Given H.M. Wolfe’s Daggermouth‘s meteoric success first as a self-published novel, it should come as no surprise that it debuted this week at No. 1 on the New York Times bestseller list with its publication by Scarlett Press, an imprint of Simon & Schuster. I even bought a copy of the book!
I bought the book because of the brouhaha leading up to its publication. You can read the gory details about the kerfuffle below:
To make a long story short, based on a larger pre-print, yet-to-be-peer-reviewed study of 14,000 Kindle ebooks examined to determine if they were AI-generated based on the AI detector Pangram (which has some false positives), Prof. Tuhin Chakrabarty accused H.M. Wolfe’s book being partly AI-generated (“60 percent” according to what was reported to Oremus). The Atlantic’s Will Oremus reported those accusations while admitting there is “no definitive proof that Wolfe used AI” or that Wolfe’s book was partly AI-generated.
Pangram is an AI detector based on a freemium subscription model. AI researchers seem to think Pangram is a reliable AI detector. Substack has even integrated Pangram into its platform offer readers AI detection. **As it proliferates, Pangram is becoming a kind of AI Panopticon. **
But, based on my review of social media threads, book authors and BookTok seem to detest Pangram and AI detectors more than they detest AI itself, or at least it’s a deadheat. (For a very thoughtful discussion of some of the issues with Pangram, Freddie deBoer’s article is a good place to start.)
Oremus’s article focused on only 1 book examined by Chakrabarty’s study: **H.M. Wolfe’s book. **Wolfe, through her lawyer, denied the accusation as Oremus reported: “The suggestion that I used generative AI to write Daggermouth is wholly untrue.”
Despite her denial, The Atlantic published the accusations against her anyway.
What are journalistic ethics in reporting accusations of a novel being AI-generated?
As the AI Book Witch Hunt continues, there may be** a day of reckoning for journalists and media covering accusations that books have been AI-generated. If journalists do not have “definitive proof,” is it ethical for them to insinuate a book is partly AI-generated**
in the face of the author’s categorical denial?
And assuming it is ethical to report the mere possibility it is, does that excuse the publication’s use of a click-bait subtitle, “Daggermouth could be the first best-selling novel partly written by a chatbot,” after the title, “** Is This What Comes After AI Slop?**” The problem with the subtitle is that, as Oremus concedes in the article, it’s also possible the book wasn’t written by AI at all. **Therein lies the journalistic ethical problem. **
As far as I can tell, Chakrabarty’s pre-print study doesn’t even identify any authors or books by name, instead providing raw numbers. But apparently, through an interview with Chakrabarty, as well as the underlying data and passages of Wolfe’s book that Chakrabarty shared with Oremus, Oremus was able to obtain identifying information from Chakrabarty about H.M. Wolfe’s book for the article.
Some commentators suggest we will see defamation lawsuits from authors accused of using AI to write their books. If more authors lose 6- and 7-figure book deals, there’s a good chance we will.
Related Stories: