 102891.
[35] ZERILLI J, KNOTT A, MACLAURIN J, et al. Transparency in algorithmic and human decision-making: is there a double standard?[J]. Philosophy & technology, 2019, 32(4): 661-683.
[36] RAAB C D. Information privacy, impact assessment, and the place of ethics[J]. Computer law & security review, 2020, 37: 105404.
[37] 张恩典. 算法影响评估制度的反思与建构[J]. 电子政务, 2021(11), 57-68. (ZHANG A D. Reflection and construction of algorithm impact assessment system[J]. E-government, 2021(11), 57-68.)
[38] KUTEYNIKOV D, IZHAEV O, LEBEDEV V, et al. Human rights during the mass introduction of artificial intelligence and robotic systems into public life[J]. Cuestiones políticas, 2021, 39(71): 609–620.
[39] MOKANDER J, JUNEJA P, WATSON D S, et al. The US algorithmic accountability act of 2022 vs. the EU artificial intelligence act: what can they learn from each other?[J]. Minds and machines, 2022, 32(4): 751-758.
[40] KULYNYCH B, OVERDORF R, TRONCOSO C, et al. POTs: protective optimization technologies[C]//Proceedings of the 2020 conference on fairness, accountability, and transparency. New York: ACM, 2020: 177-188.
[41] MOSS E, WATKINS E A, SINGH R, et al. Assembling accountability: algorithmic impact assessment for the public interest[EB/ OL]. [2025-10-05]. https://datasociety.net/library/assembling-accountability-algorithmic-impact-assessment-for-the-public-interest/.
[42] 张欣, 宋雨鑫. 算法审计的制度逻辑和本土化构建[J]. 郑州大学学报(哲学社会科学版), 2022, 55(6): 33-42. (ZHANG X, SONG Y X. Institutional logic and localized construction of algorithm auditing[J]. Journal of Zhengzhou University (philosophy and social sciences edition), 2022, 55(6): 33-42.)
[43] BRADFORD A. The Brussels effect: how the European Union rules the world[M]. New York: Oxford University Press, 2020.
[44] SLACK D, HILGARD S, JIA E, et al. Fooling lime and shap: adversarial attacks on post hoc explanation methods[C]//Proceedings of the AAAI/ACM conference on AI, ethics, and society. New York: ACM, 2020: 180-186.
[45] HAGENDORFF T. The ethics of AI ethics: an evaluation of guidelines[J]. Minds and machines, 2020, 30(1): 99-120.
[46] SHNEIDERMAN B. The dangers of faulty, biased, or malicious algorithms requires independent oversight[J]. Proceedings of the national academy of sciences, 2016, 113(48): 13538-13540.
[47] VOAS J, MILLER K. Software certification services: encouraging trust and reasonable expectations[J]. IT professional, 2006, 8(5): 39-44.
[48] BAROCAS S, HOOD S, ZIEWITZ M. Governing algorithms: a provocation piece[EB/OL]. [2025-10-05]. https://governingalgorithms.org/resources/provocation-piece/.
[49] HAJRIC E. AI and data rights considerations for U.S. policy[J]. IEEE technology and society magazine, 2021, 40(3): 93-98.
[50] BAYKURT B. Algorithmic accountability in U.S. cities: transparency, impact, and political economy[J]. Big data & society, 2022, 9(2): 20539517221115426.
[51] 毕达天, 张雪, 米艳霖, 等. 算法感知视角下跨平台信息推荐用户接受意愿研究[J]. 图书情报工作, 2023, 67(21): 100-110. (BI D T, ZHANG X, MI Y L, et al. Research on users' willingness to accept cross-platform information recommendation from the perspective of algorithm perception[J]. Library and Information Service, 2023, 67(21): 100-110.)
[52] DIAKOPOULOS N. Algorithmic accountability[J]. Digital journalism, 2015, 3(3): 398–415.
[53] 许智鑫. 人工智能时代的算法规制及其制度建构[D]. 北京: 北京邮电大学, 2021.(XU Z X. Legal regulation of algorithms and its Institutional construction in the age of artificial intelligence[D]. Beijing: Beijing University of Posts and Telecommunications, 2021.)
[54] 刘晓丽, 崔文波, 张涛. 生成式人工智能视域下情报分析算法风险多重治理机制研究[J]. 图书与情报, 2024(5): 77-84. (LIU X L, CUI W B, ZHANG T. Research on the multi-layer governance mechanism of algorithm risks in intelligence analysis under generative artificial intelligence[J]. Library and information, 2024(5): 77-84.)
[55] 张涛, 吕骞慧. 欧美生成式人工智能治理实践对中国的启示研究[J]. 农业图书情报学报, 2025, 37(4): 12-23. (ZHANG T, LYU Q H. Generative AI governance practices in Europe and the United States and the enlightenment for China[J]. Journal of library and information science in agriculture, 2025, 37(4): 12-23.)
[56] 维尔, 马图斯, 戈尔瓦, 等. 人工智能与全球治理：模式、理据和紧张关系[J]. 俄罗斯学刊, 2024, 14(3): 49-72. (VEALE M, MATUS K, GORWA R, et al. AI and global governance: modalities, rationales, tensions[J]. Russian studies journal, 2024, 14(3): 49-72.)
