GenAI Policy

GenAI Policy

There are many generative AI tools out there, from general-purpose large language models (LLMs) and image generation models to research-grade software that is related to our own work. The current policy is described below, which is in addition to—but does not replace—any  Princeton-wide policies  on the use of GenAI tools.

General Comments

  • If you use a cloud-based LLM, at minimum you must turn off data sharing so that your conversations are not used for model training.
  • With Claude, this can be found in Settings > Privacy > Help improve Claude (should be toggled off)
  • With ChatGPT, this can be found in Settings > Data controls > Improve the model for everyone (should be "Off")
  • If it is determined that you are over-relying on GenAI tools in a manner that impedes your learning or if you use GenAI tools in an inappropriate manner, your privileges for using them may be revoked.
  • It is important to be aware of how our skills may be impacted when we overly rely on GenAI tools. I ask that you read " AI gravity ", which is a nice perspective on the topic.
  • Currently, there are no plans for the group to pursue research directions that specifically involve LLMs or agentic AI. We will leave that to others.

Using LLMs in Practice

  • If you are using a cloud-based LLM for research, you might as well be using the "good stuff" and not the limited free models. Everyone in the group is welcome to have a subscription (paid for by the lab) to Claude Pro with access to Claude Code or an equivalent service of your choice. Ask Andrew for the details, as the group has several existing licenses. You should generally be using your @princeton.edu email for such services.
  • You should not solely rely on the web-based chat interface, which is limited. You are encouraged to download the desktop interface as well as the command-line version to ensure they can call tools in a meaningful way.

Encouraged Uses

You are encouraged to use LLMs in scenarios where it would increase the quality, rigor, or visual appeal of your work. In other words, if it allows you to do better science or communicate better science without removing your own autonomy in the process, this is generally worth considering. Some examples are below:
  • Use LLMs to check for errors or typos in a manuscript you are preparing. This is often quite fruitful, and while you do not need to address every point, I expect that everyone should be doing this for all manuscripts to catch issues in advance of the review process.
  • Use LLMs to help create visually appealing structure visualizations or animations of your simulation results (e.g. structures, trajectories). This can be done by having the LLM interface with relevant tools for visualization purposes (e.g. ASE with POV-Ray, Ovito, Blende via the MCP Server). In this case, the LLM is serving to elevate the visuals you would otherwise already show based on your real data.
  • Use LLMs to help make schematics for more effective presentations, such as by having it interface with shapes in PowerPoint or Illustrator directly.
  • Use LLMs for troubleshooting (e.g. compiler error, Python package installation error). This ends to work quite well.
  • Use LLMs to help with coding and/or open-source software development. If the LLM will help you carry out more robust analyses or test hypotheses more quickly, you are encouraged to use such tools. However, as I outline below, you are fully responsible for all of your work and being able to defend it fully and so you must self-reflect on your own knowledge as you do this.

Cautionary Notes

  • In general, image-based GenAI models should be avoided in scientific research except for ideation. There are better ways to make nice figures. Please no AI slop images in presentations.
  • LLMs can be used to assist with the writing process, but there are many important caveats that you must adhere to:
  • The ideas must be your own, not those of the LLM.
  • The text must be in your own words. You should not copy text from an LLM wholesale. It is obvious if you do.
  • An acceptable use could include something like: "Here is a paragraph I am working on. The second sentence does not flow very nicely. Can your help me identify what is not working and how I can fix this?" And then you could modify your text, drawing inspiration from the LLM's response.
  • An unacceptable use would be something like: "I need to write an introductory paragraph on why MOFs are great. Please write this for me."
  • Note that LLM-generated writing is, on average, still quite mediocre. You should strive to do better than mediocre.
  • If Pangram flags your writing as AI-generated, you have gone too far.
  • LLMs should never be used to manage references. Use Zotero for actual reference management. Remember that you should be writing your text with references in mind as you write, not trying to sprinkle in random references after the fact.
  • There is a zero-tolerance policy for hallucinated content of any kind, including references.
  • LLMs can be used to assist with coding, but you must still be in control of what you are doing. Here are some guidelines:
  • You must understand the code you use, and you should not rely on LLM-generated code without carefully checking the details. You do not want to get in a situation where your research results are impacted without your knowledge. If you can't clearly defend what you have done in detail, you have gone too far.
  • If an LLM generates a coding pattern that you do not fully understand, take the time to figure it out.
  • If you are using a new code for the first time, you should not rely on an LLM to do it for you. This is very dangerous since you will not know if errors are made, and you will not learn the skills you need to improve as a programmer.
  • LLMs can be very helpful in improving the content of a presentation, but like writing, you must be the intellectual driver of the content. If the audience can tell it is AI generated, you have gone too far. Do not use LLMs to make your slide deck wholesale.
  • LLMs can be used for ideation (e.g. "discussing" possible research directions, explaining difficult concepts), but note that their creativity remains quite limited and project ideas are often flawed in subtle ways. Be cautious.
  • If your writing, plots, or presentations can easily be discerned as LLM-generated, then you have likely gone too far.
  • It hopefully goes without saying, but using GenAI tools to intentionally fabricate data is considered academic misconduct and will be treated as such. Don't even think about it.