enh(policy): add Gen AI policy to review guide #344
Conversation
|
Ok friends. I've suggested a few modifications following the comments. Please let me know your thoughts. I like where this is going!! |
HaoZeke
left a comment
HaoZeke
left a comment
There was a problem hiding this comment.
There's a lot of great discussion here, most of which I basically ended up re-treading I believe, but still adding my comments :)
|
@all-contributors please add @HaoZeke for docs, review |
|
I've put up a pull request to add @HaoZeke! 🎉 |
|
pre-commit.ci autofix |
|
@all-contributors please add @jedbrown for docs, review |
|
I've put up a pull request to add @jedbrown! 🎉 |
our-process/policies.md
Outdated
| * When you submit a package to pyOpenSci, please disclose any use of LLMs (Large Language Models) in your package’s generation by checking the appropriate boxes and describing your use of generative AI in it's development and/or maintenance. Disclosure should include what parts of your package were developed using Generative AI (LLMs). | ||
| * Please also disclose this use of Generative AI tools by adding a statement to your package's `README.md` file and in any modules where generative AI contributions have been implemented. | ||
| * We require that all aspects of your package have been reviewed carefully by a human on your maintainer team. Please ensure all text and code have been carefully checked for bias, bugs, and issues before submitting to pyOpenSci. | ||
| * Your acknowledgment of using Generative AI will not prejudice the success of your submission. However, a reviewer can and will ask you to revisit your package's content if it appears that sections have been copied and pasted from other sources without human review. |
There was a problem hiding this comment.
@jedbrown @sneakers-the-rat @hamogu @yeelauren does this new language read ok to you? The goal here is to avoid rejecting a package just because someone is honest about using generative AI. That would incentivize some people to not disclose.
I think it's important that we allow for some GenAI in moderation. Let's lean into "this content has been reviewed by a human and if we determine it isn't (see below), we have the authority to discontinue the review.
Any reviewer can also say no to a review at any time. We haven't updated the reviewer guide, but we can do that as well in a future PR. But an editor can't say "no" to the review just because someone checks the box that they used Gen AI. i hope that helps. Give me a thumbs up here if you think this is reasonable.
Or comment if you have concerns. but let's try to get this to a point where we can merge and see how it goes! we can always update our policies as the ecosystem evolves and we learn more!
|
ok @hamogu @HaoZeke @sneakers-the-rat @jedbrown @eliotwrobson @yeelauren, please review the new version of our policy. Our goal now is to merge this so we can start implementing and learning from the policy. It won't be perfect. We can always make changes to it as we learn. As you review, please leave comments that are concise and directly actionable if possible. |
lwasser
requested review from
HaoZeke,
eliotwrobson,
hamogu,
jedbrown,
sneakers-the-rat and
yeelauren
eliotwrobson
left a comment
eliotwrobson
left a comment
There was a problem hiding this comment.
Some language nitpicks but the substance of the policy looks really good. It seems the focus is on just being transparent, which IMO is the right way to go.
Co-authored-by: Eliot Robson <[email protected]>
Co-authored-by: Eliot Robson <[email protected]>
Co-authored-by: Eliot Robson <[email protected]>
Co-authored-by: Eliot Robson <[email protected]>
Co-authored-by: Eliot Robson <[email protected]>
|
All - the blog post related to this policy can be found here: pyOpenSci/pyopensci.github.io#734 I just did a significant reorg and cleanup based on review. All comments are welcome there as well. I'm thinking, given the holiday next week, that we can get these items both published in 2 weeks - the first week of December! 🚀 Thank you in advance for your thoughts on both pieces (blog and policy). I think it's important that we make sure we aren't alienating people with our language. I see so many people using these tools at sprints, and I think it will hurt some and help others. I want both groups to stick around and perhaps we can influence people in positive ways! ❇️ |
Co-authored-by: Eliot Robson <[email protected]>
7 participants
This policy was developed based on a conversation here:
#331
I think that what we should do is review this policy and also consider linking to a blog post that covers some of our broad concerns and reasons for developing such a policy.