AI Use Policy

Generative AI tools (e.g., large language models, AI coding assistants) are increasingly part of research workflows. AI technologies are being developed at a pace faster than the scientific community can comprehend. These tools are already revolutionizing science and can greatly enhance the work of quantitative ecology if used responsibly. However, there are many ethical, educational, and sustainability concerns regarding the use of generative AI. The Lab’s AI Use Policy is a living document that we expect to change substantially through time. Below are current guidelines regarding generative AI use in lab research, writing, and code:

• Disclosure: If AI tools are used in a substantive way to draft text, generate code, or analyze data for a manuscript, thesis, or proposal, this should be disclosed consistent with journal and funder policies.
• Verification: Lab members are responsible for independently verifying the accuracy of any AI-generated content, code, or analysis before it is used in a publication, presentation, or dataset. Models are trained to predict the most probable next word in a sequence and can hallucinate false, fabricated, or misleading information.
• Authorship: AI tools are not authors. Responsibility for the accuracy and integrity of all work remains with the human researchers.
• Data privacy: Before using AI tools, confirm both the AI tool’s data handling and privacy terms and the privacy and terms of use governing the data itself. Some data may not be input into AI tools under any circumstances, and some tasks, including reviewing manuscripts or grant proposals for journals or funders, must never involve AI tools, regardless of a tool’s own privacy terms. Lab members are responsible for knowing which data and tasks fall under these restrictions and for holding themselves accountable to following them.
• Environment: Training and running generative AI models have large environmental consequences, including large draws on energy, land-use, and water. We should aim to use AI tools purposefully, not reflexively, and ask whether a task benefits from AI assistance or if we are reaching for it by default.
• Education: The impacts, particularly long-term impacts, of generative AI on educational outcomes and skill building are only beginning to be understood. Relying on generative AI to do our thinking for us can compromise the acquisition of the skills we’re looking to build in the lab. Evidence from a recent pre-print indicates that AI use reduces code understanding, reading, and debugging abilities without delivering efficiency gains. We try to strategically use AI to support learning and analytical skills, not to bypass the struggle that builds them.

This policy will be revisited periodically as lab and scientific norms and values evolve.