I was out last week and didn't get an issue out, which means there's two weeks of higher ed news to choose from for this one. The week's clearest story, though, is one that's been building for months and finally got the long-form treatment I was waiting for. The CSU system's $16.9 million OpenAI deal, the largest of its kind in the country, picked up serious coverage from both the New York Times Magazine and Inside Higher Ed in the last two weeks, and it dropped at exactly the moment Inside Higher Ed published new student survey data on what students actually want from their institutions when it comes to AI. The two stories don't read the same way side by side as they do separately, so I'm spending the whole issue on what the gap between them tells us.  

Accreditation Overhaul 

From Why Students Aren’t All In on AI—And What They Want From Colleges | Inside Higher Ed 

Results from the most recent New Student Voice survey find that while students are okay with AI as a learning tool, they have concerns about dependence and uneven institutional responses.   

Our Thoughts  

Last fall, students admitted to San Jose State got a welcome video from their president, congratulating them on getting in and telling them they'd picked a campus where they'd get to work with the technologies remaking the world. Sounds pretty typical for any institution welcoming a new class of students, right? The catch—she didn't say any of it. The president in that video was an AI avatar built in her likeness, sent out to deliver a message she never recorded. So far, this paints a pretty good picture of how many higher education institutions have decided to answer the AI question: from the top down, at scale, and in a form that makes for a tidy press release. “Look everyone! We’re using AI and preparing students for the future.” Except, when you ask students what they actually want from their colleges on AI, nothing they name sounds anything like an avatar or a multimillion-dollar deal with an AI company. What they’re asking for is much less flashy and a lot harder to announce in a way that gets you on the front page of a major newspaper, which is probably why almost no one is building it. 

That welcome video is one piece of San Jose State's AI Everywhere strategy, which sits inside a much larger bet across the California State University system—a $16.9 million deal with OpenAI that handed a ChatGPT license to all half a million students, faculty, and staff across its 22 campuses and let the system brand itself the nation's first AI-powered public university. CSU is the largest public system in the country, educating roughly one in ten workers in California, so its bet on AI is more reasonable than my avatar example may seem. A lot of CSU students are the first in their families to attend college, may be first-generation Americans, and many are working their way through school while AI is steadily rearranging the job market they want to enter when they graduate. For a system carrying that weight, the move to get ahead of AI and try to put that technology into students’ hands before they enter the workforce is easy to understand and hard to begrudge.  

However, students themselves are not asking for what CSU bought. When Inside Higher Ed surveyed them nationally this spring, only 7 percent described themselves as all in on AI, and the most common stance, by a comfortable margin, was worry about growing too dependent on the tools. Just one in ten said their institution was handling AI's arrival well, and the largest single group called the response they were getting inconsistent from one course to the next. Although there are numerous stories about students using AI to complete coursework, many are still interested in learning with assignments built to sharpen their own judgment rather than their skill at working a chatbot. Simply put, students aren’t asking to be trained as AI users, but instead they're asking to be developed as thinkers who happen to have new instruments in front of them. Institutions keep reaching for the first because it is so much easier to put into operation than the second. 

Of course, it’s not just students and institutions involved in this process. OpenAI is the other party in this story, one that has a market incentive to continue growing and demonstrating proof of large-scale adoption. Public university systems are excellent customers for that kind of proof because we know that when higher education institutions decide on a vendor, it takes a dozen committee meetings to change that in the future. OpenAI is already using the CSU deal as the template for similar agreements in Greece, Estonia, and the UAE, and smaller versions of it have been signed at Arizona State, the University of South Carolina, and the University of Colorado. They even run a campus ambassador program at CSU that pays students to host events and show their peers how to use ChatGPT for things like dressing up a LinkedIn profile or studying for exams. Anthropic and Adobe run similar programs. A company with that kind of capital behind it is going to do what companies with that kind of capital have always done, which is find the largest credible customer it can and turn that customer into a marketing asset. One of the business professors quoted in the New York Times piece said it plainly: these companies need to show revenue, and there is no better place to do it than a public school system. 

So, we get a deal that gives both parties what they want. OpenAI gets a 500,000-seat reference customer while CSU gets to brand itself the nation’s first AI-powered public university. The problem is that the students, whose names appear on every justification for the deal, are not sure they want this much AI at their institutions. They’re worried about dependence with only about one in ten believing their institution is handling AI well, but their interests didn’t shape the deal as we can see from the thousands of empty licenses that no student bothered to activate because the student demand wasn’t there to justify the deal. I wrote earlier this spring about ASU's Atom course builder, which the institution stitched together from its own faculty's materials without telling them. ASU and CSU look like two different stories, but they aren't. ASU built its version in secret; CSU bought its version in the open. Neither one started from a question a student or a faculty member was asking. 

CSU has an interesting history. The 1960 California Master Plan, designed by Clark Kerr, established the system to educate the state's skilled workers, and it did so with an explicit second purpose alongside the workforce one: ground those workers well enough in critical thinking and the humanities that they'd be able to push back on their own instrumentalization once they entered the economy. Now, CSU is spending tens of millions of dollars on a tool whose business model runs on exactly the instrumentalization its founding mandate was meant to prepare students to resist. To be completely clear—I don't think the people running CSU see it that way, and I don't think they're acting cynically. I think some are probably genuinely trying to prepare students for the world of AI-supported work, but nevertheless, the contradiction is there in their deal with OpenAI.   

Buried in the NY Times story though, there’s an example of how AI could be used to help CSU live up to its second purpose. A professor named Nik Janos built a handful of course-specific chatbots for his social theory class, including one trained on Marx, and used them as a way for students to interrogate the ideas they were reading. A 38-year-old returning student in that class described talking with the bot as opening a layer of understanding she hadn't gotten to on her own, and she said it shifted how she thought about the worth of her own labor and how she talks to her young son about his. That's the kind of outcome the multimillion-dollar contract with OpenAI was supposed to make possible across an entire system. One professor did it, though, in one class, because the chatbot he built started from a pedagogical question he had about his own students. There was no vendor on the other side of that decision pushing him toward a particular outcome, and there was no chancellor's office shaping the design to maximize what could be announced. The work was sized by the people it was for. That's the difference, and it isn't a difference of budget or technology. It's a difference of who the work was built to serve.  

So, what does any of this mean for the rest of us who aren't running 500,000-student systems? The lesson isn't that institutions should walk away from AI; that ship sailed before most of us were even on the docks. The lesson is that the demand side of this question is the only side that's actually about students, and it's the side almost nobody is starting on, because it isn't where the pressure is coming from. A department that gets its faculty to agree on consistent expectations for AI across its courses has done more for its students than a campuswide procurement contract will, even though it can't be announced and nobody outside the building will know it happened. Faculty who rebuild their assignments so that student judgment sits at the center, rather than designing assignments to chase AI out of the room, have done more than a graduation requirement no one can meaningfully assess. None of that work shows up in a press release. None of it makes the front page of a major newspaper. It does, however, start from the people it's supposed to serve, and right now, that's the rarest thing in this whole conversation. 

Sparks

• OPINION: If higher education wants to rebuild public trust, start with making college affordable (The Hechinger Report) - John B. King Jr., the chancellor of the State University of New York and former secretary of education argues that college affordability extends beyond just the cost of tuition. I know I’m probably starting to sound like a broken record, but public trust and cost are intertwined even though education is more expensive than many people want to believe.
• Tuition discount rate reaches 57% for private nonprofits, NACUBO says (Higher Ed Dive) - The tuition discount rate is up, and net tuition revenue has fallen. Although I just talked about cost above, I’m also not sure how sustainable these practices are long term.
• 3 Takeaways From Spring Enrollment Data (Insider Higher Ed) - A decline in master’s program candidates but growth in health professions and engineering. Depending on what happens with graduate loan caps, I think graduate enrollment may be in for a rough few years.