[Purpose/Significance] AI documentation serves as a critical tool and infrastructure for achieving risk prevention and control of AI systems, ensuring transparency, explainability, and accountability of AI systems, and promoting technology innovation. Clarifying the composition of AI documentation can facilitate effective AI governance. [Method/Process] This study analyzed AI documentation requirements in Sino-European AI legislation and conducted semi-structured interviews with three AI development companies of varying scales. Based on this analysis, the research proposed the composition of AI documentation for AI governance. [Result/Conclusion] To realize the role of AI documentation in AI governance, it is essential to clarify the composition. When defining its composition, the process should be guided by the provenance principle, forming comprehensive provenance documents that cover the full lifecycle of AI systems. Simultaneously, documentation should be differentiated based on the risk levels of AI systems and the specific stakeholders they address.
Lian Zhiying
,
Jiang Ling
. Research on the Role and Composition of AI Documentation for AI Governance[J]. Library and Information Service, 2026
, 70(9)
: 148
-156
.
DOI: 10.13266/j.issn.0252-3116.2026.09.013
[1] KÖNIGSTORFER F, THALMANN S. AI Documentation:a path to accountability[J]. Journal of responsible technology, 2022, 11:100043.
[2] FALCO G, SHNEIDERMAN B, BADGER J, et al. Governing AI safety through independent audits[J]. Nature machine intelligence, 2021, 3(7):566-571.
[3] GEBRU T, MORGENSTERN J, VECCHIONE B, et al.Datasheets for datasets[J]. Communications of the ACM, 2021,64(12):86-92.
[4] HOLLAND S, HOSNY A, NEWMAN S, et al. The dataset nutrition label:A framework to drive higher data quality standards[J]. Data protection and privacy, 2020, 12(12):1-25.
[5] MITCHELL M, WU S, ZALDIVAR A, et al. Model cards for model reporting[C]//Proceedings of conference on fairness,accountability, transparency. New York:Association for Computing Machinery, 2019:220-229.
[6] ARNOLD M, BELLAMY R K, HIND M, et al. FactSheets:increasing trust in AI services through supplier's declarations of conformity[J]. IBM journal of research and development, 2019,63(4/5):6, 1-6, 13.
[7] HUPONT I, MICHELI M, DELIPETREV B, et al. Documenting high-risk AI:a European regulatory perspective[J]. Computer,2023, 56(5):18-27.
[8] KONIGSTORFER F, THALMANN S. AI Documentation:a path to accountability[J]. Journal of responsible technology, 2022, 11:100043.
[9] MANTYMAKI M, MINKKINEN M, BIRKSTEDT T, et al.Defining organizational AI governance[J]. AI and ethics, 2022,2(4):603-609.
[10] AL ZADJALI H. Building the right AI governance model in Oman[C]//Proceedings of the 13th international conference on theory and practice of electronic governance. New York:Association for Computing Machinery, 2020:116-119.
[11] 鲁传颖.全球人工智能治理的目标、挑战与中国方案[J].当代世界, 2024(5):25-31.(LU C Y. Global AI governance:goals,challenges, and China's approach[J]. Contemporary world,2024(5):25-31.)
[12] SHEEHAN M. China's AI regulations and how they get made[J].Journal of international relations and sustainable development,2023(24):108-125.
[13] JENTZSCH S F, HOCHGESCHWENDER N. Don't forget your roots!using provenance data for transparent and explainable development of machine learning models[C]//IEEE Computer Society. 2019 34th IEEE/ACM international conference on automated software engineering workshop. Los Alamitos:IEEE Computer Society Conference Publishing Services, 2019:37-40.
[14] FROST L. Explainable AI and other questions where provenance matters[EB/OL].[2025-08-04]. https://iot.ieee.org/newsletter/january-2019/explainable-ai-and-other-questions-whereprovenance-matters.
[15] 连志英,蒋玲,张晓.档案来源观的后现代转向[J].档案学通讯, 2023(5):4-10.(LIAN Z Y, JIANG L, ZHANG X. The postmodern turn in the concept of archival sources[J]. Archives science bulletin, 2023(5):4-10.)
[16] UNESCO. Guidelines for the selection of digital heritage for long-term preservation-2nd edition[EB/OL].[2025-08-04]. https://repository.ifla.org/handle/123456789/1863.
[17] National Archives and Records Administration. Cognitive technologies white paper:Records management implications for internet of things, robotic process automation, machine learning,and artificial intelligence[EB/OL].[2025-08-04]. https://www.archives.gov/files/records-mgmt/policy/nara-cognitivetechnologies-whitepaper.pdf.
