Tell Me Your Complaints About AI and I'll Tell You Why You Think You Matter

Every complaint about using AI to do X contains fascinating and often unexamined assumptions about what makes “doing X” a valuable activity. But when “doing X” is part of one’s sense of “mattering” in the world, as Rebecca Goldstein would say, examining those assumptions threatens one’s conception of self-worth. Hence the dread.

Suppose you’re a legal academic (or another kind of academic, for that matter). When do you think that using AI to write law papers become “cheating”? Whatever your answer is, it tells us a lot about what, in your view, makes your papers valuable and your activity of “making law papers” part of your sense of mattering.

Is it the conception of the central thesis? The identification of strong arguments in support of the thesis? The studying and researching prior to that conception? The studying and researching after that, to situate and stress-test the thesis and arguments? Is it the structuring of the conceptual and explanatory moves? Is it the choice of the most important words and phrases? The overall “voice” and style? The identification of the practical applications and payoffs? The choice of a good title?

“Making” academic papers consists of many parts. Classical rhetoric can help us identify the various parts. (Rhetoric can help us identify the various parts of “making economic papers” as well: everyone needs rhetoric, more or less consciously.)

Coming up with ideas and arguments is what Roman orators called inventio, i.e. invention. One must find the issue; then look for good arguments in one or more of the several standard places or “topics” where arguments are typically found (definition, genus and species, part and whole, cause and effect, comparison, probability, etc.); then build the argument (proposition, reasons, proof, etc.); then arrange the paper (for example, one must write an effective introduction: how important is that?); choose the right style; the right words to explain one specific concept or move; etc. There’s more, of course. For example, one must choose how to summarize the main thesis in a brief verbal exchange, how to present the whole paper to a combative audience, how to respond to criticism, and so on.

These are all parts of what “making a paper” is. But which of these parts (or their many subparts) make “making a paper” valuable? Which of these tasks make the “maker of papers” matter?

One is tempted to answer: all of them. But nobody believes this answer. Academics routinely delegate some of these tasks, in whole or in part. Bibliography, footnoting, data collection, data analysis, coding, literature reviews, legal research, revisions, stress-testing, and even writing (although many law people, myself included, value this part too much to give up on it) are constantly delegated to research assistants.

But the number of “parts” that come from “outside” is much larger than the set of “parts” supplied by RAs. We workshop papers: and in those settings we get ideas, objections, examples, arguments and counterarguments, choice of phrases, applications, and many of these things change the paper in important ways. Some of us get referee reports or written comments from colleagues. We get ideas, arguments, moves, structural and stylistic choices from a number of people whom we do not considered “co-authors” of the paper and whose contribution does not feel like a threat to the value of what we do.

So where is the line? When exactly do these various forms of collaboration and exchange begin eroding one’s sense of the value of one’s own contribution?

AI feels different (and perhaps is different) because it’s not human. When we delegate to RAs, chat with colleagues, revise & resubmit, workshop papers etc., we participate in a social activity. Our sense of scholarly mattering is not entirely based on a solitary (romantic or solipsistic, you choose) quest, but on a social practice. Conversing with AI may give us the facts, stress-testing, revisions, objections, and examples that RAs, referees and colleagues give us, but cannot give us the sense of participating in a social practice.

That’s one complaint against AI and one corresponding assumption about the meaning of what we do: social practices.

AI feels different because it’s easy. Workshops are hard. RAs miss deadlines. Colleagues can be brutal. Going through these things is a struggle. Getting to a similar outcome without the struggle feels like (and perhaps is) cheating.

That’s another complaint against AI and another corresponding assumption about the meaning of scholarship: struggle.

AI feels different because it can do the writing. This is something that different kinds of scholars feel very different about. Economists care about it much less than legal academics; and mathematicians much less than economists. And, I suppose, scholars of securities regulation or tax or antitrust care less about it than scholars of legal history or legal philosophy. (To be sure, some individual writers care about style much more or much less than the typical member of their sub-tribe.)

These differences tell us something very important about these tribes’ sense of “mattering.” What is it all about? Inventio or elocutio, coming up with the idea or writing it down in an elegant and effective manner?

We might soon find ourselves in need of examining, re-examining, or even revising and reinventing our mattering assumptions. It’s a scary thought. But the sooner we start doing this work (and stop complaining) the sooner we’ll find, let’s hope, a better sense of our place in the world.