CAREER NEWS31 August 2026A white paper gives research groups four priority models, each of which uses artificial intelligence differently.ByJenna Ahart0Jenna AhartJenna Ahart is a science journalist based in Washington DC.View author publicationsSearch author on: PubMed Google ScholarSave articleView saved researchArtificial-intelligence models can be used for tasks ranging from coding to taking meeting notes, depending on a laboratory’s requirements. Credit: da-kuk/GettyAs artificial intelligence pervades scientific research, journals, institutions and classrooms are all adopting their own rules on how to use large language models. But how can individual laboratories decide on the best ways to use AI to enhance their work? It depends on the research values that your team prioritizes, a white paper suggests.The paper, posted on the arXiv preprint server by a team of space-science researchers in July1 and not yet peer reviewed, suggests that there’s no one-size-fits-all strategy for adopting AI policies across research groups. Instead, the authors define four lab-group ‘archetypes’, each of which could benefit from using AI during different phases of the research process.The idea for the paper came when the team’s members saw AI policies being implemented at astronomy and government institutions, yet were “struggling with how we can even bring up the idea with our students and postdocs”, says co-author Sarah Burke-Spolaor, an astronomer at West Virginia University in Morgantown. “We can’t say, ‘just don’t use AI’, because people are already using it.”With the many possible applications of AI tools, from generating code to helping non-fluent English speakers with writing, the team realized that “we need to talk not about AI, but about what matters to us as a group”, Burke-Spolaor says. “And as the lead researcher, what kind of group do you want to sculpt?”What’s your personality type?To help researchers broach the subject, the paper proposes four lab archetypes, each with their own priorities and possible methods of AI use. Each labs’ traits are plotted on a radar diagram — inspired by the chart used to measure song difficulty in the video game Dance Dance Revolution, says Michelle Ntampaka, an astronomer at the Space Telescope Science Institute in Baltimore, Maryland, and co-author of the white paper. The authors included a similar worksheet at the end of the paper for researchers to work out their labs’ own priorities.Collection: ChatGPT’s impact on careers in scienceIf yours is a ‘high leverage’ lab, you might prioritize maximizing scientific impact with limited resources and could benefit from using large language models for code generation, data analysis and brainstorming ideas. A ‘craftsmanship’ lab is more focused on building expertise through practice and failure without taking shortcuts, and might adopt AI for writing assistance and meeting summaries. A ‘trustworthiness’ group, which is most concerned with reproducibility and transparency, might use AI for code debugging and administrative tasks, whereas ‘data stewardship’ teams prioritize responsible data management and tend to prefer self-hosted AI models for the sake of confidentiality.Importantly, the authors note, these archetypes are caricatures of philosophies and are not mutually exclusive, with most groups probably adopting elements from multiple categories. Lab priorities can also change according to circumstances, says Tari Tan, who oversees AI-related efforts in graduate education at Harvard Medical School in Boston, Massachusetts. “There are particular contexts within a lab’s research where the focus is on making sure that humans aren’t cognitively offloading to AI” — in other words, that they aren’t letting AI do the “hard thinking” for the sake of convenience or expediency, she says. “Or, if you’re on this really big, fast-moving project, then maximizing resources and moving fast might take priority.”Training baselinesdoi: https://doi.org/10.1038/d41586-026-02543-zSubscribe to Nature Briefing: Careers, an unmissable free weekly round-up of help and advice for working scientists.ReferencesNtampaka, M. et al. Preprint at arXiv https://doi.org/10.48550/arXiv.2607.20836 (2026).Download references AI and the PhD student: friend or foe? ChatGPT for students: learners find creative new uses for chatbots Universities are embracing AI: will students get smarter or stop thinking? Need a policy for using ChatGPT in the classroom? Try asking students Collection: ChatGPT’s impact on careers in scienceSubjectsCareersEducationMachine learningLatest on:CareersEducationMachine learningJobs Faculty Positions at Center for Infectious Disease Research, Westlake UniversitySeeking scientists in pathobiology, immunology, vaccinology, epidemiology, drug discovery, focusing on microbial infections and inflammations.Hangzhou, Zhejiang (CN)Center for Infectious Disease Research, Westlake University