Following up on this, philosopher Dan Kaufman, who taught for a quarter-century at Missouri State University, shares a draft of his provocative remarks being presented today at Florida Gulf Coast University. Some of this seems to me right, some debatable, all interesting. Comments are open, and signed comments will be preferred. Professor Kaufman’s text follows:

AI and the University

The university was once an institution whose purpose was the education and acculturation of elites, which is why classics dominated the curriculum. This conception of the function of higher education was revised in the mid 20th century to include science and engineering, a development that was controversial in itself as evinced by the (in)famous F.R. Leavis/C.P. Snow “Two Cultures” debate. But in the United States, the real transformation occurred after the Second World War and the GI Bill, when the university began to turn, gradually, into an institution of mass education. And while its research mission largely remained unchanged, its main pedagogical charge today is to prepare and credential people for the white collar professions. This new profile and mission, when combined with contemporary market forces, have put the entire system at risk and left it even more exposed and vulnerable to the impact of AI than it would have been otherwise. Indeed, I don’t see how the university survives in its current form.

Once higher education became mass education, efficiency and cost took on an overriding significance. The research, residential, and amenity-rich infrastructure these places provide is an extraordinarily expensive proposition, with the costs being passed directly down to the student. This was fine when undergraduates were overwhelmingly from privileged backgrounds, and half of the point of going to university was for the experience, socialization, and connections doing so would garner — I was in no hurry to graduate from the University of Michigan in 1989 — but in the current environment, the mad inflation of undergraduate tuition and residency has sent students and their families scrambling for faster and cheaper ways to complete an undergraduate degree.

Community colleges have risen to the challenge, transitioning from the largely vocational role they once played, into a cheap alternative for the first two years of college. So swiftly has this occurred, that within the twenty four years that I taught at Missouri State University — from 1999-2023 — I watched it change, in real time, from a traditional, four-year, residential institution into what is essentially a two-year finishing school. Beyond gutting programs whose enrollments depend on their place in the general education curriculum, the question of why the university should maintain all of these fantastically expensive student-life amenities for students who will not be living on campus — as few upper-classmen do — becomes harder and harder to answer.

Speaking of general education, it isn’t long for this earth. Even back in my day, the question of why an aspirant hotel manager or accountant should have to slog through two years of general education studies, in addition to a major and minor, had become difficult to answer in a way that did not strain credulity, but today, the rationales have become so thin that they no longer bear any scrutiny at all. “Well-roundedness” is all very well, but not to the tune of $100-200K, per student. Even putting aside the current moment of degenerate political leadership, in which rationales like “Education for citizenship” and “Developing the next generation of leaders” sound like a sick joke, whatever momentary plausibility they might have had evaporates, once you realize that they imply that those without college educations — around 60% of the population — are unsuited for citizenship or leadership, which is demonstrably untrue. And I shouldn’t have to explain why saying, “You need to attend our institution for four years, so we can justify millions of dollars in infrastructure” is not going to resonate with those who are so strapped that they don’t know whether they’ll be able to pay their mortgage or health insurance premiums from one month to the next.

But, you’ve come to hear about AI, and my point is that given the current state of higher education, adding it to the mix is like pouring gasoline on a smoldering pile of coals. The introduction of LLMs into an institution of higher learning would be disruptive under any circumstances, but the transformation of the university into a white-collar vo-tech had already begun to send the resulting, incoherent, bloated mess of an institution on its way out the door. Adding AI will simply accelerate the process. Whether this is to be celebrated or bemoaned, I leave up to you, but that it will happen, I have little doubt.

Most of the critical focus has been on the “consumption” side of the ledger, and since I am now retired, I can only sympathize with the legions of teachers who find themselves wringing their hands over how to credibly assess students in the age of generative AI. Even before the widespread use of this software, cheating had become a serious problem, due to the availability of online encyclopedias, which made nonsense out of any assignment intended to be completed at home. In my classes, all the assessment used to be take-home — papers, exams, etc. — so as not to sacrifice precious lecture hours, but when more than half of what I received one semester was copied straight from Wikipedia or the Stanford Encyclopedia of Philosophy, I switched entirely to in-class modes of assessment, a practice I maintained until my retirement. (Giving half the students zeroes — my preferred penalty for cheating — was not an option the Dean would allow.) But, from what all of my friends who remain in the business are telling me, generative AI has made the situation much worse.

The problem is so obvious that it shouldn’t need repeating: the point of education is for students to acquire skills that they previously lacked, and if they cheat their way through school, they won’t. But, this is an even bigger problem, now that going to college has been reduced to a purely transactional arrangement.

Someone who pursued higher education on the old model might still get a lot out of the experience, even if he or she learned nothing academically speaking. (I’m imagining a real-world version of Bertie Wooster.) But, if the sole reason for going to college is to learn how to be an accountant, hotel manager, social worker, etc., then the fact that a person has not learned how to do the relevant tasks renders the entire exercise — and expenditure of money — pointless. And if too many generations of graduates come out of the university not knowing how to do whatever it is that the job for which they have been preparing requires, then it will not be very long before employers stop viewing a college degree as any kind of credible credential. Indeed, this process is already underway.

Much less discussed is the impact of generative AI on the production side of higher education, and it is on this that I want to spend the remainder of my time. 

We already treat undergraduate teaching — and especially, introductory-level teaching — as a burden to be offloaded to the university’s lowest-level, most poorly paid employees. At places with graduate programs, this means they are taught by probationary faculty, adjuncts or graduate students, while in the schools lacking them, it is just the first two. (Of course, if your department is small and unimportant enough, as mine was, all of the faculty may have to do at least some introductory-level teaching, though it still is widely resented.) With the imperative being to teach stuff as cheaply as possible, handing the job over to chatbots seems inevitable, and is already happening in places. Will doing this make the skill-acquisition problem even worse than it already is? It seems a near certainty. Will anyone be willing to pay 100-200k for it? It seems unlikely. Does this mean that large numbers of faculty — and the people managing the administrative infrastructure around them — will be left unemployed? Undoubtedly.

As bad as all of this is, however, what the widespread use of generative AI will do to academic research is worse.

In my own discipline — philosophy — there is a disturbing trend in which my colleagues not only publicly admit to using chatbots to conduct their academic research and write their papers, but are proud of doing so and suggest that AI-written philosophy represents an inevitable future. If this is happening in a subject as unfriendly to automation as philosophy, you know it is much worse in those areas that are more amenable to it.

I have no doubt that some people will be interested in reading scholarship “written” by a chatbot, but the idea that this could sustain professional academic disciplines is fanciful at best. Given how chatbots work, there is only so much recycling, re-aggregating, and redistributing of existing material that is possible before it all becomes rather thin stuff, and the more scholars use AI to do their research, the less original writing will find its way into the bots, and the more anemic academic publishing will get. Eventually, given enough time, little to nothing will be left but recycling, re-aggregation and redistribution.

The argument one always runs into, when engaging with the subject at this level, is that the purpose of scholarship is to produce truths, and that it doesn’t matter where they come from. When having such conversations, AI defenders will ask, “If a cure for cancer was discovered by a chatbot, wouldn’t it be just as useful as if it had been discovered by a person?” and the question is a perfectly fair one, but only in a very limited sphere. Certainly, the value of some scholarship is a function of this kind of simple utility, but much of it is not. Indeed, outside of medicine and engineering (broadly construed), I would suggest that the value of most scholarly research cannot be measured by such a calculus. Please understand that I am not invoking the airy — and in my view, purely rhetorical — idea that academic scholarship is intrinsically valuable, a claim that is both undemonstrable and in a transactional framework, unpersuasive. Rather, I submit that in many if not most cases, scholarship has no immediate or obvious material benefit but rather, adds to our collective understanding in ways that may prove useful in the future, but often in a manner that we cannot specify or predict.

But, even with regard to scholarship whose value is a matter of simple and demonstrable utility, the offloading of it to generative AI raises the long term problem just mentioned. The more the material on which the bots are trained is itself produced by bots, the less original, human input goes into the system, and the more iterations of reaggregation occur, the less new knowledge will be produced. Sure, right now, chatbots are going to “notice” useful aggregations of existing scholarship that we might have missed and which may produce novel technologies, cures, and the like, but the longer and more involved in the process they are, the more this will yield diminishing returns, until eventually, little to nothing novel comes out of scholarship at all. Certainly, human imagination has a lot to do with this, but perhaps even more important are our values and ambitions which, of course, lie beneath every human endeavor, not just scholarship.

At this point, I can only hope that those running our colleges and universities will come to understand the essentially human and social role that they play in our civilization, and sooner rather than later, given the speed with which this second transformation is occurring. It took decades to turn our colleges and universities into white collar vo-techs, but I fear that their automation by chatbots may take only a single one. This hope, however, is tempered by the sober realization that at least thus far, not only do our colleagues not appear to see the proverbial writing on the wall, but far too many of them think that handing over their work and their institutions to generative AI is actually a positive development.

If nothing is done, and we continue on our current course, I see the following happening. (These are, of course, guesses not predictions.)

–Spending four of one’s most productive years in residence for purposes of acculturation, socialization, contemplation, and leisure is by its nature an elite activity and will become one again. It will be done primarily at SLACS and the most prestigious, well-endowed universities. 

–Past the point of entertainment, the humanities and liberal arts will largely become the province of “educated amateurs.” They will not remain professional, academic disciplines in most places.

–Education and training for the blue- and white-collar professions will be done in dedicated trade schools, and will take a year and a half to two years to complete. If there are masters levels of study, they will be combined into a single 3-4 year program.

–Research and Development will be done at dedicated institutions, financed by a mixture of government and industry funding. They will be charged both with teaching the next generation of researchers, as well as conducting the research itself.

–The number of four-year colleges and universities, commensurately, will shrink drastically, with the state university system and middling private institutions being hit the hardest.