Wikipedia:The LLM-written ANI report
Don't use an LLM to write an ANI report
From Wikipedia, the free encyclopedia
 | This is an essay. It contains the advice or opinions of one or more Wikipedia contributors. This page is not an encyclopedia article or a Wikipedia policy, as it has not been reviewed by the community. |
What is the LLM-written ANI report?
Administrators and other regulars at WP:ANI have become familiar with a recurring pattern of complaints. It looks like this (and may have emojis starting each point which may/may not actually apply to the issue):
I am requesting administrative review of [editor] due to their pattern of X, Y and Z at [topic area]. Here is another sentence briefly giving context about the dispute. Maybe there's a third sentence.
1. The first complaint
"Quoted text from an edit"
2. The second complaint
According to WP:POLICY, editors should never Thing B. WP:GUIDELINE says that article content should not Thing D. On Wikipedia, [wishful conjecture that does not follow from the prior two statements].
This is a serious violation of WP:BLP
3. The third complaint
- Possible hounding across multiple articles WP:HOUND
- Reverts edits and tells people to discuss on the talk page without engaging in discussion WP:DISCUSS
The complaint without the number
On articles such as [subject1] and [subject2], [editor] has:
5. We remembered the number again
During my discussion with [editor], another user [well-known admin] collapsed my text and said I was using an LLM. This made it difficult to continue discussion. [Well-known admin] contributed to a hostile environment. WP:AGF, WP:CIVIL
Requested action
I am requesting administrative review of [editor] and [well-known admin]'s behavior to establish:
- whether this constitutes disruptive behavior or tendentious editing
- whether to issue a topic ban or warning
- how to ensure that editors can contribute constructively going forward
–Preceding unsigned comment added by OPUSERNAME
There's small variations. Sometimes there will be more WP:LLMSIGNS like stray markup or incorrect line breaks. Sometimes the signature will be included. The exact content and number of sections can vary. Sometimes there will be diffs, sometimes there won't (likely a function of how savvy the user is to insist that the LLM use provided diffs in its writeup). The tone will always be crisply professional and authoritative, even when the cited policies and guidelines are not. The reports look like this and this and this and this. It's not clear whether it's one LLM in particular that produces requests like this, or a subset of them, or all of them, nor is it clear where the LLMs got the idea that this is what an ANI report looks like. It's not. Normal ANI reports look like this and this and this and this.
Why these reports are a problem
All of the general considerations regarding LLM use in discussions on Wikipedia apply to ANI. But beyond that, in producing these reports, LLMs are throwing together dozens of superficially plausible allegations at once, which requires a lot of effort for the community and the accused editor to respond to. While some attempts to use LLMs at ANI are sincere, it is also very easy to turn these tools into a vector for harassment and abuse. It is not fair or reasonable to ask editors who volunteer their time on Wikipedia to put up with the prospect of such harassment. If you have a genuine concern about someone else's activity on Wikipedia, follow the rules and draft a complaint yourself. If you need help, ask other editors, not the AI.
Notes
- ^ These discussions chosen by twirling the mouse wheel at random on the pages already opened to the previously cited LLM cases, apologies to any parties in these cases.