The Book You Liked Before Friday

Suppose you bent down a page. Something in a novel caught you, and you wanted to find it again. I am writing to you, the reader who liked the thing before everybody needed an opinion about whether liking it had been a terrible mistake.

On Friday, the Académie Goncourt removed Thélyson Orélien’s C’était ça ou mourir from its prize longlist. The Haitian-Canadian writer denies using AI. An accusation that the novel was almost entirely machine-written had circulated through an anonymous account using a detection tool called Pangram. Other detection programs reportedly reached the opposite conclusion. Separate plagiarism allegations concerning earlier work also entered the affair.

I haven’t read the novel. I cannot tell you whether it is good, who composed each sentence, or what evidence will finally establish. You may already know more about the book than I do. That used to count for something in a discussion of literature.

The academy cited both the plagiarism accusations and analyses it considered trustworthy in explaining its decision. Its purpose, it said, was to protect the prize’s integrity and honor literature written by women and men. I understand the purpose. If a prize is for human writing, the people awarding it have to care how the writing got there.

But I want to stay with you a minute, before the argument puts on its good trousers and leaves for Paris.

Perhaps you recommended the book to someone. Perhaps you were pleased to have found it before that person did. I know this small, disagreeable pleasure. I have had it about writers, songs, and bars that were never improved by my presence. Now someone tells you the work may be counterfeit, and along with the disappointment comes the fear that you have been made to look stupid.

I would want to get that feeling over with. Waiting to find out what happened would be considerably less pleasant.

You are allowed to be angry if somebody sold you machine-made prose under a false account of its authorship. I would be. The fact that a sentence moved me would not settle whether I had been deceived. Knowing who made something, and under what conditions, can change what I think of it. I am not going to tell you that only the words matter, as though a book arrived through the window without a name or a price.

But the anger needs a fact to stand on. Otherwise we are simply passing the embarrassment along until it reaches somebody who cannot pass it any farther.

Orélien says the allegation is racially motivated. He has defended his repetitions and imagery as part of a Haitian and Caribbean language tradition. I cannot establish the motives of the people accusing him. I can hear the ugly possibility in requiring a writer to make his voice recognizable as human to software before the rest of us will listen to it.

You might dislike his repetitions. Good. Readers have been disliking writers for centuries without laboratory assistance. Tell me where the language wore you out. Tell me why the passage that worked the first time became tiresome the fourth. That would be an argument about the writing, and I could learn something from you even while thinking you were wrong.

A detector’s verdict is a different claim. It says something about the making of the work. When programs disagree, I am not entitled to keep only the result I wanted and proceed as though the question has been settled.

And no, this does not get the plagiarism allegations out of the room.

Those concern earlier work and deserve examination on their own terms. If a passage was taken from another writer, there are particular words, dates, and sources to compare. I want that comparison. I don’t want evidence about one act quietly promoted into proof of a different act because we have become impatient with the number of things there are to know.

The original Canadian publisher, Boréal, says it cannot confirm or deny the allegations at this stage. It is concerned about the plagiarism claims and has condemned the personal attacks. This is not an acquittal. Nor does it give the rest of us a license to supply the missing findings with confidence and a bad temper.

You might reasonably ask why I am being so careful. I am rarely careful about anything that involves a literary prize. Usually I regard the ceremony as a room full of people trying not to look at who got the better chair.

Because I can see how much I would enjoy the verdict if I already disliked the book.

I could stop explaining my dislike. No need to show where the prose went slack or the character stopped behaving like a person. I could announce that the machine had detected another machine, and my bad mood would become expertise. This would be a great convenience to me. I would prefer not to have it.

It would be worse for the next writer. I can imagine somebody trimming repetitions, flattening a peculiar rhythm, taking out a sentence that feels right because it might look wrong to a detector. Not because that has been shown to fool the software. Because nobody wants the accusation, and frightened people will try whatever seems available.

I don’t know how many writers would do this. I know it would make a rotten set of instructions for a young one: develop your own voice, then make sure it does not attract attention from the apparatus certifying that voices are still allowed.

Meanwhile you still have the experience of reading the book. You don’t have to defend the author against every allegation to admit that you had it. You don’t have to pretend the feeling proves human authorship, either. It is something that happened to you, and I would hate to see you lie about it merely to arrive on the respectable side of an argument a little sooner.

I want fraud investigated. I want borrowed words traced to the people who wrote them. I also want a writer able to be strange, repetitive, sentimental, even lousy, without those qualities becoming a confession to using software.

If the evidence establishes deception, you can change your mind. You can regret the recommendation. You can decide never to buy another book by the man. There is plenty of room for a judgment that arrives after the work of judging.

What I cannot bear is watching you apologize for having paid attention. You gave a book some hours of your life. Now there is more to find out. That ought to mean examining the allegations, not going back through your own mind to remove the pleasure before somebody notices it.

Additional reporting: France’s prestigious Goncourt prize drops novel after author accused of using AI

Source: Novelist accused of using AI to write book removed from French prize list