If large numbers of college-educated people find themselves unemployed or downwardly mobile, it could create a crisis of confidence for higher education.

The potential for artificial intelligence to revolutionize our economy is obvious enough, and even if we can’t predict exactly how and where it will lead us, we’ve trodden similar paths before with the telegraph, the steamboat, the railroad, the telephone, the automobile, the computer, the internet, and social media. Among other things, we can confidently predict that AI will go through stages similar to e-commerce and other new technologies: overinvestment in a wide variety of uses, followed by a bust in which a lot of people lose their shirts because they poured money into half-baked business models, followed by a period in which new behemoths arise. Some companies will be Amazon and Netflix, and some will be Pets.com, Netscape, and Blockbuster.

We can also foresee — and even see underway — how AI is changing a lot of other things besides just how businesses produce goods and services. AI can create fake images and news, it can revolutionize cheating in school, it is creating an arms race in the job market between AI-generated resumes and AI-staffed human resources resume-screening systems, etc. All of this has been unfolding just over the past three years.

What is much harder to describe, although we can feel them coming, are the social consequences of AI. In the news and information space, it’s hard to say which is worse: gullibility that allows lies to spread, or pervasive cynicism that prevents the truth from gaining traction (the answer is most likely both at once). In the world of learning, there are real hazards in creating a tool that is so good at faking education that actual humans don’t bother acquiring one. I don’t even want to touch on the pornographic uses and how that escalates the already excessive incentives to follow that path as a substitute for human intimacy, with all of its risks of rejection, heartbreak, and imperfection. On the other hand, if the pervasive artificiality of AI leads more people to mistrust written and video communication and give more face-to-face interactions a chance, that could be a good thing.

The political earthquake, however, could come from economic displacements that may be arriving fast and furious. They will be unlike what we have witnessed before. One of the most salient features is the threat to white-collar jobs, be they knowledge jobs or more clerical jobs. Historically, we have seen many examples of how new technologies destroy some jobs while creating new ones. The pattern has very often been that blue-collar jobs (artisans, tradesmen, factory workers, unskilled workers) get consumed rapidly, while new ones are created more slowly or unevenly, with unpleasant results for blue-collar workers who find themselves compelled to relocate, emigrate, or accept lower-status or lower-paid jobs. By contrast, the typical pattern for white-collar jobs has been the opposite: that new technologies that enhance productivity end up swiftly creating more jobs than the ones they destroy, and that the workers who get displaced tend to be well-situated to adjust. These aren’t ironclad rules, but historically, technological advances have worked out better for the more-educated classes than the less-educated classes and the people who work with their hands. Over and over, we’ve replaced muscle and fingers with mechanization.

This time will be different. AI’s threats to blue-collar jobs are much more limited; it will be far easier for machine thinking and learning to replace people who think, learn, write, compile, and analyze for a living. Now, in a moral sense, that’s neither better nor worse than displacing blue-collar workers; it’s just different. But in its social implications, it could be very different. Political revolutions have often been made, and led, by educated people who find themselves without the status and income that they expected or once had. Our political system has never really had to deal with this phenomenon on a large scale. It is also hard to predict exactly how it plays out for the two parties, given the increasing sorting of such people onto the left. It’s a particularly bad moment for the nation’s most prominent conservative speaker to campus audiences to be assassinated.

There’s also a more specific threat that will arise if large numbers of college-educated people find themselves unemployed or downwardly mobile, and that is a crisis of confidence for higher education. We all know the various arguments for a college education in terms of intellectual, moral, or social formation, but we also all know that the vast majority of people go to college primarily as an economic and social-status proposition: If you get a degree, it’s a ticket to a better life than if you didn’t have one. That bargain is why our society has accepted massive price increases, huge loan burdens, and colossal public subsidies for the colleges. The difficulty in turning down that bargain is how the colleges have succeeded at bundling along with their status-conveying role all manner of anti-social political indoctrination. People put up with a lot to get that degree. But what if the degree increasingly turns out to have been a lie? It’s one thing to live through an economic downturn in which the college-educated suffered with everyone else. But to live through the loss of the relative status of college graduates is bound to be especially radicalizing. The result for the university system, already finding itself locked in combat with Donald Trump and his working-class supporters, could be catastrophic.