Mechanisms and modes in AI for Science(AI4S) research: A perspective from knowledge production modes

Abstract

Abstract:

Artificial Intelligence (AI) has demonstrated immense potential in the scientific domain. AI for Science (AI4S) is rapidly advancing and gaining widespread adoption, driving a comprehensive transformation in scientific research paradigms. However, the characteristics and mechanisms of AI4S in empowering scientific innovation remain unclear, and its strategic direction has yet to be systematically planned. Focusing on the 2024 Nobel Prize in Chemistry-winning projects, AlphaFold and Rosetta, this study employs a comparative case analysis to examine the distinct features and mechanisms of AI-enabled scientific research. It proposes two AI4S empowering models: "AI+" and "+AI." From the perspective of knowledge production modes, the study further analyzes the knowledge production characteristics of these two models. It identifies the emergence of "intelligence" as a fifth major actor in the "AI+" model, with knowledge production exhibiting traits of growth, independence, collaboration, and platformization. Additionally, the "AI+" and "+AI" models each have their own suitable application scenarios and frameworks. Finally, this paper offers policy recommendations, including differentiated strategies for AI+ and +AI, prioritizing small-model development, and advancing infrastructure construction. These findings contribute to both AI4S-related research and knowledge production mode theory, while also providing insights for China's development of the AI industry, exploration of AI applications, and advancement of AI models.

Key words: AI for Science, knowledge production, empowerment mechanism, AI Model

摘要：

人工智能(AI)在科学领域展现出巨大潜力,人工智能赋能科学研究(AI for Science,以下简称AI4S)正快速发展与普及并推动科学研究范式全面变革,但当前AI4S赋能科研创新的特征与机制依然不明晰,AI4S布局方向仍未形成系统谋划。聚焦2024年获得诺贝尔化学奖的AlphaFold与Rosetta项目,本研究采取对比案例研究方法,分析AI赋能科学研究的不同特征与机制,提出“AI+”与“+AI”两种AI4S赋能模式。基于知识生产模式视角,本研究进一步分析两种模式的知识生产特征,提出在“AI+”模式中存在“智”第五大主体,其知识生产具有成长性、独立性、协同性、平台性特征,且“AI+”与“+AI”模式各自具有适配场景与应用模型。最后,本文提出AI+与+AI差异化战略、重视小模型开发与推动基础设施建设等政策建议。以上结论对AI4S相关研究及知识生产模式理论均有一定贡献,也对我国发展AI产业、挖掘AI场景、开发AI模型具有启示意义。

关键词: AI4S, 知识生产, 赋能机制, AI模型

CLC Number:

G301

LI Tuo-yu, WU Jia-yi. Mechanisms and modes in AI for Science(AI4S) research: A perspective from knowledge production modes[J]. Studies in Science of Science, 2026, 44(2): 235-244.

李拓宇, 吴佳一. 人工智能赋能科学研究的机制与模式分析——基于知识生产模式视角[J]. 科学学研究, 2026, 44(2): 235-244.

Figures/Tables 9

References 21

[1]	Wang H, Fu T, Du Y, et al. Scientific discovery in the age of artificial intelligence[J]. Nature, 2023, 620(7972): 47-60. DOI
[2]	The Nobel Prize in Chemistry 2024 - Award ceremony speech - NobelPrize.org[Z]. 2025: 2025.
[3]	王飞跃, 缪青海, 张军平, 等. 探讨AI for Science的影响与意义:现状与展望[J]. 智能科学与技术学报, 2023, 5(1): 1-6. DOI
	Wang F Y, Miao Q H, Zhang J P, et al. Exploring the impact and significance of AI for Science: Current status and prospects[J]. Chinese Journal of Intelligent Science and Technology, 2023, 5(1): 1-6.
[4]	张越, 郭玥, 余江. 通用人工智能驱动的科研新范式:理论与实践[J]. 科学学研究, 2025, 43(4):673-682.
	Zhang Y, Guo Y, Yu J. A New AI-driven research paradigm: Theory and practice[J]. Studies in Science of Science, 2025, 43(4):673-682.
[5]	李国杰. 智能化科研(AI4R):第五科研范式[J]. 中国科学院院刊, 2024, 39(1): 1-9.
	Li G J. Intelligentized scientific research (AI4R): The fifth research paradigm[J]. Bulletin of Chinese Academy of Sciences, 2024, 39(1): 1-9.
[6]	孙坦, 张智雄, 周力虹, 等. 人工智能驱动的第五科研范式(AI4S)变革与观察[J]. 农业图书情报学报, 2023, 35(10): 4-32. DOI
	Sun T, Zhang Z X, Zhou L H, et al. The transformation and observation of AI-Driven fifth research paradigm (AI4S) [J]. Journal of Library and Information Science in Agriculture, 2023, 35(10): 4-32.
[7]	王飞跃, 缪青海. 平行科学:大模型时代AI4S的前沿技术与框架体系[J]. 人民论坛·学术前沿, 2024(14): 64-79.
	Wang F Y, Miao Q H. Parallel science: Frontier technologies and framework systems of AI4S in the Era of Large Models[J]. Frontiers, 2024(14): 64-79.
[8]	Nonaka I. A dynamic theory of organizational knowledge creation[J]. Organization Science, 1994, 5(1): 14-37. DOI URL
[9]	Alavi M, Leidner D E, Mousavi R. A knowledge management perspective of Generative Artificial Intelligence[J]. Journal of the Association for Information Systems, 2024, 25(1): 1-12. DOI URL
[10]	Jarrahi M H, Askay D, Eshraghi A, et al. Artificial intelligence and knowledge management: A partnership between human and AI[J]. Business Horizons, 2023, 66(1): 87-99. DOI URL
[11]	程鹏, 童松, 崔鹤, 等. AI时代下科学知识生产的伦理困境与策略[J]. 科学学研究, 2024, 42(11): 2254-2261.
	Cheng P, Tong S, Cui H, et al. Ethical dilemmas and strategies in scientific knowledge production in the AI Era[J]. Studies in Science of Science, 2024, 42(11): 2254-2261.
[12]	Carayannis E G, Campbell D F J. Open innovation diplomacy and a 21st Century fractal research, education and innovation (FREIE) Ecosystem: Building on the quadruple and quintuple Helix innovation concepts and the “Mode 3” knowledge production system[J]. Journal of the Knowledge Economy, 2011, 2(3): 327-372. DOI URL
[13]	Gibbons C L P S. The New Production of Knowledge: The Dynamics of Science and Research in Contemporary Societies[M]. Sage Pub., 1994.
[14]	Alavi M, Leidner D E. Review: Knowledge management and knowledge management systems: Conceptual foundations and research issues[J]. MIS Quarterly, 2001, 25(1): 107-136. DOI URL
[15]	肖峰. 生成式人工智能与知识生产新形态——关于三阶知识生产的探析[J]. 学术研究, 2023(10): 50-57.
	Xiao F. Generative Artificial Intelligence and the new form of knowledge production—An analysis of three-order knowledge production[J]. Academic Research, 2023(10): 50-57.
[16]	Lu C. Rethinking artificial intelligence from the perspective of interdisciplinary knowledge production[J]. AI & Society, 2024, 39(6): 3059-3060.
[17]	姜华. 从辛弃疾到GPT:人工智能对人类知识生产格局的重塑及其效应[J]. 南京社会科学, 2023(4): 135-145.
	Jiang H. From Xin Qiji to GPT: The reshaping of human knowledge production by Artificial Intelligence and its effects[J]. Social Sciences in Nanjing, 2023(4): 135-145.
[18]	喻国明, 苏芳. 范式重构、人机共融与技术伴随:智能传播时代理解人机关系的路径[J]. 湖南师范大学社会科学学报, 2023, 52(4): 119-125.
	Yu G M, Su F. Paradigm reconstruction, human-machine integration, and technology companionship: Pathways to understanding Human-Machine relationships in the Era of intelligent communication[J]. Journal of Social Science of Hunan Normal University, 2023, 52(4): 119-125.
[19]	李梦薇, 徐峰, 高芳. 人工智能应用场景的界定与开发[J]. 中国科技论坛, 2021(6): 171-179.
	Li M W, Xu F, Gao F. Definition and development of AI application scenarios[J]. Forum on Science and Technology in China, 2021(6): 171-179.
[20]	Treviso M, Ji T, Lee J, et al. Efficient methods for natural language processing: A survey[J]. Transactions of the Association for Computational Linguistics, 2023, 11: 826-860. DOI URL
[21]	Wang Q, Cardiff B, Frappe A, et al. Tiny models are the computational saver for Large Models[J]. Computer Vision - ECCV 2024, PT LVI, 2025, 15114: 163-182.

维度	知识生产1.0	知识生产2.0	知识生产3.0
驱动力	学术理论驱动	问题需求驱动	社会需求和多方合作驱动
生产特点	封闭单一、学科驱动式生产	跨学科、应用导向式生产	跨领域、动态开放、创新网络协同式生产
生产主体	大学内部	多主体协同	多主体合作网络
依赖的知识	传统科学知识与理论	跨学科知识与理论	跨领域知识、系统理论
结果产出	基础学科知识	应用性、跨学科知识	综合性知识
创新性	线性创新	横向协同创新	非线性网络协同创新

维度	知识生产1.0	知识生产2.0	知识生产3.0
驱动力	学术理论驱动	问题需求驱动	社会需求和多方合作驱动
生产特点	封闭单一、学科驱动式生产	跨学科、应用导向式生产	跨领域、动态开放、创新网络协同式生产
生产主体	大学内部	多主体协同	多主体合作网络
依赖的知识	传统科学知识与理论	跨学科知识与理论	跨领域知识、系统理论
结果产出	基础学科知识	应用性、跨学科知识	综合性知识
创新性	线性创新	横向协同创新	非线性网络协同创新

数据来源	数据分类
一手数据	AI for药物研究领域专家访谈,包括AI研发专家与药物研发专家
	AI与物理、AI与医学、AI与药学、AI与能源等学生访谈
	AI4S相关会议资料
二手数据	通过案例企业网站、公共媒介获取的资料
	通过权威媒体、专业媒体人报道获取的资料
	通过案例企业内部文件获得的资料
	通过热门网络论坛获取的评价信息

数据来源	数据分类
一手数据	AI for药物研究领域专家访谈,包括AI研发专家与药物研发专家
	AI与物理、AI与医学、AI与药学、AI与能源等学生访谈
	AI4S相关会议资料
二手数据	通过案例企业网站、公共媒介获取的资料
	通过权威媒体、专业媒体人报道获取的资料
	通过案例企业内部文件获得的资料
	通过热门网络论坛获取的评价信息

	共同点	差异点
	共同点	AlphaFold	Rosetta
主导者	AI与传统领域的融合与协作	AI主导	行业或领域主导
驱动力	AI技术的实际落地与行业应用	AI技术(数据与算法)驱动	行业需求驱动
赋能特点	提高效率或解决问题	强调跨界融合和系统性重塑。	强调具体环节优化或提升效率
适用范围	跨领域实践	广泛,通常涵盖多个领域、系统或行业之间的融合	具体,通常限定在某一行业或特定环节的优化中
赋能目标	提升生产力和创新能力	打造一个以AI为核心的新生态系统	提升现有行业的效率与效果,逐步优化产业运作
AI赋能机制		AI+	+AI