Is uncited LLM usage considered a CoC violation under the Inauthentic Usage policy?

Question

In the new CoC, there's an Inauthentic Usage policy that prohibits "Plagiarizing or copying content from websites, books, or other online and offline tools without proper attribution in a manner that violates our referencing standards."

One way to interpret that is that uncited LLM usage is a violation of that, as it would not comply with the referencing standards and likely falls under "online and offline tools".

If that interpretation is accurate, anyone posting uncited LLM output (including in election questionnaires, Discussions, etc.) would be in violation of the CoC. Further, it would mean that (a) sites could not allow uncited LLM usage* and (b) that moderators are required to do something about uncited LLM-generated content, as per the Moderator Agreement, mods will enforce the CoC to the best of their abilities.

Is this interpretation of the Inauthentic Usage policy in the CoC correct, and if it is not correct, how should that be interpreted in regards to uncited LLM-generated content?

---

*Individual sites must follow the CoC, they can't make policies overriding it.

Answer 1

TL;DR: Yes, the current Inauthentic Usage policy prohibits posting AI-generated content without appropriate attribution.

Human-driven contribution is a critical facet of what makes Stack Exchange work. When we originally proposed the Inauthentic Usage policy, the goal was to create a reference in the Code of Conduct that clearly states our expectation that all contributions to the site, except where otherwise authorized, were created by a real person.

When a reader comes to the site with a question, they expect that the answer they see is honest about how it was written, and has been vetted by a community of human readers. They also expect that the answer is clear about its sources – and that, in general, it is primarily the author’s own creation. While we allow quoted and properly referenced material to supplement member contributions, posting GPT-generated answers without attribution goes against future readers’ expectations for what they’ll encounter when they come to the site.

It’s also worth acknowledging that most sites have prohibited posting GPT-generated answers categorically, whether or not they are appropriately attributed. If you plan to post an answer, please pay close attention to the specific guidelines of the site that you are participating on, because they may be more restrictive than the network’s Code of Conduct.

Note that there are certain occasions on which automated participation is acceptable, such as Charcoal, a community-operated spam-fighting system. Such cases, however, operate with our explicit permission.

Answer 2

While Slate wrote this before the new CoC, I'd take their answer to Ban ChatGPT network-wide, which says:

we do consider AI generated content to be "the work of others" and the requirements for referencing must be followed for all such content on the network.

(you could also take Animuson's answer to Is attribution required for machine-generated text when posting on Stack Exchange?, where they say:)

If it wasn't created by you, attribution is always required here. We do not make exceptions because the content being attributed is not copyrighted.

to say that- yes, usage of content from a Generative AI without following /help/referencing is not allowed by the Inauthentic Usage Policy, which also points to /help/referencing in the clause of interest:

Plagiarizing or copying content from websites, books, or other online and offline tools without proper attribution in a manner that violates our referencing standards.

I guess you could nit and say that them pointing to the same Help Center page about content "written by others" doesn't qualify for whether it meets the condition of "websites, books, or other online and offline tools", but it's pretty clear to me that Generative "AI" are tools, and that ChatGPT is an online tool, and I think you'd have to do some pretty incredible mental gymnastics to say otherwise.

You could additionally try to make some deeper argument about the Generative AI plagiarizing their sources, but I don't think that's necessary for the purposes of policy interpretation here.