genai policy

This is an evolving document, for obvious reasons. But it reflects our thoughts as of August, 2026.

As a lab, we value the parts of the research process that require us to engage critically with our respective subjects. Brainstorming, writing, and comprehension of relevant topics are some of the most important skills needed for contributing novel ideas to any field. We believe that GenAI interferes with these crucial stages of the research process and opt out of using it for writing, brainstorming, developing methodologies, and any task that requires diligent human thought.

We also understand that generative AI, used when communicating with each other, can negatively impact our relationships. See here for something that explores this in detail. To that end, we commit to always being explicit about when we are using generative AI in any form of communication, with ourselves or others outside the lab. This commitment is one small way we’re committed to clawing our way back out of the— abyss— :).

Ideally, in fact, we avoid this all together, but we also understand that these technologies can be extremely helpful various stages of life and research. In the context of our work, we believe there exists mundane tasks that must be done but don’t build the skills necessary to be an effective researcher. It is in these situations that we find GenAI to be most valuable.

Overall, then, we are cautious when using GenAI tools. We encourage careful discretion when using generative tools, clear documentation of how and when these tools are used during the research process, and explicit acknowledgement when used in communication with other people. We recognize that GenAI can generate incorrect but plausible responses and always review responses for accuracy.

Below you can find some more specific use cases we’ve been thinking about:

Example uses we feel like we’re generally down with but acknowledge its a slippery slope on any of them we know, we know

• Debugging
• Formatting for certain venues
• Revision / robustness of approach (e.g. am I violating statistical assumptions with the model I am making use of)
• Plot generation/exploratory tools
• Finding potentially useful papers*
• Image generation for slides?

Example uses we think are less productive in the short or long term

• Drafting papers
• Ideating
• Slide creation (although animation using GenAI is OK, generating your conclusion slide, e.g., is not)

*Sources and citations should always be manually verified