The Future of Plagiarism

Elizabeth Eisenstein, in The Printing Press as an Agent of Change, quotes a thirteenth-century Franciscan monk who observed that there are four ways to make books. One can simply copy the work of another without alteration. One can put together excerpts from several works without adding anything of one’s own. One can put together excerpts and add comments. Or one can write a work of one’s own and add excerpts from others. The monk meant “making books” literally: putting words on paper or parchment by hand. Before the printing press, the distinction between writing and publishing did not exist in any of its modern forms. To write, in a sense, was to publish. Strikingly, the monk also assumed that “making books” mostly consisted of copying. It did not occur to him that a book might be entirely original. Modern ideas of authorship emerged only after Gutenberg, and then only gradually.

But modern “authorship” itself has been anything but a straightforward concept. Ann Blair’s new book In the Scholar’s Workshop shows just how greatly, even after the advent of the printing press, scholars often depended on hidden collaborators: editors, research assistants, servants, spouses, ghost writers, and many others. We don’t have to dive into poststructuralist debates over authorship and language to ask if “authors” actually “write” the works attributed to them. Maureen Dean, who achieved her fifteen minutes of fame in Watergate as the glamorous spouse of Nixon’s White House Counsel John Dean, was once heard to boast that not only had she not written her novel Washington Wives; she hadn’t even read it.

Today, the twin revolutions of the internet and artificial intelligence are adding radical new complications to the story and doing so at warp speed. Is a piece of writing in any sense “one’s own” when AI generates it in response to a prompt? What if one gives AI notes and an outline, and then asks it to produce the prose? What if one writes a draft and then asks AI to add evidence and polish the writing? My Princeton colleague Graham Burnett has suggested that AI is rapidly developing the capacity to churn out many forms of scholarship, such as basic literature reviews, or even historical monographs, “at the push of a button.” In the future, he implies, there will be no point in humans producing work of this sort on their own, and therefore, presumably, no reason to treat that work as individual property worthy of protection.

At the same time, AI is also making copying easier to detect. At the simplest level, it can identify literal, unattributed borrowing. But it can do more as well. Suppose I wrote an article that copied no sentence verbatim from existing scholarship yet closely paraphrased earlier works while adding no genuinely new ideas of my own. Proper citations might spare the article from the charge of plagiarism. But it would hardly qualify as genuinely original scholarship—and AI can demonstrate the point almost instantly.

So how, today, should we define plagiarism? It is not in fact entirely clear. My own institution defines it as “the use of any outside source without proper acknowledgment,” which is not really very helpful. The phrase covers everything from absent-mindedly copying a single banal, anodyne sentence, to passing off an entire piece of work as one’s own. And the recent slew of headline-making cases has only made things worse by adding toxic political fog to the issue. Just a few years ago, it came to light that my Princeton colleague Kevin Kruse had copied a few passages in his dissertation and first book from the work of others. The borrowings were minor, and even the historians he copied from dismissed the infraction as insignificant, but conservative opponents of Kruse quoted Princeton’s regulations and demanded his termination. Meanwhile, after the Claudine Gay and Jason Arday cases, various progressives and liberals have twisted themselves in knots to argue that “mere overlap of sentences” is not really plagiarism. I am here quoting an essay by the ostentatiously self-exiled philosopher Jason Stanley, which called Arday the victim of a political witch-hunt.

As someone who takes writing seriously, Stanley’s phrase “mere overlap of sentences” exasperated me. Sentences—even straightforward sentences describing things like the set-up of a laboratory experiment—are not always easy things to craft. Writing sentences that are clear, precise, and engaging takes both effort and inspiration. Passing someone else’s sentences—i.e., someone else’s effort and inspiration—off as one’s own is, in fact, intellectual theft. Still, obviously, some instances of intellectual theft come closer to shoplifting a packet of Kleenex, while others are more the equivalent of grand larceny.

I do fear, however, that plagiarism is an area in which AI will fundamentally change almost everything. As I have written in this Substack. I have my reservations about Graham Burnett’s techno-optimism. But even if AI is not going to churn out acceptable academic monographs “at the push of a button” any time soon, it will nevertheless transform our understanding of what counts as original work, and therefore what counts as intellectual theft. The technology is simply too good, and too easy to use. Everyone is going to use it, whether we like it or not. In the future, very little writing is going to be purely the product of individual effort and inspiration. And when machines can draft competent prose, summarize scholarship, rewrite paragraphs, and imitate styles in seconds, the central ethical question will no longer be whether particular sentences have been copied. It will be whether a piece of writing embodies original thought. That is a much harder standard to define—and one that universities, publishers, and scholars have only begun to confront.

And here is a good example of the problem. I finished the previous paragraph with the words “the product of individual effort and inspiration” and then asked ChatGPT to provide a conclusion. The AI wrote the paragraph’s final three sentences. I wouldn’t have ended the paragraph in quite the same way myself, but Chat came pretty close, and I saved myself some time and effort. Are those three sentences “mine”? Is it intellectual theft for me to pass them off as mine? Would it be plagiarism for someone else to take credit for them?

We are heading into absolutely uncharted intellectual territory. Here there be monsters. The ride is going to be scary.