Discussions of AI in science often follow a familiar arc: AI begins as a tool that assists human researchers, then gradually takes on more responsibility until it becomes an autonomous scientist. A new preprint on arXiv challenges that framing by putting human-AI collaboration at the center.
The paper introduces the '9A framework' as a way to think about and design collaborative science. Rather than treating collaboration as an early stage on the road to autonomy, the authors present it as a distinct mode of work in which human and AI contributions are interleaved. The framework is meant to guide the design of such collaborative systems.
The significance lies in the shift of attention: instead of asking how to remove humans from the loop, designers might ask how to keep them meaningfully involved. The paper argues that human-AI collaboration is not merely a step toward autonomous AI scientists but a valuable mode in its own right. As a preprint, the work has not yet been peer-reviewed.