The Library Card Fallacy
Everyone is writing the obituary for higher education. I am more interested in the architecture of what comes next.
There is a scene in Good Will Hunting that critics love to quote at each other. Will tells the Harvard kid in the bar he could have gotten the same education for a dollar fifty in late charges at the public library. The line lands because it flatters something we want to believe. That knowledge is the great equalizer. That the building, the brand, the seal on the diploma are markups on a product anyone could acquire for free if they were willing to put in the work.
People are saying the same thing now about AI. The library card has been replaced by a chat window. Why pay for a degree when the entire corpus of human knowledge is one prompt away.
It is not the same argument. And the difference is the whole point.
The library had provenance. Someone vetted what got on the shelf. An editor, a publisher, a peer reviewer, a librarian making collection decisions, a chain of custody you could trace and challenge. Imperfect, biased, gatekept in ways worth criticizing. But the books in the building had been touched by hands that were accountable for putting them there. When Will told that kid he could read his way to an education, he was making a claim that depended on the shelf having been curated by people whose job it was to curate it.
AI has no shelf. AI is a confidence machine producing fluent prose at a speed no human can fact check in real time, drawing on a training corpus that mixes the canon and the garbage in the same breath, with no way for the user to tell which is which from inside the conversation. The chat window is not the library card. The chat window is what you would get if someone burned the library down and rebuilt it overnight out of every book, every tabloid, every comment section, every confidently wrong Reddit post, and then handed you a polite tour guide who has never once admitted not knowing something.
That is the upgrade we are being sold.
And here is the part that matters. Even when the library was curated, it was never enough. Provenance is necessary. Provenance is not sufficient. If you do not know what you do not know, you do not know what to search for. You do not know how to sequence what you learn so the second thing builds on the first. You do not know which expert is credible and which is performing credibility. You do not know when you are wrong, and you do not know how to find out. Will Hunting could read his way to an education because Will Hunting was a once in a generation autodidact in a movie. Most people are not Will Hunting. Most people need a designed path through the material, a human who has walked it before, and a credential at the end that means something to someone who was not in the room.
Access was never the bottleneck. Discernment was. Application was. Sequencing was. The library card was a permission slip, not an education. The chat window is a permission slip without a librarian.
And the entire higher education sector has spent the last three years arguing about which permission slip is more dangerous instead of asking the only question that matters.
What do we do now?
Enough Autopsies
It is open season on higher education. Every week brings a new piece arguing the degree is dead, the model is broken, the bubble is bursting, the kids are not coming, the boomers are retiring, the bills are coming due, the ROI is negative, the credential is hollow. Most of it is at least partially true. None of it is interesting anymore.
I am tired of the autopsy. I am interested in the architecture.
Here is what the pile on misses. The need higher education was invented to fill has not gone away. The world still needs people who can think, discern, apply, lead, build, heal, teach, govern, and live as citizens of a place. It needs more of those people, not fewer. The fact that current institutions are doing a worse and worse job of producing them is not a reason to celebrate the collapse. It is a reason to design what replaces the parts that are not working.
So let me do the thing MOST ARE NOT DOING. Let me tell you what it could look like.
What A Modern Institution Is Actually For
An institution that is not producing primary knowledge is not in the knowledge business. It is in the design business. Its job is to take knowledge that is now everywhere, much of it free, much of it suspect, much of it produced by machines that do not know they are wrong, and turn it into something a human being can use to become more capable.
That is the work. It has always been the work. We just had the luxury of pretending the work was the lecture, because for a long time the lecture was the only place the knowledge lived. The luxury is gone. The work remains, and it is harder than it used to be, not easier.
The Ivies and the elite R1s produce primary knowledge. They have laboratories, research budgets, and graduate students whose job is to push the edge. They should keep doing that. They are a tiny fraction of the sector and they were never the right model for the rest of it. The other four thousand American colleges and universities are teaching institutions. Their value is in what they design, not what they discover. The institutions that admit this and build around it will thrive. The institutions that keep cosplaying as miniature Harvards will not.
The Five Role Stack
Here is what a teaching institution actually needs in the room when it builds a learning experience. Five roles, working together, and one of them is not human.
The knowledge expert. Still essential. Not the performer of the lecture, the source of intellectual integrity. They know what is true, what is contested, what is current, and what a learner cannot be allowed to skip past. In the new model they have a second job. Telling the AI what good looks like in their field, where it tends to be wrong, and what a learner should never accept from it. Without that, AI generated material drifts into plausible nonsense, and the expert is the only person in the room who can tell.
The learning designer. The person who knows how human beings actually acquire skill. Sequencing, scaffolding, retrieval, transfer. This is a discipline, not a part time assignment for a faculty member with a Friday afternoon free. The learning designer’s hardest new job is deciding where AI belongs in the loop and where it does not. AI as a tutor that asks the learner the next right question is a miracle. AI as an answer machine that lets the learner skip the cognitive work that produces actual learning is the opposite of education. Knowing which is which is the discipline.
The design designer. The person who understands the medium itself. How a screen works. How attention works. How an AI tutor should be scoped so it teaches instead of answers. How to build an interface where the learner argues with the AI, gets corrected by it, and learns to argue back. This role barely exists in higher education and it is the most important one MANY are not hiring for.
The technology stack. Not the thing the IT committee buys. The substrate the others design within. Data, platforms, infrastructure, the systems the learner lives inside. Institutions that treat technology as something they procure rather than something they co design will lose to the ones that treat it as a partner.
AI itself. Named as its own role, because pretending it is just a tool understates what it does. AI drafts material the experts refine. It tutors the learner at three in the morning. It generates practice problems that adapt to where the learner actually is. It analyzes patterns across thousands of learners and tells the designers what is working. It sits in on the work the way a brilliant, tireless, occasionally hallucinating junior colleague would. Pretending it can replace the other roles overstates it (At least for now). The honest description is partner, supervised closely, with the humans accountable for the output. An institution that has not figured out how to put AI in the room as a partner is an institution still arguing about whether to let the learners use it. Those are not the same conversation.
When the five roles work together, the output is not a course. It is a credential a learner can defend in the world. The graduate is not someone who sat through the right number of lectures. The graduate is someone who can discern, apply, adapt, and work alongside AI without being replaced by it.
What The Graduate Has To Be
The whole point of redesigning the experience is the human being walking out the other side. And the only useful question to ask of any institution claiming to redesign itself is whether the person on the far end is more capable of operating in the world they are actually entering.
That world does not need more people who can recite. It needs people who can discern. It needs cultural competence, because the workforce is global, the customers are global, and the problems are global. It needs environmental awareness, because every industry, not just the obvious ones, is being reshaped by climate and resource constraints and the long bill from a century of externalized costs. It needs workforce readiness, which is not the same as being trained for one job, it is the capacity to keep learning across a career that will span fifty years and a dozen industries that do not yet exist. And it needs citizens. Not in the narrow legal sense. In the older sense. People who can participate in the institutions of a society and recognize their obligations to people who are not them.
A learner processed through a curriculum optimized for content delivery does not come out the other end as that person. A learner designed for, by a team that takes the five roles seriously, has a much better chance.
What Is In The Way
I am going to name a few things directly because the rest of the sector will not.
Some of the historical (former) regional accreditors are well intentioned but miss the mark. The framework still rewards inputs. Seat time. Library volumes counted as if a library is a stack of books in a building. Thanks you NECHE for removing that vestige as a Standard). Governance processes that require eighteen months to make a decision the world will have moved past in six. Accreditors will tell you they support innovation. The form they make institutions fill out tells a different story. Until the framework measures what graduates can do, accreditation unintentionally creates a moat protecting the institutions that least want to change. I applaud those rethinking their approach!
Faculty senates have to decide what they are. A faculty senate that stewards academic integrity and the quality of the learning experience is one of the most important bodies on a campus. A faculty senate that uses shared governance as a procedural weapon to slow every change until the change is no longer possible is something else. The learners can tell the difference. The market can tell. Pretending otherwise is not protecting the academy. It is hastening its decline.
The rankings industry has corrupted the strategic conversation for thirty years and the sector has let it. Institutions optimize for selectivity, which means they brag about the learners they reject. They optimize for spending per student, which means they are rewarded for being expensive. They optimize for peer reputation, which is a closed loop in which the same hundred presidents rate each other based on what they read about each other in the same magazine that produces the rankings. None of these metrics measure whether a graduate can do anything. The institutions still playing that game are not strategic. They are nostalgic.
The politicization theater. Higher education is being conscripted into a political war it cannot win, and institutions letting themselves be conscripted are losing in both directions at once. From one side, executive orders, state laws, and funding threats aimed at forcing compliance with a prescribed view. From the other, sustained pressure to make institutions instruments of activism, to take official positions on every issue, and to measure faculty and learners by ideological alignment instead of intellectual capability. Neither side is interested in the learner. Both sides are interested in the institution as a piece of political real estate worth occupying. The thing the institutions are actually for is producing a graduate who is resilient, culturally competent, and can think. Someone who can hold two contradictory ideas in their head long enough to evaluate them. Someone who knows the difference between a fact, an interpretation, and an opinion. That graduate is not produced by an institution that has picked a side. Refusing the conscription is not neutrality or centrism. It is the harder thing. It is saying the institution exists to develop human capability, not to be a weapon in someone else’s argument.
The AI panic. I write a lot about this in other posts, so I will keep it brief. The institutions that have responded to AI primarily as a cheating problem have told you everything you need to know about how they understand their own product. If your assessment can be defeated by a chat window, the assessment was never measuring what you said it was measuring. The honest response is not a detection arms race. It is a redesign of what you assess and how. AI did not break the model. AI exposed it.
The Build
So here is the part that most are not writing, blogging, or reporting.
The future of higher education is not the death of the institution. It is the redesign of the institution around what learners actually need and what AI actually changes. That redesign has a shape. Five roles in the room. A graduate measured by what they can do. A credential that means something because the work behind it is real. A partnership with AI that is honest about what AI is good for and what it is not. An accreditation framework that catches up or gets routed around. A faculty culture that decides whether it is in the protection business or the building business.
This is not a hot take. These are operating instructions. The institutions that pick them up are going to be standing in twenty years. They will be smaller in some cases, larger in others, different in almost every case. They will have credentials that employers actually understand. They will have graduates who can walk into a room with an AI agent on one side and a difficult problem on the other and contribute something neither of them could produce alone.
They will be teaching institutions that are proud of being teaching institutions, instead of apologizing for not being research or exclusive ones!
The library card was a promise that the building had been curated. The chat window is a promise that the answer will be confident. Neither one is an education. The education is what an institution does with the now abundant raw material to produce a resilient, culturally competent, career ready, environmentally aware citizen who can think.
That is still the work. It has always been the work. The institutions that name it, staff for it, design for it, and measure themselves by it are the ones who get to write the next chapter.
The rest are writing their own obituaries and pretending it is somebody else’s fault.
Cannot agree more with this: “ If you do not know what you do not know, you do not know what to search for. You do not know how to sequence what you learn so the second thing builds on the first.” I think this is why we still need better education in the AI era than any other time!