Stop Plagiarizing Claude
Some (human generated) thoughts on how to credit Claude for its contributions to your writing.
Prompt: A photo of a pile of dusty law reviews.
In the cutthroat, but very small, world of law review publishing, few crimes are more outrageous than plagiarism, so the news that a lawyer is publishing an article that was secretly co-written by Claude in a reputable, online law journal has brought with it a torrent of criticism. (It has also called into question the authenticity of the author’s other, recent law review articles, including one published in the prestigious Harvard Journal of Law and Public Policy.)
As someone currently trying to get a law review article placed in the best journal possible, I am intimately aware of how difficult and stressful it is. You spend months writing something fantastic, weeks editing it, days perfecting the citations and bluebooking and, most importantly, hours dreaming up a pithy pun worthy of the title (for the uninitiated, there is an entire universe of law review article puns based on a single Supreme Court case, Katz v. United States). To find out you were beaten by a law review article written by Claude feels like cheating. But is it?
Personally, I don’t use AI in my writing because it substitutes my voice with someone else’s voice. Is Claude’s writing frequently better than mine? Sure. But like watching two computers play chess, what’s the point of reading something ghostwritten by someone else? Are you here because you want to hear what I, a lawyer and author of a book and numerous law review articles, has to say about AI plagiarism? Or do you want to hear what Claude thinks? Both me and Claude have views, but if you want to know what Claude thinks, you can ask it. If you want to know what I think, you can keep reading.
The question of attribution for law review articles co-written by Claude might remind readers of the famous Monkey-Selfie Case, where a selfie snapped by a black macaque monkey named Naruto led to a years’ long copyright lawsuit. Since then, the US Supreme Court has held that an AI also cannot hold copyright.
But there’s a big difference between copyright and plagiarism. Plagiarism isn’t about the rights of the author to make money, it’s about the integrity of the field. If law scholarship is a collective endeavour, one that is integral to society, than the members of that field must know where cutting edge ideas are coming from. They must know how to attribute ideas to the correct author. If I don’t know who wrote an article, I can’t put that article into context or ask the author questions about it.
So by all means, use Claude as a research assistant, bounce ideas off it, and even have Claude help with the bluebooking. But don’t plagiarize Claude when you write a law review article. Not because you’re stealing from Claude, but because you’re hurting our wonderful, collective endeavour. And what lawyer would ever want to hurt legal scholarship, a field that has given us such gems as, “The Common Law Origins of the Infield Fly Rule,” and “J. R. Poisson v. Etienne d’Avril”?
Like pornography, plagiarism can be difficult to define, but we often know it when we see it, whether it’s stealing a student’s ideas, or conveniently failing to mention that our latest article was ghostwritten by a certain Large Language Model with a penchant for em dashes, pithy takes and flattering even our silliest ideas.
So, how to credit Claude? Here are some thoughts about what the system should be (and no, I didn’t ask Claude for help making this list):
Claude as first author - If you asked Claude for some good ideas for a law review article, then had it write and format the article for you, don’t publish it. Claude can’t add anything to the field of law scholarship because, in a way, Claude is the field of law scholarship.
Claude as co-author - If you had the idea and asked Claude to contribute to the research, writing or (substantive) editing process, credit Claude as a co-author, then specify how it contributed (structuring a research plan, conducting the lit review, writing paragraphs, analyzing data, etc.)
Claude as informed and helpful colleague - Did you bounce your ideas off Claude, like you would a friendly colleague in a bar, getting its thoughts and feedback? Then thank Claude in a footnote, just like you (hopefully) would a colleague.
Did Claude just help with a web or database search, formatting pages or bluebooking? Then there’s no need to credit Claude, though it would be more transparent if you did.