Who decides when AI helps with science?
Zheng, Pan and Lan propose a nine-part guide to research with AI. It covers how work is split, how people and tools stay in touch, and how results are checked. This is a design plan, not a trial showing better results.
What the source reports
Zheng, Pan and Lan offer a nine-part design for people and AI to work together in science. Their plan asks teams to name the task, the actor, the decision owner, and the proof they will accept. It also asks how people will coordinate and how they will check results. The Beyond Prompting digest turns these ideas into four questions for a handoff card. That card is the curator’s exercise, not a tested part of the paper. The source is a preprint abstract and metadata, not a full review or a trial. It does not show that teams using the plan produce better science.
Original source
- Title
- Designing Human-AI Collaborative Science: The 9A Framework
- Author
- Zihan Zheng, Linyue Pan, Xin Lan
- Publication
- arXiv
- Date
- Friday, October 16, 2026
Also covered by the other newsletter.