The K3@MIT Gathering AI use policy
The K3@MIT Gathering AI use policy
This is not an event about proving K3 problems using AI. Our main goal is to facilitate the human process of doing mathematics together, an experience which we deeply value. We anticipate that many students currently use AI in some capacity in their current research workflow. To ensure effective working relationships and to foster trust between collaborators, we will require all working groups to submit a (relatively) comprehensive AI use policy that all group members agree to follow with respect to their project. The details of your policy are between you and your group members. That said, we have a few guidelines that you should consider to help you arrive at an effective policy in a timely manner.
1. Please respect your group members’ views. Everyone has complex, usually strongly-held views about AI, and speaking them aloud is an expression of vulnerability. Endeavour not to interrupt one another when speaking and ensure that every group member has been listened to. When deciding on a final policy, make sure everyone has a chance to comment and are willing to say that they’re comfortable with the group working under that final policy; in particular, we suggest acquiescing to the least permissive policy suggested by your group members.
2. Address at least the following categories of AI use:
a. Using AI to draft a proof before editing it for final inclusion;
b. Using AI to produce figures, tikz code, or other graphical elements of a paper;
c. Using AI as a “sounding board” to be suggested which ideas are good and which are poor;
d. Using AI to proofread final documents and proofs for grammar/style/proper english;
e. Using AI to proofread final documents for mathematical correctness;
f. Using AI to help produce code if necessary for the project;
g. Using AI for literature search;
h. Using AI for extracting/producing specific examples from the literature;
i. Using AI for translation;
j. Using google search (which has embedded AI);
k. If any AI use is permitted, which AI tools/companies are acceptable.
3. Think about what a complete AI disclosure would read if your project produced a paper, and be sure you’re comfortable with your name being on a paper with such an AI disclosure.
To give an idea of the style and comprehensiveness we expect, we have a sample policy below. Your group is welcome to deviate from the below sample in style and content, but it should be similarly comprehensive with the aim of fostering complete trust.
Sample AI use Policy:
Our group will not use AI systems in the writing of proofs, content, or code at any stage during our project, both in drafts and in any final documents. This includes using AI to write, proofread, or verify correctness of proofs, using AI to proofread or edit other writing, using AI to produce figures or LaTeX code, or using AI to produce other code for use in our project. All text, figures, and computer code at all stages of the project should be 100% human-generated. To avoid potential pollution of this goal, our group will not use AI systems directly as a sounding board, to ask questions, or to summarise papers. Our group permits the use of AI for finding examples in the literature, for finding papers and other sources, the use of google search, and for translation from a source text into our native languages. If a translated work is cited in our work, we will endeavour to verify the translation manually by using more traditional tools. Any AI use not covered explicitly by this policy should be okayed by all group members before use, and if any member objects that use will be avoided. After such a decision we will amend this policy to cover that use case. Group members using AI for something will okay the specific product before use and will add the name of the product to this policy.
The event was organized by Ayodeji Lindblad, Ollie Thakar, and Evan Scott. If you have any questions, please contact us at organizers-K3@MIT.edu