[ADMIN] Adding an MDAnalysis AI tools policy. #5210
Conversation
Adds a markdown document that defines MDAnalysis' current stance on AI tools for contributions.
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## develop #5210 +/- ##
===========================================
- Coverage 92.72% 92.71% -0.02%
===========================================
Files 180 180
Lines 22475 22475
Branches 3190 3190
===========================================
- Hits 20841 20838 -3
- Misses 1177 1179 +2
- Partials 457 458 +1 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
Co-authored-by: Rocco Meli <r.meli@bluemail.ch>
| AI assitance is deemed acceptable. However, if code generate exceeds minimal, sporadic amounts (e.g. repeated or large multi-line blocks), | ||
| it would be considered fully AI-generated and, as defined in section #1, is not acceptable. | ||
|
|
||
| As per section #0, where possible please state that you are using AI assistance via an IDE. |
There was a problem hiding this comment.
Remove "where possible" as this puts the burden on us. Instead, it should be a contributors responsibility to know when they are using AI features.
| As per section #0, where possible please state that you are using AI assistance via an IDE. | |
| As per section #0, please state when you are using AI assistance via an IDE. |
There was a problem hiding this comment.
With IDEs, my main worry is that it's just easy to forget - like I'm a vim guy, but sometimes I use vscode and then the auto complete happens but I forget about it.
|
|
||
| ### 6. Consequences for failure to comply to this policy | ||
|
|
||
| Any failures to comply with this policy should be reported to the MDAnalysis core developers. |
There was a problem hiding this comment.
Do we need to link to contact information?
There was a problem hiding this comment.
Yeah the whole thing needs links, and then itself being linked in other locations. I'm hoping that to be a task once we've agreed on the text.
| - Code submitted for inclusion in MDAnalysis | ||
| - Documentation of any kind | ||
| - Communication content (e.g. issues, discussions, pull request descriptions, and reviews) | ||
|
|
There was a problem hiding this comment.
I'd add something like the following as a preamble:
| We recognize that the use of AI in open source software raises many difficult questions and acknowledge that views, approaches, and technical capabilities are rapidly changing. | |
| We invite comments and further discussion (via our [discussion channels](https://www.mdanalysis.org/pages/community/)) and expect to periodically review this policy. | |
There was a problem hiding this comment.
I had the same thought yesterday. We should make clear that the policy will evolve over time, as we learn and the landscape changes.
There was a problem hiding this comment.
Do we want to date or version this policy?
There was a problem hiding this comment.
I think that's a great idea. I agree with your point that this should be updated over time.
tylerjereddy
left a comment
tylerjereddy
left a comment
There was a problem hiding this comment.
FWIW, the sklearn PR template currently has this:
AI usage disclosure
I used AI assistance for:
- Code generation (e.g., when writing an implementation or fixing a bug)
- Test/benchmark generation
- Documentation (including examples)
- Research and understanding
|
|
||
| ### 1. Fully generated code is not allowed | ||
|
|
||
| Fully AI-generated contributions, such as content generated by prompting a large language model (e.g. claude code or chatgpt) are not permitted in MDAnalysis. |
There was a problem hiding this comment.
I'd mention agents explicitly here as well. Additionally, it might be useful to add a short line on why for now they are not allowed to(i.e. concerns about licensing etc.).
There was a problem hiding this comment.
I'm afraid I'm rather uneducated when it comes to agents etc... do you have some suggested wording for this?
I really like this approach. It gives people more of a chance to acknowledge usage of AI in different aspects where they might initially say "no AI" for a single response. Potentially saves us some time in review. |
|
P.S. I'm going to merge in grammatical corrections, but not content just yet - that way everyone gets a chance to comment on the same thing. |
Correcting some typos found during review.
A policy for AI generated tools.
LLM / AI generated code disclosure
A likely-LLM powered tool in google docs was used for spell checking and pointing out gramatical issues.
PR Checklist
package/CHANGELOGfile updated?package/AUTHORS? (If it is not, add it!)Developers Certificate of Origin
I certify that I can submit this code contribution as described in the Developer Certificate of Origin, under the MDAnalysis LICENSE.
📚 Documentation preview 📚: https://mdanalysis--5210.org.readthedocs.build/en/5210/