AI May, Paradoxically, Increase Demand for Higher Ed

There are scenarios where college demand increases in the face of AI

Please like, share, comment, and subscribe. It helps grow the newsletter and podcast without a financial contribution on your part. Anything is very much appreciated. And thank you, as always, for reading and listening.

---

---

About the Author

Jimmy Alfonso Licon is a philosophy professor at Arizona State University working on ignorance, ethics, cooperation and God. His forthcoming book, Better Not to Know: Why Knowing Less is Sometimes Best, is with Peter Lang Publishing. Before that, he taught at University of Maryland, Georgetown, and Towson University. He lives with his wife, a lawyer, at the foot of the Superstition Mountains. He also abides.

---

A decade ago, AI expert Geoffrey Hinton predicted that artificial intelligence would decimate radiology. Why employ expensive specialists when algorithms could read X-rays and CT scans faster and more accurately than humans? The logic appeared airtight at the time: AI substitutes for pattern recognition, radiology is pattern recognition, and so, therefore radiologists would go the way of elevator operators. And yet, a decade out, and the opposite has happened. Between 2012 and 2022, the number of radiologists in the United States increased by roughly 15 percent, and their median salaries rose faster than inflation. AI made radiologists more productive, shifted their work toward higher-stakes interpretation and consultation, and created demand for specialists who could integrate imaging AI into clinical workflows. The very technology predicted to eliminate the profession instead raised the premium on radiological expertise.

This outcome holds a lesson that extends far beyond medical imaging. Every few months, someone announces that AI will finish what the internet began and render higher education obsolete. Why pay tuition when a model can summarize Kant, draft business plans, debug code, and generate essays on demand? (Even though something similar could be said for the many, many free, high-quality college courses available online for the last decade or so). If AI substitutes for cognitive labor, and college primarily produces cognitive labor, then demand for college should fall. That inference is not crazy, but it should not be taken as inevitable either. The economic future of higher education turns on factors more complicated than substitution, and include complementarity, scarcity, signaling, and the structure of knowledge itself. And once those mechanisms are examined, it becomes plausible that AI will increase demand for certain forms of higher education rather than shrink it.

Frontier Knowledge Becomes Scarce When General Knowledge Becomes Cheap

Begin with a point about the nature of knowledge that Friedrich Hayek emphasized throughout his career. AI systems aggregate vast quantities of general information. They are extraordinary at synthesis across domains. But Hayek’s insight was that much of matters economically is local, tacit, and, contextual—something that is very hard to aggregate and automate, like an inside joke you ‘had to be there’ to appreciate. Markets coordinate this dispersed knowledge through prices precisely because no central authority can internalize it all.

Universities, at their best, are institutions that generate, curate, and transmit frontier knowledge and locally embedded expertise. Laboratories, research, and collaborative projects produce knowledge that does not yet exist in any training corpus. So, if and when AI dramatically lowers the cost of accessing what is already widely known, the effect is that the relative value of what is not widely known increases in relative terms.

This is a clear substitution effect, such that, if AI automates general knowledge, then returns shift toward what is difficult to encode and what requires situated judgment. And institutions that are organized around producing and validating such knowledge—research universities, advanced professional schools, specialized laboratories—may thus become more valuable. Consider how this worked in radiology. As diagnostic AI improved, the work of reading standard chest X-rays became partly automated. This allowed radiologists to focus on complex cases and communicating uncertain findings to clinicians under time pressure.

Powerful Tools Demand Sophisticated Users

A second mechanism involves credentialing. Public discourse often assumes AI (fully) substitutes for educated labor. However, the history of general purpose technologies at least suggests complementarity is just as, if not more, common. Electrification increased the marginal productivity of engineers and computers allowed finance professionals to focus on quantitative finance, risk modeling, and algorithmic trading. The more powerful a tool is, the more the risks of misusing it and the higher the value of the judgment of someone using it.

Firms adopt AI to, among other things, reduce costs and manage risk. And when AI introduces new risks like hallucinations, security vulnerabilities, regulatory violations, companies need highly trained professionals to supervise and validate outputs. The legal and reputational stakes often increase alongside automation such as in fields like medicine, finance, aviation, and law, where greater automation has historically been accompanied by more credentialing and oversight. Often regulatory and insurance structures require formal training. AI can thus increase demand for credentialed labor precisely because it raises the stakes of decision-making.

When Cheap Signals Flood the Market, Costly Signals Appreciate

A third mechanism involves signaling. Education has always served dual functions of both building human capital and screening ability and dispositions. Degrees signal do the work of signaling what graduates know and personality traits like persistence, conscientiousness, and baseline intelligence. This is why (many) employers rely on a college degree, even one where the major is irrelevant to the job, in part because the completion of a multi-year program under institutional constraints credibly reveals something about the person.

AI complicates this environment for simple reason that if generating polished essays, code, and research summaries becomes trivial, low-cost signals proliferate. And in such an environmental, employers face a harder sorting problem. When cheap signals flood the market, economic logic predicts a flight to costlier, harder-to-fake signals. A degree from a reputable institution—earned over years under structured evaluation—remains relatively expensive and difficult to counterfeit. Simply put, one would expect that if AI makes short-run competence easier to mimic, employers lean more heavily on long-run indicators of reliability. Not every institution benefits from this dynamic, but it may help high-trust universities, and hurts low-trust institutions.

The Differentiation Thesis

None of this denies that AI could very well upend higher education. If AI compresses the value of lecture delivery and standardized content transmission, those institutions distinguishing themselves only on those margins are vulnerable because what they offer is substitutable. The future may involve fewer institutions functioning primarily as content distributors.

The mistake in many claims about the demise of higher education due to AI is that higher education is treated as a homogenous good, and that simply isn’t so—it really never has been as anyone who has applied to different colleges and universities already knows. Here economic reasoning pushes toward differentiation with some segments of higher education perhaps shrinking, and others offering frontier research environments and credible costly signals expanding. In such a scenario, it is pretty economically straightforward that AI would intensify the premium on exactly what strong universities do best, namely reputational networks that coordinate labor markets, frontier knowledge production, and cultivated judgment under uncertainty.

Implications

So which institutions thrive? The vulnerable institutions here are likely to be those that compete primarily on credential accessibility or content delivery without deep research missions, strong reputational networks, or meaningful skill certification. Whereas, in contrast, the institutions of higher education positioned to benefit are likely to be those research intense universities and colleges producing frontier knowledge, professional schools tightly integrated with credentialing and liability regimes (law, medicine, engineering), and specialized programs training people to work with AI systems in high-stakes domains.

This post is not meant as a prediction—the history of tech prediction is a dumpster fire and only a fool engages in that—but instead a challenge to the lazy assumption that AI mechanically reduces demand for higher education. That assumption treats education as mere information transfer and ignores the economic structures around scarcity, signaling, liability, and coordination.

The arrival of AI may not collapse demand for higher education, but instead force it to reorganize itself to better meet demand of higher education in an AI-saturated world. And on several margins like Hayekian knowledge constraints, complementarity in production, signaling under information overload, and liability-driven credentialing, the case for increased demand in the higher education space may be what happens when you apply economic reasoning to a new technological. And if you don’t believe me, just ask the radiologists.