Wikipedia talk:Wikipedia Signpost/2022-11-28/Opinion

← Back to Opinion

Discuss this story

Since The Signpost invited Ladsgroup to give us permission to publish his essay, I didn't want to modify it before publication. But it is worth mentioning that the stylometry techniques discussed, whether AI assisted or not, can be used to link real-world identity to on-Wiki identity by comparison to off-wiki writing such as blog posts or other published material. Something to think about? ☆ Bri (talk) 17:16, 28 November 2022 (UTC)

The third paragraph of the second section addresses that, I think. I'd go on to add that it's not just oppressive governments or big companies that could do that, but also basically anyone with a grudge and a target - could be Kiwifarms, 8chan, terrorists, extremist partisans, and other such undesirables (in fact, I'd say they're even more dangerous than Big Brother or Big Tech). This is definitely something everyone should be careful about, both within Wikipedia and elsewhere on the Internet. Doxxing and IRL harassment based on Wikipedia edits has already happened on Wikipedia, in relation to the 2020 Delhi riots article - a well-respected editor was doxxed and harassed by far-right "news" site OpIndia.

Stay safe, people... W. Tell DCCXLVI (^{talk to me!}/_c) 18:19, 28 November 2022 (UTC)

• I see this article as being essentially fair warning of the technology we can't stop - either Wikipedia uses it for making Wikipedia a better place, or it's used by outsiders for their own interests; or probably both. We really can't afford to ignore it. It will have different impacts on various groups, however.
  • The main group that it should have a large negative effect on is commercial editors such as undeclared paid editors. It should be easier to catch these folks when they sockpuppet. We should embrace this use of stylometry, but not overuse it.
  • People who write here under a pseudonym, but who are published elsewhere under their own name, may have undeserved troubles because of stylometry.
  • As with any tech that can be used for doxxing, women may be subject to harsher consequences. But I don't see them being a big part of the above 2 groups.
  • I don't see much problem for editors who have experimented with sockpuppeting, say for 8 edits on 4 articles 10 years ago. Matching the 2-3 accounts (which won't have much text to analyze) would be difficult. And who would gain much from matching the 2-3 user names? In any case admins generally aren't looking for these folks.
  • And there is a kicker. This and other technologies will progress - so, if you're going to try to fight it, you'll have to fight the versions that will come out, say 5 years down the road. Smallbones_(smalltalk) 04:55, 29 November 2022 (UTC)
• I wonder just how unique these writing styles will be. I suspect if you took four hundred wikipedians who write in English the two from New Zealand would come up with some shared features as would the two from Dublin and the two from Glasgow. Especially if they used the same spellchecker. ϢereSpielChequers 19:35, 13 December 2022 (UTC)