From AI for Science to Agentic Science: A Survey on Autonomous Scientific Discovery
Jiaqi Wei, Yuejin Yang, Xiang Zhang, Yuhan Chen, Xiang Zhuang, Zhangyang Gao, Dongzhan Zhou, Guangshuai Wang, Zhiqiang Gao, Juntai Cao, Zijie Qiu, Xuming He, Qiang Zhang, Chenyu You, Shuangjia Zheng, Ning Ding, Wanli Ouyang, Nanqing Dong, Yu Cheng, Siqi Sun, Lei Bai, Bowen Zhou
2025-08-21
Summary
This paper discusses a new stage of artificial intelligence in science called "Agentic Science," where AI goes beyond just helping researchers to actively conducting scientific discovery on its own. It's like AI becoming a research partner, capable of coming up with ideas, planning experiments, running them, analyzing results, and improving the process, all without constant human guidance.
What's the problem?
The problem is that while AI has been used in science for a while, it's mostly been a tool that humans use. We haven't fully explored or understood how AI can take on the full responsibilities of scientific discovery, which traditionally requires human intelligence, creativity, and autonomy. This paper aims to organize and understand this shift from AI as a tool to AI as an independent scientific agent.
What's the solution?
The paper proposes a framework called "Agentic Science" to understand this evolution. They connect different ways of looking at autonomous AI in science, like focusing on the process, the autonomy itself, or the specific ways AI works. They then lay out five key abilities that allow AI to act autonomously in science, describe the typical steps of scientific discovery as a four-stage process that AI can follow, and review how these autonomous capabilities are being used in fields like biology, chemistry, materials science, and physics.
Why it matters?
This matters because it shows how AI is becoming a powerful, independent force in scientific progress. By understanding Agentic Science and its components, researchers can better develop and integrate these advanced AI systems, potentially speeding up discoveries and tackling complex scientific challenges that were previously out of reach for human researchers alone.
Abstract
Artificial intelligence (AI) is reshaping scientific discovery, evolving from specialized computational tools into autonomous research partners. We position Agentic Science as a pivotal stage within the broader AI for Science paradigm, where AI systems progress from partial assistance to full scientific agency. Enabled by large language models (LLMs), multimodal systems, and integrated research platforms, agentic AI shows capabilities in hypothesis generation, experimental design, execution, analysis, and iterative refinement -- behaviors once regarded as uniquely human. This survey provides a domain-oriented review of autonomous scientific discovery across life sciences, chemistry, materials science, and physics. We unify three previously fragmented perspectives -- process-oriented, autonomy-oriented, and mechanism-oriented -- through a comprehensive framework that connects foundational capabilities, core processes, and domain-specific realizations. Building on this framework, we (i) trace the evolution of AI for Science, (ii) identify five core capabilities underpinning scientific agency, (iii) model discovery as a dynamic four-stage workflow, (iv) review applications across the above domains, and (v) synthesize key challenges and future opportunities. This work establishes a domain-oriented synthesis of autonomous scientific discovery and positions Agentic Science as a structured paradigm for advancing AI-driven research.