Editor’s Note: Retraction of article containing fabricated quotations
- Thread starter JournalBot
- Start date
Eh, no. I and many others read the article and know it was Benji Edwards and Kyle Orland. Both are equally responsible for the content.
Upvote
6
(37
/
-31)
I'm not sure reviewer #2 being replaced by an AI agent has risen to the level of science yet.
Upvote
-10
(4
/
-14)
So my question is - did the authors themselves source these fake quotes directly from a gen AI tool? Or did they source them from a third party article which itself got them from a gen AI tool? I feel like this retraction statement could be taken either way, but it's a pretty significant difference for saying what the authors did. I agree with others that there should be a full moratorium on using any gen AI when writing articles here, period. But is that what they did, or did they just not adequately verify the source of the quotes they used?
Upvote
80
(82
/
-2)
Upvote
12
(20
/
-8)
It's damning that he has to speculate at all. The authors should have replied to him quickly and let him know how things got wrong on their end. This is a story about him, for fuck's sake. Was no attempt made to reach him for comment while the story was being written?
Upvote
182
(185
/
-3)
I am gravely concerned that this happened not on a syndicated article, not from some random freelancer, but on an article by two of the most recognizable authors on the site. This has deeply shaken my trust in Ars.
I'm willing to allow some time, but I do expect a full postmortem. I want to know:
I'm willing to allow some time, but I do expect a full postmortem. I want to know:
- how this happened
- what is being done to ensure that it never happens again
- whether or not disciplinary action is being taken, and all details that are safe to share about it
Upvote
295
(296
/
-1)
I appreciate Ars being transparent and open about this process, and for quickly owning up to the issue. But let's be clear: this sort of stuff cannot be allowed to recur. AI slop is everywhere, and free. There is no goddamn reason why I should be exposed to any of it for a service that I'm paying for. Content generated by AI shouldn't just be excluded from the final product of an article, AI should be excluded from the entire process.
Upvote
122
(125
/
-3)
Agreed. Though, the blog could have been ingested by an LLM and it would still spit out fabricated quotes. It's not like LLMs are 100% accurate even when having things put directly into their context.
I once fed ChatGPT some bulleted lists and asked it to turn into into paragraphs for dissemination. On proofreading the output, there were several instances of completely fabricated data points. I corrected the mistakes and sent it on, fully disclosing how I had used ChatGPT.
Of course, I'm not writing articles for a respected media outlet.
Upvote
47
(49
/
-2)
Except rootbeer and the Federation are both net goods... unlike LLM AI. I'll also add that I've not touched the stuff. Well, to be more accurate I've seen it pop up unbidden, then found ways to block it in every case I could, and I even toyed around with it in a few chats to see just what I could expect, but came away unimpressed. I've never used it to write anything for me, never asked it any questions I wanted answered, never even used it for proofreading. I just prefer to do my own thinking. I know this comes across a bit insulting to those who have found themselves relying on it, but I'm not really sure how else to phrase it. If nothing else, when I do "unload" my thinking, I'd prefer it to be to an entity that actually is thinking and not just doing statistical analysis.
Last edited:
Upvote
45
(49
/
-4)
There are a lot of people here who have no idea what article they are referring to because they pulled it. Instead they should have left it up with a statement at the top that the content is under review and why. Locking comments seems appropriate under those circumstances.
This is literally not the first time they have just pulled an article and made a context free apology.
Upvote
142
(142
/
0)
I'd say the truth is... sigh... somewhere in between in this case. There ARE a large number of people in the U.S. who refuse to ever admit to a mistake, but I don't think it's the population at large, just a significant number.Let me tweak this for you please...from to "the population believes" to "U.S. federal and state government officials believe".
Upvote
32
(33
/
-1)
The quotes only existed on Ars when the article was released. I searched for them elsewhere. Every indication is that they were fabricated by the author or authors, apparently using AI tools.
Upvote
122
(122
/
0)
My initial reaction was also that Namingway was referring to Kyle Orland as "Senior Space Editor" which confused the heck out of me.
Only when I was about to write a reply and re-read it I realized he must have been referring to Eric Berger.
Upvote
38
(40
/
-2)
My own curiosity wants to know why the authors felt they needed to use AI anyways. It's not like this was some breaking story that had to be published on Friday. Couldn't it have waited until Tuesday if it meant they didn't use AI? Or did they just rush to publish it so they could take off for the long holiday weekend?
Upvote
95
(95
/
0)
The "One (or two)" part that you quoted makes it highly relevant "how" it happened. Were both authors involved in the article and jointly wrote/edited the whole thing? Did one start and hand it off to the other to finish? Who introduced the hallucinated quotes? Also, was any other party involved in editing the article, and did they have a responsibility to check the quotes or sources? I don't know Ars' editing process.
Those things matter a lot for how Ars proceeds, and I think it is more than reasonable that Ars should try to get that part right. I do expect to hear more, once they have done so, but it's more important that they get it right than that they give me the full story on a Sunday of a holiday weekend just two days after the issue occurred.
Upvote
92
(92
/
0)
Occam's razor says that this is probably the case. But given that this whole topic is a miasma of AI generated bullshit, it's not beyond the realm of possibility that when writing this article, the authors found an article with fake AI-generated quotes, used them, and said article has since been deleted.
Better evidence is the fact that the fake quotes in the retracted article contained a "Shambaugh wrote", with a link to Shambaugh's own blog post which didn't contain those quotes. It's hard to imagine they'd get those quotes from a third-party article but link back to Shambaugh's post.
Upvote
42
(43
/
-1)
I wonder myself if the current owners of Ars have been putting pressure to "integrate AI" into the workflow around the place, like oh too many companies have been.
Upvote
106
(112
/
-6)
@Ken Fisher , I know that employee discipline processes take time, so I personally will be waiting two weeks to see the consequences of at least one of your writers lying to you, deceiving readers, and permanently damaging Ars' credibility. if everything is business as usual, with no public consequences, I will be cancelling my subscription on March 1st. I'm not going to pay for fabrications and lies
Upvote
97
(117
/
-20)
A plausible, if dumb, explanation is that since the article is a collaboration between two authors, that both of them thought the section in question came from the other author, not recognizing it had been generated, or that the other had not checked it.
A clear process failure to be sure. I would suggest that Ars make it explicit AI policy that no front page copy is to be drafted, edited, or assisted by AI tools of any kind.
Because it's easy to say "oh, I just got a summary of this, this isn't AI content". But when that summary is wrong, we get things like the penguin tariffs and the article at issue here.
Ultimately, a post mortem of the incident from whoever is responsible for AI reportage at Ars following this, or from a senior staffer such as Ken or Lee feels necessary.
I would prefer a Mea Culpa from the authors, with a clear outline of what their workflow was and how it will be changed to address the issue, since Benji seems like an alright chap, and I rather like Kyle.
But, as Jim Salter pointed out, this is one of the worst breaches of journalistic integrity you could have, just shy of intentional misinformation and hit pieces. So something has to change to prevent fabricated quotes from being published again.
Upvote
51
(60
/
-9)
On the moment when you see these kind of AI-written articles appear on ARS, you have to admit that the dead internet theory is real.
Upvote
108
(109
/
-1)
I don't work in journalism, so take my comment with however much salt you feel is appropriate. That said, I would argue that posting the retraction at the start of the article, and repeating it (or expanding on it) at the end, is the right thing to do. You don't want people reading the article, taking it in, and only right at the end finding out that it's been retracted. The retraction needs to be front and centre, not a side note (publishing an article detailing the errors with a direct link to the original) or an afterthought (posting the retraction at the end of the article).
I do, however, agree that simply memory holing the original is not the right thing to do. I can understand why, but it's not being fully transparent about what's going on. I also don't think that hiding or deleting the original comment thread is a good look. Lock it, sure, with an explanation; but tossing it into the memory hole with the original article smells like trying to sweep the whole thing under the carpet. It's... not a great approach, IMO.
Upvote
157
(157
/
0)
This remains nonsensical and is not how AIs work. It's not "deciding" anything.
No, fabricating quotes is literally what got Jayson Blair turbofired from the NYT, in addition to the plagiarism.
There's a reason that Berger's SpaceX posts regularly generate 500+ comments.
Upvote
102
(102
/
0)
I've been a Pro subscriber for quite a while now, and I want to preface what I say below by stating I'm not threating my subscription over this. Journalism is extremely undervalued and it would be reactionary to dismiss the whole outfit over one or two writers.
I also don't expect Ars to fire writers on such short notice. I'm mad, but it'd be irresponsible to fire staff without at least investigating what happened. I don't want a head on a pike, I want to know how this happened and what Ars will do going forward to prevent it happening again.
That being said, I expect better from Ars' writers. If I found out one of Beth Mole's medical nightmare stories didn't actually happen, or some component of it was fabricated, I don't think I could ever enjoy their pieces again.
Unfortunately, until that happens I cannot trust these two journalists. I'm not interested in reading potential misinformation on hot topics. If I wanted that I'd still be using twitter. So, for their sake, please publish a follow-up to this.
I also don't expect Ars to fire writers on such short notice. I'm mad, but it'd be irresponsible to fire staff without at least investigating what happened. I don't want a head on a pike, I want to know how this happened and what Ars will do going forward to prevent it happening again.
That being said, I expect better from Ars' writers. If I found out one of Beth Mole's medical nightmare stories didn't actually happen, or some component of it was fabricated, I don't think I could ever enjoy their pieces again.
Unfortunately, until that happens I cannot trust these two journalists. I'm not interested in reading potential misinformation on hot topics. If I wanted that I'd still be using twitter. So, for their sake, please publish a follow-up to this.
Upvote
208
(211
/
-3)
Hey, uh, I'm pretty sure I found the retraction through my Mastodon feed, and boosted it. But now that post is gone. Did you actually delete the mastodon post+thread for YOUR OWN RETRACTION?
This is feeling much less transparent than I wanted to believe. Thank the gods for RSS.
E: probably not deleted, just never posted there, which. Ok.
This is feeling much less transparent than I wanted to believe. Thank the gods for RSS.
E: probably not deleted, just never posted there, which. Ok.
Last edited:
Upvote
27
(34
/
-7)
While both authors and the editor who reviewed and approved it for publication have responsibility, that doesn't mean that both authors equally contributed to how things have gone off the rails.
My own guess is, that this is still being investigated by Ars and that result of that investigation will be a factor in the followup and any possible actions taken against either author.
It is quite possible that only one of the authors did wrong, and the other may have read the blog, not remembered exactly the text and assumed the quotes in text contributed by the other were correct given his general memory of what he read. And trusted his coworker did not need to be checked on this.
If so, that could impact measures taken against each of them differently than if it turns out both were knowingly in violation of the policy on the use of AI, and failing to check work provided by an AI.
Upvote
106
(106
/
0)
This seems to me to be an even and fair take on the issue. I've already submitted my suggestion for how to prevent this (ban the use of AI in researching, writing, and proofreading articles, except in-so-far as it must be done to research content made by AI itself.
Upvote
57
(57
/
0)
Those of us who are already shaking our heads at the access journalism that goes on here at times agree with you.
Upvote
53
(56
/
-3)
This isn't a great retraction. It reads more like a standard business 'we got caught, here's a vague blanket apology that clarifies nothing' (I'm not saying ARS as a whole is complicit, just that it's a limp apology).
In the academic world, along with, I would say standard journalistic integrity, you'd expect the retraction to include what article is being retracted would be mentioned, along with the quotes, a clarification of what the misquoted person did say, and a link to the original article so anyone who read it knows what was true or false.
ARS could definitely do better than this.
In the academic world, along with, I would say standard journalistic integrity, you'd expect the retraction to include what article is being retracted would be mentioned, along with the quotes, a clarification of what the misquoted person did say, and a link to the original article so anyone who read it knows what was true or false.
ARS could definitely do better than this.
Upvote
132
(133
/
-1)
- Status