Authorship and the AI Workflow
When the human in the loop blurs into the machine
Edu-consultant Adam Pryor recently wrote in a Second Draft Labs column “‘Can you tell what you wrote and what the AI wrote?’ “I want to answer that directly: no.” Pryor wrestles with explaining the concept of the AI-human hybrid workflow while dragging 25 centuries of publishing behind him. Framing and classifying what might be a new normal is daunting.
Treated wrongly, it’s a matter of bait and switch. I have no problem with writers who use such a workflow. But it completely reframes the concept of authorship and moves it in the direction of branding. When I commit my time and attention to an author—as opposed to a publisher, such as the New York Times, I expect that not only the facts and concepts presented are the work of the human whose name is at the top, but also the literary choices made in the prose.
As a writer myself, I understand the broader editorial workflow. I accept that editors, fact checkers, and sometimes lawyers, and spouses, contribute to the finished product. But there is an unspoken convenant between the reader and the author that the work, in the generally accepted meaning, belongs to the person behind the byline. Not that they merely “stand behind” the work, but that it is the product of their own mind. Authorship smeared into generative AI can trend toward plagiarism.
I suggest that we not force readers to accept a “new authorship,” but to modify credit. If a work is from hybrid sources, the human in the loop should relinquish the mantle as a conventional author and perhaps present themselves as something akin to an orchestra conductor or project manager, or even a contributing editor. The involvement of an AI entity should be explicitly mentioned,
This is a role in search of a title. Suggestions?
