When Artificial Intelligence Tells Lies
When the Dutch author Thomas Heerma van Voss read a description of his latest novel Het Archief (‘The Archive’) in Harper’s Bazaar, he did not recognize his own book. The freelance digital editor had blindly trusted ChatGPT. The fashion magazine’s reaction to the mistake consisted of glib one-liners. Does anyone still have the slightest appreciation of truth?
A friend texted me that she saw my name on the list of the best five Dutch books in Harper’s Bazaar. Ah, the fashion magazine? I quickly grabbed my laptop, searched and scrolled, and indeed, there was my novel The Archive. So far, so good – until I saw what was said about my book.
“The Archive follows a young man who, after his father’s death, discovers an archive full of documents, in which he learns more about his father’s life. As he rummages through the papers, he gets closer and closer to unravelling secret aspects of the past.”
I have regularly come across sloppy summaries over the years, but this had little to do with sloppiness. None of this text seemed to apply to my book. The father of my protagonist is still alive throughout the novel. There is no archive full of documents through which the narrator rummages. There are no secret aspects of the past.
I had a look at the rest of the list. About the biography of film director Theo van Gogh, “without a doubt one of the best Dutch-language books of 2024”: “De Bolle Gogh (‘The Chubby Gogh’) tells the story of Joris, a young painter who struggles not only with his art but his weigh as well. He is a bit ‘chubby’ and that doesn’t always make his search for success and love easy.”
Hey? What? I began to suspect how this piece – the work of a freelance digital editor who had already written hundreds of contributions for Harper’s Bazaar – had come about. Especially when I saw that Chantal Akermans’ translation from the French Ma mère rit (‘My Mother Smiles’) (“life with a mother who always kept smiling no matter what”) was also labelled as one of the best Dutch books of the year.
Michal Citroen’s extensive history of the hiding of Dutch Jews in wartime now turned out to be a story about a woman who undertakes “an intriguing search for her roots”. Marijn Kruk’s journalistic research into the rise of the extreme right in the West was summarized as follows: “Opstand (‘Rebellion’) takes place in a future where an oppressive regime has everything in its grip. In this best Dutch book, we follow a young woman who resists oppression and goes in search of her own place (…) When she joins a rebellion, she faces difficult choices.” Conclusion about the book: “As soon as you pick up the book, you can’t put it down.”
I have regularly come across sloppy summaries over the years, but this had little to do with sloppiness
I contacted the Harper’s Bazaar editors. I shared my objections and my strong suspicion that these texts were written by artificial intelligence. My request: whether this could be taken offline as soon as possible and then rectified. My question: what on earth had happened here?
Even before I received a response, the best five Dutch books of 2024 had disappeared. Oops! The page you are looking for does not exist.
Then someone whose job description was managing editor digital reported in my mailbox. She was quite pleased that I had pointed this out to Harper’s Bazaar, which describes itself as ‘the oldest and most iconic fashion magazine in the world’. “Apologies for the incorrect description of your book and the other books.” No explanation, no shame. As if I had drawn their attention to some incorrect detail.
Had this managing editor read my entire mail? Why didn’t she respond to my comment about artificial intelligence?
Her message ended with the announcement that I would soon receive an adapted text about my book, and if I could please vet it? In other words: I was allowed to do unpaid final editing work so that they could cover up their own botch job.
I was allowed to do unpaid final editing work so that they could cover up their own botch job
The tricky thing is that the use of ChatGPT can rarely really be proven; different texts are always generated, and any writer involved can simply continue to deny it. Yet my doubts continued to grow. Especially when I asked ChatGPT what my book and those of Marijn Kruk and Jaap Cohen were about.
Texts appeared on my screen that were almost identical to those in Harper’s Bazaar’s top five. This time, I had written a novel that “focuses on a young man’s search for the truth about a tragic incident from the past.” Again, there is talk of family secrets and rummaging through the papers of a father who has already died.
I emailed Harper’s Bazaar again. This time I also messaged the freelancer on duty. That freelancer also responded that she was quite pleased that I came forward. “There had indeed been some errors regarding content, so we temporarily took it offline to correct it.” What errors regarding content? In this annual review, not a single sentence had made any sense. Didn’t she put pride in her work, was this answer also written by artificial intelligence?
© Joris Casaer
I sent her another email, and repeated my questions: how did she conclude that The Chubby Gogh was about a young, somewhat chubby painter named Joris? To which young woman and what future was she referring in the book Rebellion? No reply. Not even after my subsequent barrage of reminder emails. I had sunk my teeth into it.
The managing editor did send a reply, although she did not respond to anything substantive. She kept saying that I could expect an adapted text about my book and whether I wanted to review it, firing further glib one-liners at me. “We have ascertained that one of our editors has been careless in her research for this article. We deeply regret that. It goes without saying that the accuracy of our articles is of paramount importance to us. We will resolve the matter internally with our editors.”
What research? And ‘careless’, really, was that the adjective for such a compilation of fabrications? What does that ‘address the matter internally’ look like, how can such a text slip through despite your lofty words? And what about the rectification?
None of my questions were answered. What I did receive were: the repeated assurance that at Harper’s Bazaar the accuracy of articles is ‘of paramount importance’, the suggestion that they have already done me a great service by taking their own nonsense offline, and finally the request that I ‘respect’ how they will resolve this matter ‘internally’. “To clarify, as in the previous email: the research was carried out carelessly, I will check the errors with the editor and ask again to please respect this.”
Another message: “Rest assured that we are taking this case seriously, but unfortunately, mistakes are also made in human work.”
It is problematic when a major fashion magazine feels called upon to publish a top list of books, while no one even bothers to read the back cover texts of the works in question
The point is, of course, this can hardly be called human work. That someone summarizes books carelessly, that a freelancer cannot completely read all the books on the list is understandable. But it is strange, no, it is downright problematic when a major fashion magazine feels called upon to publish a top list of books, while no one even bothers to read the back cover texts of the works in question.
It wasn’t until a week later that the managing editor of Harper’s Bazaar – a magazine with a circulation of 25,000 in the Netherlands and on its website in its own words with ‘Guaranteed engagement’ – sent the announced rectification. At least that’s what she herself called it, because what she sent was not a rectification at all: it was the same top five, this time with coherent summaries. Not a shred of anything about the previous version of the piece, a message of any errors, no excuse, as if it had never happened.
Meanwhile, I see online that the journalist involved has written countless new pieces for Harper’s Bazaar. No trace of any rectification. The slogan on the site: for those who dare to dream.
If this gives you pause to think: what’s the big deal about some inaccurate little list of books in a fashion magazine? The answer is that the whole incident is an extremely big bad deal.
A great deal of outrage was rightly expressed about Mark Zuckerberg’s measure at the beginning of this year to no longer use fact-checkers for his company Meta. Facebook and Instagram will thus follow the open sewer X: the apps will become a receptacle for the most idiotic and hurtful disinformation.
Newspapers and magazines hereby play a crucial role in such cases. And that role comes down to everything that Harper’s Bazaar failed to do.
Not being held accountable. Not making any excuses. Hiding behind empty, impersonal language, and where possible blaming the person pointing something out to you for a lack of ‘respect’. Only making changes when you are caught, and even then, quietly.
These are all steps that further erode the truth, a process that is already having terrible consequences worldwide. And in that way, people keep stumbling on, towards their demise.
Leave a Reply
You must be logged in to post a comment.