Three bad ideas that are good for US higher education
Logpodge #10 is all about strawberry statements
Note: Logpodge posts are a break from my usual long-form writing and book reviews, presenting short, topical essays. If I wanted to optimize my Subtack, I would drip these out as separate posts, but as Iβm looking for readers, not followers, I send them out collected in a single post to avoid spamming your inbox.
Like Dave Karpf, Iβm a Substack free-rider, having decided not to get on the paid newsletter treadmill. That will make it easy for me to be among the first rats off this platform, if and when Substack shifts to monetizing eyeballs instead of giving readers. The fact that Substack lets you trust me with your email address is what convinced me to follow Lance Eaton from LinkedIn to Substack nearly two years ago.
Anyhoo, subscribing is the best way to continue to find my writing when Substack follows its fellow VC-funded platforms into the internet shitterβs vortex and I jump ship.
Offering the machine what doesnβt matter
I am a fan of
heterodoxy when it comes to AI and higher education, not because I agree with her, but because she offers ideas that challenge the usual terms of what Henry Farrell calls AI Fight Club. AI Fight Club is like Chuck Palahniukβs version, except the fights are rarely physical, and everyone talks about the club, which pits those who see AI-based educational tools as our inevitable future against those who resist AI in large or small ways.1In proposing that CSU automate general education through a microservices architecture consisting of AI applications, Robbins offers a concrete alternative (a genuinely awful one) to the already existing structure of general education at many universities (also genuinely awful). The awfulness of existing programs has been created by the adjunctification of teaching, the curricular standardization of big state systems, and the increasingly transactional expectations of students. Now that OpenAI has built Shel Silverstein's "Homework Machine" and offers it to college students at no cost, there really is no way to look at the first two years of college and believe all would be fine if we just had a few more resources to spend on adjuncts or digital classrooms. Love it, hate it, or try to ignore it, generative AI is forcing us to face how little actual education has been going on in general education programs.
Robbinsβs proposal presents an unflinching alternative to a fight club compromise or continuing to ignore problems that are now impossible to ignore: just have AI do it. My amendment to her 7-step process of having AI function as a general education program is to eliminate two middle steps: Step 4: Develop the Project-Based Learning System and Step 6: Integrate Human Mentorship and Collaboration. Using AI along with humans for two activities is not efficient. Instead, I think we should decouple credentialing and learning, bringing the trend over the past (three? five? ten?) decades to its conclusion. This would give the βstakeholdersβ who want to build an AI-empowered university system what they care about most while leaving students free to figure out what they want to learn, with a few mentors around to help.
While students figure out how best to navigate the automated machinery that will grant their degrees and badges, the faculty create loose structures of in-person connections where students who are interested pursue knowledge via seminars and projects under the mentorship of experts. Let the machinery do the credentialing as efficiently and effectively as AI-powered digital platforms can manage it, while faculty and students figure out real life and real learning together in face-to-face interactions.
It goes without saying that this terrible idea would never work. But creating a sharp division between building the AI-empowered university system and learning might get us somewhere.
Embracing randomness
At
, Chad Orzel writes about admissions to highly selective colleges. College Admissions is a Hard Problem lays out a quantitative analysis of selective admissions free of the anxious pontificating and obsessing over our obsession with standardized tests that usually informs such discussions. Orzel lists several bad ideas for improving admissions, concluding with βsome sort of thresholded lottery system, where everyone who meets a minimum standard goes into a random drawing for the available spots in the class, including the awful idea of randomized.βI am interested in this one because students taking my course this fall, all of whom will have just been admitted to a selective program at an Ivy League university, will be talking and writing about lottery-based admissions. One of the first concepts we will wrestle with is βpredictive AI,β which does not get the attention it used to, thanks to ChatGPT.
Arvind Narayanan, one half of the AI Snake Oil duo, is an advocate for randomized admissions. He and his co-author, Sayash Kapoor, write about it in theirΒ book, which students in my course, How AI is Changing Higher Education, will be reading in actual, physical form this fall. As I wrote about in my review essay about AI Snake Oil and The Ordinal Society, the role that pseudo-Bayesian probability plays in AI hype has made it hard to talk about how uncertainty relates to AI. Enthusiasts are just so sure AI can do that, any day now. βThe habit of confident prediction, especially when expressed probabilistically, gives a rational sheen to the most unhinged speculation,β I wrote. Narayanan and Kapoorβs discussion of uncertainty is the antidote to the habit of confident prediction. Randomized admissions is a way of making ideas about contingency and probability feel concrete and personal, at least I hope so.
Having succeeded in winning the somewhat rigged lottery that is the current admissions system, Iβm curious what students will make of the idea of a thresholded lottery, which, as Orzel notes, appears to be the least-bad option for those who want change. Orzel gives some of the reasons to hate the idea, which for him mostly involve the way the current system helps βfitβ students to institutions, and vice versa. Iβm skeptical that schools have any way to meaningfully evaluate such a thing as fit, and the larger the applicant pools and the greater the automation of the screening process, the more skeptical I am.
So, the challenge to my students, and really to all involved in selective academic programs, is how are you going to change admissions processes as applying becomes a matter of applicants using generative AI to stand out and screening becomes automated through the application of predictive AI? If embracing randomness is a bad idea for meeting the challenges of AI, whatβs a better one?
Columbia didnβt settle. Columbia aligned.
Reading The Strawberry Statement at age 13 was important to the early stages of my understanding of higher education. Written by James Simon Kunen when he was 19, the book describes his experience as an antiwar protester at Columbia in 1968. It has been decades, so the details are hazy. I remember it as funny and mostly not about the Columbia protests. But it left me with the sense that college happens outside classrooms as well as inside, and that universities are important to the American experiment in democracy. Also, it gave me the sense that Columbia University mattered.
The bookβs title comes from a statement made to the student newspaper by a faculty administrator in the midst of the protests. The bookβs cover reminds me that the administrator was Herbert Deane and that the exact quote was: "A university is definitely not a democratic institution. When decisions begin to be made democratically around here, I will not be here any longer,β followed by the statement: "Whether students vote 'yes' or 'no' on an issue means as much to me as if they were to tell me they like strawberries."
Columbia could use leaders like Herbert Deane today. At least he had the courage of his anti-democratic convictions and spoke the truth. Instead, they have Claire Shipman, the board of trustees co-chair, who stepped down into the role of acting president in order to defend what matters most: the revenue flowing from the federal government to the university. Her statement that the negotiation completed earlier this week βsafeguards our independence, a critical condition for academic excellence and scholarly exploration, work that is vital to the public interestβ lacks courage because it is untrue. By capitulating, Colombiaβs leaders sacrificed the public interest and their institutionβs independence in a desperate attempt to end the very real pain the Trump regime was delivering.
Nothing about the agreement safeguards anything because, as Columbia economics professor Suresh Naidu writes:
Trade negotiators from longtime partner countries, government contractors, law firms, federal employees, permanent residents, the Federal Reserve chair Jerome Powell, even the Transportation Security Administration labor union are all experiencing contractual vertigo, finding out that the administration will not honor previous agreements.
An agreement that will not be honored and cannot be enforced is not an agreement. As The Specter, βa faculty and staff-driven newsletterβ at the university, put it, βColumbia didnβt settle. Columbia aligned.β Those feeling a sense of relief because the money will flow again are hoping that alignment is all they need for things to return to normal. Except that now the money flows both ways as Columbia agreed to pay $200 million in βfinesβ in exchange for receiving money had Congress allocated.
This is not normal. Though, asΒ Adam Tooze provocatively reminds us, what he calls βTrumpβs βad hoc regulationβ or βgovernanceβ has a history.β Perhaps a better word is precedent, because this sort of ad hoc deal-making βcharacterizes much of American corporate, business and public life today.β This βlawfareβ happens outside courts when entities and individuals invoke the law in ways that βcome with a connotation of menace, threat, extortion, ruinous and arbitrary fees, obscure deal-making, hidden clauses, life-ruining nuisance suits, and bizarre somersaults from the freedom of speech to accusations of terrorism.β This is an interesting concept applied to the rule and practice of law in early twenty-first-century America, but it does not apply to Columbiaβs situation.
What Tooze elides is that Trump brings state power to his side of the negotiating table. The outcomes of βlawfareβ as practiced by everyone else are sticky. Parties resolve their conflict by settling, like Columbia thinks it did. Shipmann and her advisors were acting with what Thorstein Veblen called a trained incapacity, βthat state of affairs whereby oneβs very abilities can function as blindnesses.β Conditioned by the rules and norms of lawfare, Columbiaβs leaders, and they are not alone, cannot see what is sitting at the table across from them. The chaotic nature of Trumpβs exercise of power is that nothing he says sticks, even when you get it in writing. They think they are looking at Toozeβs βroutine of civil lawfare,β but it is, in fact, autocracy.
Anyone who hopes that the independent monitor, as written into the agreement, will matter is fooling themselves. The history of autocratic power, as well as Trumpβs long history of disregarding his promises, signed or not, shows why. Naidu suggests the quickly broken oath made by James II at his coronation in 1685 to respect Parliamentβs authority is a relevant example. I hope so. James was deposed about four years later in the Glorious Revolution with little bloodshed (his army abandoned him), and he lived the rest of his life in Saint-Germain-en-Laye, a sort of Mar-a-Lago of his day, just outside Paris.
What should Columbia have done instead of aligning? The answer, quite clearly, is what Harvard has been trying to do. Although the oldest corporation in the western hemisphere is not exactly a bastion of free inquiry these days, it has seemed willing to fight to preserve its independence in a way that would serve the public interest. Harvard professor Ryan Enos, who rallied the Cambridge Common in April, is not optimistic about whatβs happening behind closed doors. If Harvard chooses to align, it will be a disaster for everyone except those who value a dollar today over all else.
Fighting a regime that is both kleptocratic and autocratic is a bad idea, one that will lead to terrible outcomes for the institution and, quite possibly, for Alan Garber and other leaders of the Harvard Corporation.
I wonder what thirteen-year-olds forming an impression of which colleges matter will think of what Columbia has done, and what Harvard decides to do.
The best way to support the writing I do here is to share.
Two bonus questions
Why is The Verge the only media organization writing about the other big news out of Columbia?
I first started reading The Verge when they employed James Vincent, a brilliant writer on AI whose tenure at the news organization started well before ChatGPT and ended last year. He is the author of Beyond Measure: The Hidden History of Measurement from Cubits to Quantum Constants, one of my favorite popular science books of the last ten years.
Since Vincent left, their AI coverage remains quite good, but The Verge is for people who are into technology and the tech industry. They donβt write much about technology and education. Which makes it weird that they are one of only three major news outlets covering the biggest higher education technology story of the summer, which not so coincidentally, involves Columbia University.
On June 24, Columbia University experienced "a technical outage that disrupted certain University IT systems" that they said was "caused by an unauthorized party with the apparent intent to disrupt our systems."
On July 1, Bloomberg reported that the disruption was the work of a hacktivist and verified that the "disruption" included the apparent theft of all admissions records from 2019 to 2024. Reporters communicated with the alleged data thief, who claimed to have stolen "approximately 460 gigabytes of extracted data detailing financial aid packages, employee pay, and at least 1.8 million Social Security numbers belonging to employees, applicants, students, and their family members." It appears the hacker also accessed, or tried to, the systems of the University of Minnesota, NYU, and two other universities.
The Vergeβs Elizabeth Lapatto, who has written two long-form pieces on the breach, asks a very good question: Why are other news organizations, including The Chronicle of Higher Education and Inside Higher Ed, not covering what may be the biggest privacy breach in history at an American university?
The silence is especially strange given Bloomberg's reporting that the hacker is a vigilante looking to expose Columbia's admissions practices as violating the Supreme Court's ruling against affirmative action in 2023. That is, this theft appears to be a political act, indirectly related to the Trump regime's pressure on Columbia that led to the βagreementβ this week.
As Lapatto points out, the exception to the blackout is the New York Times, which has been all over the story, taking the angle that what matters here is the contents of Zohran Mamdani's college application, which the hacker made public, and not the politically motivated attack on the computing systems of a beleaguered elite university that has been in the news all year. Weird, but understandable in light of the Timesβs habits and obsessions. It is, after all, the local paper of the financial capital of the world. Mamdami unsettles them in a way Trump does not. Like Claire Shipman, the Old Grey Lady is blinded by her trained incapacity.
The silence from other outlets is weirder. Inside Higher Ed, which is generally willing to rock more boats than the Chronicle, has written before about cybersecurity risks, including this excellent piece a year ago. Is it that summer is vacation time, so there is no one is around to write the story? Newsroom budget cuts mean they can't cover every little thing? Are investigators asking them to hold off?
Seriously asking here. What's going on?
What is a bigger problem for teachers than ChatGPT?
I rail against CSUβs decision to spend millions on ChatGPT during a budget crunch and argue that small, incremental projects that bring teachers and technologists together to explore small open models is a better way for universities to experiment with AI. But the truth is, ChatGPT Edu is not getting a lot of takers. But itβs not because Silicon Valley skeptics have the upper hand in academic technology budget meetings. The reason is that it's a hard sell to get a university to pay for something their students already get for free, and most faculty donβt even want. When they stop giving ChatGPT away for free, is when they will look to sign more enterprise agreements.
As Marc Watkins pointed out earlier this summer, ChatGPT is a sideshow compared to the AI features being added to the digital platforms that teachers and students already use. We should be talking about this as much as we are about ChatGPT, but that does not seem to be happening.
Here is his recent piece in the Chronicle providing good advice for teachers who find AI features waiting for them in the fall. And here he is writing about the problem on
earlier this summer:
Thanks for reading ππ ππ¨π . If you like the Logpodge format of shorter essays, here is another: Dinasaurs Donβt Do Digital.
Robbins is absolutely right about the (lack of) value of external consultants in a moment when no one has any real idea of what AI means for higher education. Why pay millions of dollars for fake certainty and dubious charts from people who know nothing about your campus and even less about AI than your own faculty and IT staff?
If you feel like paying someone to say this to you, please hire me as an external consultant. My hourly rates are quite reasonable, although I am terrible at the consultantβs primary job, which is to get hired again next year to deal with the same problems they were hired to solve the first time.
All your critiques are fair and welcome. I donβt think most people know how bad things are NOW. Improvement is going to take acknowledgement of the current awfulness first.
On admissions, I am always interested in the mindset of Arizona State - success is how many people you admit not how selective your admissions are