Aspirational Attribution: A Response to Lemley & Ouellette, Plagiarism, Copyright, and AI

Let another man praise thee, and not thine own mouth; 
a stranger, and not thine own lips.1

For better or worse, plagiarism norms have existed since time immemorial. Ancient Greek philosophers accused each other of plagiarizing ideas, and the Roman poet Martial coined the term, while offering to sell his byline to his plagiarist. While plagiarism admits different definitions at different times among different groups of people, its essence is always unattributed copying. Authors beget expressions and ideas, plagiarists adopt them as their own.

I must confess to a morbid interest in plagiarism, the only capital scholarly offense. While the academy can forgive just about any other transgression, plagiarists are pariahs. Even an unsubstantiated accusation leaves an indelible stain, so when the academy deigns to absolve a plagiarist, it must pretend the offense never occurred.

And yet, the justification for plagiarism norms is peculiarly unclear. Plagiarism scholarship is voluminous, but unusually uniform. Almost everyone agrees that plagiarism is bad, so we should prohibit and punish it, but no one bothers to explain why, and the few exceptions merely prove the rule.2 Accordingly, I applaud Professors Mark A. Lemley and Lisa Larrimore Ouellette’s refreshingly skeptical essay Plagiarism, Copyright, and AI, which argues that academic plagiarism should be an ethical, but not a legal, offense.     

Of course, I was delighted to see that Lemley and Ouellette found my scholarship helpful. As an otherwise lonely voice in the wilderness, it’s nice to hear an echo, however faint. Perhaps unlike other scholars, I was only disappointed that they cited my work, rather than plagiarizing it. After all, only plagiarists love words and ideas so much they want to make them their own.

Like many people, Lemley and Ouellette are worried about AI. Specifically, they are worried about the effect of AI on attribution. Academic plagiarism norms require the attribution of all expressions and ideas copied from another source. But when students and scholars use an AI model to generate a text, they often don’t know the original source of the expressions and ideas it includes, or even that a source exists. Lemley and Ouellette argue that it’s still plagiarism, and that academic plagiarism norms should require both students and scholars to identify and attribute the source of expressions and ideas in AI-generated texts.

I suspect most people will find this claim uncontroversial or even obviously true. Of course students and scholars should be required to attribute expressions and ideas, especially when they appear in AI-generated texts of dubious reliability. In fact, they should be writing the texts themselves; attribution is the bare minimum. At the very least it would help them avoid AI “hallucinations” and other errors.

I’m not so sure. While I agree with Lemley and Ouellette that plagiarism is not and should not be a legal offense, I cannot agree that academic plagiarism norms are still justified by efficiency and morality. They argue that plagiarism norms are efficient because attribution gives marginal scholars a salient incentive to produce additional scholarship, but provide no evidence to substantiate their claim, even though scholars have many other incentives to produce scholarship. And they argue that plagiarism norms are moral because authors and readers are entitled to attribution, without providing any consequentialist justification at all. I don’t find either of these arguments convincing. 

A decade ago, Lemley scandalized intellectual property scholars by arguing that consequentialist and deontological theories of intellectual property are fundamentally incompatible. Consequentialism holds that intellectual property rights are justified only if and when they cause good outcomes, while deontology holds that intellectual property rights are inherently justified.3 Or rather, consequentialists think intellectual property is a means to an end and deontologists think it’s an end in itself. Accordingly, consequentialists and deontologists “have nothing to say to each other” because they “simply cannot speak the same language.”4

Of course, the observation that consequentialist and deontological theories of intellectual property are in tension is banal. The controversial part was Lemley’s characterization of consequentialism as “evidence-based” and deontology as “faith-based.” Unsurprisingly, deontologists took it as a slight. It’s one thing to say scholars are wrong, but quite another to call them ideologues.

Ironically, I fear that Lemley and Ouellette’s defense of academic plagiarism norms is based on faith, rather than evidence. They believe that mandatory attribution of expressions and ideas will encourage marginal scholars to produce more scholarship but provide no evidence. And they have faith that scholars and readers deserve attribution of expressions and ideas but don’t explain why.

As a zealous consequentialist, I think intellectual property rights are justified only if and when they produce good outcomes. So, copyright is justified only if and when it actually provides a salient incentive for marginal authors to produce additional works of authorship, and the same is true of plagiarism norms.5 I assume that copyright is sometimes justified, but observe that it often isn’t, and I assume the same of plagiarism norms. The hard question is when they are justified and when they aren’t. And that question can only be answered with evidence, not assumptions or beliefs. I am not convinced that the evidence shows that academic plagiarism norms actually provide a salient incentive to marginal scholars. And I don’t know what evidence could show that authors and readers are entitled to attribution.

Lemley and Ouellette object that I haven’t produced any evidence proving that academic plagiarism norms don’t increase the production of scholarship. Guilty as charged. But why should the skeptic bear the burden of proof, rather than those defending the legitimacy of privately-enforced property rights in public goods? Academic plagiarism norms have existed in one form or another for at least 2000 years. If they actually increase the production of scholarship, surely evidence of their efficiency ought to be readily available.

Of course, as Lemley and Ouellette observe, it’s at least plausible that plagiarism norms could give marginal scholars a salient incentive, because scholars like attribution. But it’s equally plausible that copyright could give marginal scholars a salient incentive, because scholars also like money. Surprisingly, authors are not necessarily rational economic actors. While English writer Samuel Johnson famously observed, “No man but a blockhead ever wrote, except for money,” the blockhead wrote for free all the time. Most scholars do the same, participating in a gift economy in which they write for free. Maybe they expect payment in credit? After all, it’s a truism that attribution is the coin of the scholarly realm. If so, the credit is usually, “I’ll gladly pay you Tuesday for a scholarship today.” Most scholarship is never cited, and often never even read. If scholars are motivated by the promise of attribution, hope has truly triumphed over reason. Surely they have other reasons to write.

Take legal scholarship. Sure, the possibility of attribution gives marginal legal scholars an incentive to produce additional works. But they have other incentives as well, many of which are considerably more salient than attribution. For example, legal scholars must write in order to publish articles, find a teaching job, attend conferences, land speaking engagements, get tenure, be promoted, receive research funding, move to a different school, and so on. By contrast, while attribution is always nice, its benefits are indirect, inchoate, and unpredictable.

As a consequence, when it comes to attribution, most legal scholars are largely “inframarginal.” While they care about attribution, they care about other things more, usually a lot more. After all, it’s hard to know if another scholar will cite your work, not to mention whether anyone will notice or care when it happens. While I remain open to the possibility that academic plagiarism norms solve market failures in scholarship, as far as I can tell, the evidence is inconclusive at best and unconvincing at worst.

But Lemley and Ouellette also argue that academic plagiarism norms are justified because authors and readers deserve attribution. Or rather, they argue that academic plagiarism norms are justified because attribution is helpful for authors and readers, which amounts to the same thing. After all, how does attribution help authors and readers? It helps them know who owns an expression or idea and is entitled to attribution when it is used.

Maybe Lemley and Ouellette think scholars deserve to own expressions and ideas. If so, I have nothing to say in response because that is a deontological proposition, not a consequentialist one. But I do have some tangential observations that may be helpful, or at least provocative.

As I have previously observed, everyone believes in the legitimacy of the kind of property they want to own, and scholars are no exception to the rule. Scholars want to own expressions and especially want to own ideas. Accordingly, scholars tend to believe in the legitimacy of academic plagiarism norms because they’re the best way of claiming and enforcing ownership of ideas. Where copyright says no, plagiarism norms say yes. If knowledge is a commons, then academic plagiarism norms are the fences that scholars use to enclose it. Unsurprisingly, like all enclosure schemes, they tend to benefit incumbents at the expense of new entrants: “For unto everyone that hath shall be given, and he shall have abundance, but from him that has not shall be taken away even that which he has.”6

Let’s assume academic plagiarism norms are justified, and scholars are entitled to own attribution rights in expressions and ideas. Who decides what they own and what they don’t? Anyone who has ever attended a faculty workshop knows that scholars are inclined to see their ideas everywhere they look. And their peers are inclined to agree with their assertions of ownership, either out of self-interest or in the interest of preserving peace.

What’s more, there are lots of good reasons to object to attributing expressions and ideas, even when academic plagiarism norms say attribution is mandatory. You might think someone doesn’t deserve attribution. You might think someone else deserves attribution more. You might think no attribution is necessary. You might find a scholar’s work unhelpful. You might find a scholar odious. And so on. At the very least, the plagiarism police need to recognize a fair use exception.

Returning to Lemley and Ouellette’s proposal, I fear it is unworkable and unwise. At least in theory, academic plagiarism norms require the attribution of all expressions and ideas, except for those unique to the author. But there is nothing new under the sun.7 Anything and everything you might conceivably say has already been said by someone, somewhere, sometime, or might as well have been said. Indeed, academic plagiarism norms at least purport to require attribution even if the author was unaware of the copied work at the time of writing.

Accordingly, Lemley and Ouellette argue that academic plagiarism norms should require students and scholars who use AI to produce scholarship to attribute all of the expressions and ideas the AI model copied from another source. But how and why? They could google each phrase or passage in search of precedents, or ask an AI model to do the same, but what would be the point? Anyone else can do the same. What’s more, AI models are machines for reproducing conventional wisdom. If an AI model produces an idea, and that idea is similar or identical to one previously expressed by a person, how are we to know whether the idea is unique to the person or merely an expression of conventional wisdom? Or rather, if an AI model produces an idea, shouldn’t we assume it’s conventional wisdom, owned and ownable by no one?

For the record, I don’t oppose attribution. On the contrary, I encourage it! Authors can and should attribute expressions and ideas whenever they believe it’s helpful or deserved. But I think attribution should be voluntary, not mandatory. As the late messiah admonished us, “freely ye have received, freely give.” Authors should be free to decide for themselves whether and when attribution is necessary. Let us attribute out of love, not obligation. We will all be happier and wiser for it.

* * *

Brian L. Frye is a Spears-Gilbert Professor of Law, University of Kentucky J. David Rosenberg College of Law.

• Proverbs 27:2 (KJV).
• For an especially eloquent dissent from conventional wisdom, see K.R. St. Onge, The Melancholy Anatomy of Plagiarism (1988).
• See Mark A. Lemley, Faith-Based Intellectual Property, 62 UCLA L. Rev. 1328, 1341 (2015) (“A utilitarian IP framework has a metric for deciding whether we should give control over those terms to the people who claim them. But if IP is a Right, granted to the first creator not for a purpose but simply because they are first, it is hard to find a similar limiting principle.”).
• For what it’s worth, I suggested that consequentialist and deontological theories of intellectual property can be reconciled by adopting a consequentialist public theory and deontological private theories. See generally Brian L. Frye, Machiavellian Intellectual Property, 78 U. Pitt. L. Rev. 1 (2016).
• Notably, Lemley agrees. See Mark A. Lemley, Faith-Based Intellectual Property, 62 UCLA L. Rev. 1328, 1340 n.48 (2015) (“Even the justification for moral rights seems to me more utilitarian than moral—we should give rights of attribution and integrity to creators if those are in fact the things that motivate them to create, as some recent scholarship suggests.” (citing Jeanne C. Fromer, A Psychology of Intellectual Property, 104 Nw. U. L. Rev. 1441, 1443–44 (2010)).
• Matthew 25:29; see also Robert K. Merton, The Matthew Effect in Science, 159 Sci. 56, 57–58 (1968) (presenting evidence that academic citation practices benefit incumbents at the expense of new entrants).
• Ecclesiastes 1:9 (KJV).