Academic Doomerism
AI doomerism is everywhere these days. Every field, and recently humanity as a whole, has had its "we're finished" post. In contrast to the run-of-the-mill hot take, Jason Potts has written an economics paper on why academia is doomed. So let's dive in.
What does a university sell?
Potts says a university is really a platform. It is a hub that connects many different groups: undergrads, grads, teaching staff, research staff, employers, government, alumni, donors, parents, etc. Each group needs the others. Undergrad tuition helps pay for the research infrastructure for the professors whose research reputation drew the students there in the first place. International student fees provide funding for local students. Unfortunately, this isn't a diversified portfolio kind of situation, but more of a weakest-link setup. If you pull on one thread, several others start to come loose as well.
(Side remark: I will admit that I have always struggled to find the true customer the universities served. It is not the faculty, not even the students... This platform/hub explanation makes sense, of course. But I wonder if this might be a convenient cover up to make the disorganized/disarrayed/organically-complicated state of universities look better.)
Since the university has this multifaceted platform/hub status, free availability of teaching technology (like books, online classes, the internet) never hurt the university. Potts argues, the university never really sold content, it mainly sold matching and verification. A degree is a certification. It tells an employer that this person has some ability the employer can take for granted.
Potts argues that AI is the first technology to disrupt that promise directly. AI makes it cheap to produce essays, code, homework, which the shools use for grading and certification. Potts calls this signal collapse.
He highlights two other disruptive developments. First, the university gets cut out of its traditional brokering position. It used to connect students to teachers and the researchers to funding. Now a student can get free tutoring from LLMs, and a researcher can work alone with an AI assistant instead of a lab full of people. Second, Potts argues that AI changes where the money comes from. It used to come from knowing a fixed body of facts, but now it comes from being able to learn something new fast. (I don't buy the second half of this. I think the comparative advantage shifted towards creativity and judgement, as I argue in the discussion at the end of this post.)
Potts splits the universities into three types: elite, specialist, and all else. Elite schools are mostly fine, because their real product was always the prestige/selectivity and the classmates you meet there. Specialist schools (like medicine) are OK because of their hands-on status and that outside boards do the checking for them. All others are in trouble.
The warrant was already broken
Potts treats the degree as a strong asset, a Hart asset that an institution must protect above everything else, and argues that this is now facing a strong threat with AI. But that is not true, as universities had already devalued their "credible warrant of quality" feature by diluting their degree programs starting around 2010.
I lived through this. University administrators got greedy and chased more enrollment and more tuition. As a result, programs multiplied, courses got easier, and grade inflation become the norm. This had the same effect Potts describes for AI: a good grade stopped meaning the student actually mastered the material. The credential lost its meaning from the inside, at the hands of the very institution meant to protect it. I saw this firsthand. The SUNY Buffalo CS degree lost real standing with companies like Bloomberg in New York over those years, many years before the AI threat materialized. Every company started their own rigorous interviewing process rather than taking university certification at face value.
That this devaluation already happened matters for what to do next. If the degree still had high prestige, one of Potts's fixes, bringing back the oral exam, would work cleanly. But a school that spent 15 years training students, parents, and employers to expect easy A's cannot win back trust just by adding oral exams. It has the uphill battle ahead to convince the market to trust it again. Trust is easy to lose, and hard to gain back. Unfortunately, this means that the ordinary school in Potts categorization has nothing left in reserve. It's the school least able to take this hit, and it's now going to get hit with the AI wave now.
What should the universities do
I suggest something Potts does not discuss. The universities should lean heavily towards humans' comparative advantage over AI.
AI can already write good code and do decent technical work, but it keeps failing on things without clear right answers. This makes creativity, judgment, and the authentic human voice the scarce resource (where the opportunity cost is lowest). And these are what a university should be teaching toward.
Potts does not talk about the human side of the story at all. I spent sixteen years as a professor before moving to industry. What I remember most, and miss most, is watching a student's eyes light up when an idea finally clicks. A good teacher doesn't just regurgitate the course content, rather they pass on the love for the subject. When you see someone who has spent thirty years on hard problems still lighting up talking about them, you want that for yourself as well. No AI model can match that inspiration, and light that fire. My own advisor did that for me, and watching him approach problems shaped how I think to this day.
Wrestling with a problem, persistently trying out various strategies, being unafraid of making mistakes, and progressing incrementally to understand the underlying ideas produces a certain kind of endurance, which enables us to be comfortable with the struggle.
The best parts of any real education happen off the page. None of this shows up in a course catalog, and none of this can be faked. It can only be earned slowly through hard effort, through apprenticeship (which is Lindy), and often through osmosis from another caring human being.
Every time that a human being succeeds in making an effort of attention with the sole idea of increasing his grasp of truth, he acquires a greater aptitude for grasping it, even if his effort produces no visible fruit.
--Simone Weil
Comments