The disruptive force of artificial intelligence is engulfing the world of book publishing.
In March, publisher Hachette Book Group withdrew publication of an upcoming horror novel after online detectives accused the author of using generative AI to write parts of the book. The author denied the accusations and instead blamed an acquaintance who edited the book. Last week, author Steven Rosenbaum faced backlash after readers discovered that some quotes and source material in his book, True Futures: How AI Will Reshape Reality, appeared to have been hallucinated by chatbots or fabricated. It includes a quote from Lisa Feldman Barrett, a prominent Northeastern university professor.
Rosenbaum responded to the allegations by stressing that he “used ChatGPT and Claude during the research, writing, and editing process,” according to a statement to the New York Times. He took “full responsibility” for the error and said he was working to correct it in future editions, the newspaper said.
This crisis has also permeated the much diluted world of literary awards culture. This month’s literary magazine is Granta has found itself at the center of controversy after awarding the annual Commonwealth Short Story Prize to a story believed to have been generated with the help of AI.
Some argue that AI is actually leading to the Napster moment in the late ’90s, when the music file-sharing software upended the economic and administrative structure of the music industry by allowing millions of users to download and distribute copyrighted songs online. This episode showcased the disruptive power of technological innovation on the creative industries.
Generative AI is now forcing a long-overdue reckoning in book publishing about what counts as original human work and how technology should be ethically introduced or disclosed at every stage of the book production pipeline.
Vance Ricks, professor of philosophy and computer science at Northeastern University, said the intrusion of AI into book publishing could impact the importance we place on reading and writing. Large-scale language models show that these tools can write, or at least produce, text that has “nothing to do with anything experiential or sensory,” he said.
“Then you have to ask, what does it mean for a human to write something, and how does that differ from what it means for a machine to write it?” he said.
As a long-standing habit, readers assume that the authors of published works are responsible for their words and ideas, Ricks said. That’s why there have been cases of outrage when internet sleuths make discoveries or when writers reluctantly reveal secret uses of AI in their creative processes, he said.
This deception is all based on expectations of common practice, Ricks said.
For example, experts point to the fact that the industry has long battled shrinking editorial resources, coupled with what a recent New York Magazine report describes as a largely honor-based approach to fact-checking and verification, making book publishers particularly vulnerable to the risks posed by generative AI.
Experts say that without a set of best practices, recent events touch on some of these concerns, especially as it pertains to the integrity of authors and the ethical expectations and obligations that authors and their editors undertake.
“Everyone is scrambling to figure it out,” says Stephanie Young, associate professor of English and W. M. Keck Foundation Professor of Creative Writing. “Publishing is very much in decline at the actual editorial level,” she noted.
Young, whose work focuses more on the “prize economy” of literature than mainstream publishing, suggested that the crisis has exposed deeper structural pressures on publishers, with much of the editorial burden tending to fall on overworked literary agents, who can undergo multiple rounds of substantive editing and drafting before a book reaches an editor.
Young, who believes that AI tools can serve important and educational purposes, said he is interested in how the debate surrounding the use of AI in publishing speaks to persistent “illusions” about ideas of authenticity, expressiveness, and originality in writing.
“I think we have this idea of creative writing centered around a very distinctive, authentic voice, but really, these are all genres,” Young said, “and as humans, we’re influenced by the genres that are in front of us, and we write within those genres. So it kind of makes sense, in the same way that award-winning short stories are very identifiable genres, right?”
She pointed out that there are traditions that reject authorial control entirely, such as the “automatic writing” of surrealism, so-called “found object” art, and other conceptual practices dating back to the early 20th century that explore appropriation, remixing, and chance.
As we zoom out into the larger world of mainstream publishing, we tend to rely more on identifiable formulas and the lines between “real” human writing and AI-assisted writing become more blurred. Mainstream publishing has become a “genre-driven, market-driven project” and is already shaped by “standardized voices” and patterns that AI can easily imitate, she said.
But there may come a point, Rix said, when the use of AI in published work resembles the relationship between celebrities and the communications teams behind them.
“I imagine that one day it might be more widely understood that, just like in speechwriting, having your name associated with something basically just means you take credit or take credit, and it doesn’t necessarily mean that you put in the work or effort to organize those thoughts and words that way,” Ricks said.
But until that day comes, Ricks said authors should be able to identify and prove what tools they used in the creative process, whether it was other humans, AI, or both. Similarly, Mr Young suggested that a more robust conversation is needed about how.
But fear of a backlash may be holding them back. “I don’t think it’s a good thing if people around you think they shouldn’t tell you that they’re using an AI system or how they’re using an AI system, but they’re doing it anyway,” Ricks says. “You can have a preference for people not to use them, or only in a very limited way, but you can also try to chip away at the feeling that it’s the most serious moral sin a person can commit.”
One way publishers can better detect AI in authors’ writing is simply by building better relationships with authors, Rix said. That way, publishers can more easily detect sentences that appear to be incorrect. These publishers should also be open with readers about their AI policies, what they allow, and how they vet their use of AI, he said.
Given the economic realities for publishers, Ricks said there are also ethical questions about whether publishers will one day use AI to replace editors.
“The challenge for publishers is what kind of oversight will be provided.” [we are] What kind of instruction and instruction are you willing to pay for? [we are] I’m willing to pay for it,” he said. [publishers are] I’m just going to accept it. [they’re] You will sometimes feel embarrassed when something like this happens. ”
The use of AI in the publishing industry also raises the question, “Why do we want people to read books in the first place?” Ricks says.
Reading “is a technology. It’s a form of creative expression and a way to convey information,” he said.
To that end, “we want to uphold the value of reading and what we can learn from the written word in any policy or practice,” he said.
