a Wikipedia article that was citing an AI-generated "source"

March 7, 2024 5:12 PM

Video: Become a Wikipedian in 30 minutes (with transcript)! web3isgoinggreat's Molly White's fantastic explanation of how to get started and lots of other useful advice. White suggests that new editors to get started writing about something they know nothing about: for example, the silverspotted tiger moth. If you are a subject matter expert in a particular topic, you don't know what other people don't know and may not be able to write a good 101-level entry.

Molly White:

instead of finding an article that needs help and then trying to make an edit from there you can try to find an interesting fact and then add that to the encyclopedia. One great way of doing that is just go to your local library and see what they have in their non-fiction section that looks interesting to you. I did that this morning, and I found this book: a huge tome on garden insects which I guarantee has information in it that hasn’t been put onto Wikipedia.

posted by spamandkimchi (21 comments total)

White suggests that new editors to get started writing about something they know nothing about:

An unfortunate number of Wikipedia entries were clearly written following this advice.
posted by Dip Flash at 5:53 PM on March 7 [8 favorites]

That sounded way crankier than I meant. I think the advice is great for getting people started, but sometimes the results are kind of thin. But anything that gets people started and feeling comfortable is great.
posted by Dip Flash at 5:56 PM on March 7 [1 favorite]

The framing of this post threw me. The title and the "in 30 minutes" framing made me think that the video and related links were going to be a spammy guide to easily contributing to Wikipedia by having an LLM generate plausible bullshit for you. And then, once I realized what it was, it struck me that the two types of guides (the real and what I imagined) really would look the same on the surface. "Look how easy it is to contribute!" is both the cause of the injury and the enticement to produce the cure...
posted by whatnotever at 6:13 PM on March 7 [2 favorites]

That would be fine if was just a "people's encyclopedia" - that doesn't sound like something that requires fidelity to truth, logic and reason. It just has to be good enough.

However, people are citing Wikipedia as an authoritative source, which embiggens the pressure considerably. It's possible for it to be its own source, in some cases.

Using/allowing some bullshit AI-written article is a huge dickslap in the face to the truth. It's a pity you can't criminally prosecute those involved in that sort of shenanigans.
posted by JustSayNoDawg at 7:44 PM on March 7 [1 favorite]

Like many others (I assume) I haven't looked at the "official" encyclopedias in a while. But then again I graduated over a decade ago. Are Encyclopedia Britannica's online articles not a valid information source? I guess it just might not be as suitable for things like "list of all 100 episodes in this TV show from 2005" or other niche interests.
posted by picklenickle at 8:20 PM on March 7

An unfortunate number of Wikipedia entries were clearly written following this advice.

I don't write entire articles, but the majority of my meager contributions to WP are started when disappearing down a WP rabbit hole and coming across an arcane article with some combination of spelling/grammar mistakes, commercial links, and just plain bad writing. I make my edits, sometimes adding something hopefully motivating to its Talk page to improve something specific that I know nothing about, and going on my way. Pretty sure I haven't ever been reverted.
posted by rhizome at 9:15 PM on March 7 [3 favorites]

I guess it just might not be as suitable for things like "list of all 100 episodes in this TV show from 2005" or other niche interests.

I just checked, Encyclopedia Britannica doesn't even have an article on the 1998 Hell in a Cell match between Undertaker and Mankind, where Wikipedia has 8,000 words.
posted by Literaryhero at 9:26 PM on March 7 [7 favorites]

Reading mathematics pages on Wikipedia is illuminating. They almost all seem to be written by post docs who are domain specialists. Which makes them basically useless.
posted by constraint at 11:19 PM on March 7 [14 favorites]

Coincidentally (I haven't watched Molly's video yet) I created my first Wikipedia page just the other day - after decades of just making minor edits.

I tried to follow the various onboarding guides on Wikipedia itself, and the experience was genuinely enlightening - a lot of people have clearly invested a huge amount of work and making that flow clear, while also strongly supporting Wikipedia's editorial values.

Finding citations to support everything I wanted to include was a ton of work!

Here's my page, for the Bach Dancing & Dynamite Society jazz venue in Half Moon Bay, California.
posted by simonw at 1:24 AM on March 8 [7 favorites]

Used to hand-crank HTML 2.0 pages (using a lorra blink tags) ~30 years ago but I've been inhibited from doing anything more than piffle about the edges of Wikipedia because of the [[wikipeculiar]] tags in the source editor. TIL we can use a wysiwyg VisualEditor. So thanks, that was handy!
posted by BobTheScientist at 2:36 AM on March 8

Reading mathematics pages on Wikipedia is illuminating. They almost all seem to be written by post docs who are domain specialists. Which makes them basically useless.

This is how I feel about the eternalism article, which I often end up on because it’s commonly used in sci-fi as anything from a trapping (Warframe) to an underlying structure (Story of Your Life) and I still cannot get my head around it.
posted by brook horse at 4:52 AM on March 8

people are citing Wikipedia as an authoritative source, which embiggens the pressure considerably. It's possible for it to be its own source, in some cases.

I've cited them before in news articles, but only after reviewing the underlying sources. For example, there are lists of previous rocket launches that are accurate and confirmable through other sources, and it is convenient to cite Wikipedia because the author has done a good job condensing the information.

There are other times I would never cite them, in particular, anything historical that is open to interpretation. I don't trust them for that at all.
posted by wolpfack at 5:19 AM on March 8 [2 favorites]

AI-generated text is spam that can pass a Turing test.
posted by AlSweigart at 6:25 AM on March 8 [1 favorite]

Are Encyclopedia Britannica's online articles not a valid information source?

I wouldn't know, the article for Encyclopædia Britannica on the britannica.com site is locked behind a paywall.

Encyclopædia Britannica, Inc. is a private company. How long do you think it'll be before they cut costs by using AI-generated text themselves? Where is the View History page for their articles so we could see what edits have been made and when?

"They wouldn't make such a short-sighted mistake" is naivety from a simpler time, up there with Google's "Don't be evil."
posted by AlSweigart at 6:36 AM on March 8 [2 favorites]

Here's my advice for actively improving Wikipedia articles:
1. Think of someone you know or someone your friends know who has a Wikipedia entry.
2. Make contact and ask three questions:
-- What is incorrect or missing on your Wikipedia page right now?
-- What have you done since the last time someone updated your page?
-- Where can I find webpages that independently verify that information?
3. Update their page.
posted by Hogshead at 8:20 AM on March 8 [2 favorites]

That's more or less what I do! It can be really useful because a lot of people who might be well known enough to be on Wikipedia nonetheless don't know much about Wikipedia and so feel it's hopeless to get anything changed. I find that starting with biographies of not-well-known people where you can prove notability and following a pretty basic template can also be a great way to start. There are so many little ways to contribute, like maybe article writing isn't for you but have you heard the good news about the typo team?
posted by jessamyn at 8:26 AM on March 8 [1 favorite]

brook horse: completely tangential to TFA, but i don't think eternalism is fairly described as a "trapping" in Warframe -- it's pretty deeply woven into the worldbuilding!
posted by adrienneleigh at 10:59 AM on March 8

have you heard the good news about the typo team?

It me! I may have to join a club that will have me as a member.
posted by rhizome at 3:00 PM on March 8 [2 favorites]

I noticed that a family member’s college had like 12 “notable faculty” on its entry and I tried writing an entry in the identical style for a different professor of some mild notability as an author. I was told that he is probably not notable enough, for reasons that I don’t really understand. The experience was weird and unpleasant.
posted by sdrawkcab at 5:52 PM on March 8 [2 favorites]

Notability is a disastrous failed experiment that bears substantial responsibility for the project's dying-small-town vibe today. Unfortunately, due to structural flaws in Wikipedia governance (which I would summarize as low barriers to exit and high barriers to voice making course-correction nearly impossible), the prospect of rolling that failed experiment back is extremely remote.

It's a shame there hasn't been a serious effort at a friendly fork in many ages.
posted by Not A Thing at 6:12 PM on March 8 [1 favorite]

brook horse: completely tangential to TFA, but i don't think eternalism is fairly described as a "trapping" in Warframe -- it's pretty deeply woven into the worldbuilding!

I fully believe and accept your authority on this. I am repeatedly on the eternalism Wikipedia page because I don’t understand the Warframe lore. I am bad at video games and miss approximately 96% of the story in transmissions because i am trying not to die ✌️
posted by brook horse at 12:37 PM on March 10 [1 favorite]