Whatever you think of AI’s potential upsides—and I believe they are significant—the most urgent questions around artificial intelligence in education have to do with minimizing harm. How do we keep students from using it to cheat on essays and assignments, even on in-class quizzes and tests, as a quarter of high schoolers copped to in a recent Fordham Institute survey? And how do we keep kids safe when they’re likely to be tempted to chat with bots about all kinds of personal issues?
But the most important questions related to AI and education have to do with the future of work. It’s never been particularly easy to predict the jobs of the future, but it’s arguably an order of magnitude harder now, given the uncertainty about which occupations may shrink or disappear amid the accelerating tech cognitive revolution. Especially for older students, we want schools to prepare them to make a living. Hence any reasonable guesses at future careers are super valuable.
So I was happy to learn about the new book Messy Jobs: The Work That AI Cannot Reach, by Luis Garicano, Jin Li, and Yanhui Wu, which was highly recommended by Tyler Cowen, the information omnivore, libertarian, and AI enthusiast, and has received positive reviews in The Atlantic, City Journal, and elsewhere.
It’s a winner that you should put atop your reading list. The first half, in particular, supplies a persuasive explanation of what AI is likely to do to the labor market. The three authors, all economists, expect neither a utopian future where sky-high growth rates leave us all enjoying a life of leisure on the beach, nor a dystopian world where nobody can get a job or earn an income. That’s because plenty of human work will remain, some of it (including in education) not directly boosting economic output.
It’s not just the obvious jobs that will be hard to automate, such as the skilled trades and the caring professions. It’s also much of knowledge work. We shouldn’t fear a white-collar job apocalypse, say this trio of economists.
They suggest that the folks in Silicon Valley and elsewhere who predict mass unemployment may have a good understanding of the technology they are building but don’t have a grip on what office workers actually do. Frankly, few of us really do. How many times have you thought about the jobs your college buddies work in and wondered, “What the heck do they do all day?” As a think-tank president, I’m sure my college buddies have these questions about me!
But labor economics includes scholars who study exactly these questions, including the “production function” of white-collar work. Which turns out to be a lot more complicated than we might assume.
The authors argue that the jobs that will not go away are “messy.” As they explain:
All knowledge work varies along one dimension: messiness. At one end of the spectrum, there is a single defined task to execute, such as helping clients file their taxes. You get the pay slips and expenses via email, use rules to fill out a form, and get a result. At the other end of the spectrum, there is an enormous bundle of complex tasks. Running a factory, or a family, involves many different tasks. They are very hard to specify in advance, and they are difficult to split from each other so that some can be automated autonomously. Many involve a lot of conflict.
Jobs that amount to a single routine task, or even a bundle of routine tasks, can and will be automated. Note that many such jobs have already been outsourced, such as call-center roles that have moved to India or the Philippines.
But many jobs are “tight bundles” that are hard to break apart and difficult to automate. And mostly what those jobs have in common is lots of dealing with humans, with all our messiness and neuroses, and with the complicated internal politics of large organizations.
If, therefore, you spend a lot of time sitting in meetings, working toward decisions with colleagues and clients, selling to customers, or solving problems, then good news: Your job will be hard to automate!
What does that mean for schools? Should we make kids do lots of group work and endure endless “meetings”? Some maybe, but that’s only part of what’s involved. The authors expect that the trends identified years ago by Harvard professor (and now dean) David Deming will still apply. They write that “a consistent pattern runs through decades of data: The labor market increasingly rewards people who combine analytical thinking with the ability to work well with other people.” Demming tracked American jobs in the 1980s, 90s, and 2000s and found that careers that scored high on both “cognitive skill” and “social skill demands”:
…grew by roughly 12 percentage points over three decades. Jobs that were cognitively demanding but low on social interaction (number crunching) shrank by about 3 percentage points. The demand increase was not for “soft skills” but for people who could both think and coordinate.
That might sound like common sense, but it’s regularly overlooked. Too often, I hear “durable skills” advocates make it sound like people skills are all that matter, as if AI will replace everything else. But to use AI effectively—especially to spot errors or misjudgments—we need people who possess deep domain expertise. That means fully developing students’ cognitive capacities and helping them become experts in something. (It’s akin to the necessity of knowing stuff in order to learn more stuff.)
Importantly, bear in mind that both cognitive capacity and people skills are also super important when it comes to participating in democracy, i.e., to competent citizenship.
The early career pipeline
The good news from Messy Jobs, then, is that it’s unlikely that all the office jobs will disappear. But there’s bad news, too: Concerns about early career pipelines are real, legitimate, and hard to solve. The authors explain that many white-collar professions have used an apprenticeship model, whereby consulting firms, law firms, or hospitals charge much more for the labor of their junior workers than they pay them. In return for reaping much of that financial value, they train those junior workers to become senior workers.
But if AI can do many of the junior-level tasks as well or better than junior workers, this model breaks down. It’s hard for firms to charge dearly for the labor of their junior employees. Which means they can’t afford to invest in their training and development.
The authors don’t offer much in terms of solutions other than to point out that AIs still need a lot of spot-checking and oversight, and that’s something junior employees could learn to do reasonably well. But this is a real problem!
The future of teaching
I close on a positive note. Even if you disagree with Garicano and colleagues about office jobs, other careers are clearly going to survive the AI challenge: those in the caring professions. And guess what? America’s K–12 education system is one of the largest employers in that field. Unless you think AIs are set to become very effective at teaching most students (which I do not, in no small part due to the motivation problem), we will still need humans to help students learn, grow, and develop.
If college graduates face fewer options for early-career jobs, our schools should be well positioned to recruit stronger candidates, akin to what can happen during a recession when the overall quality of teaching applicants increases markedly. With less competition from consulting shops and law firms, Teach For America seems well-positioned for a resurgence, as do other means of drawing talent into classrooms.
And if AI does boost economic growth, that should yield the revenue to pay teachers and other education professionals reasonably well, especially if we choose to prioritize quality over quantity. As the authors of the book write, Baumol’s “cost disease” (whereby labor-intensive parts of the economy struggle to keep up with rising productivity overall) can become Baumol’s blessing, as well-paid human-facing jobs survive the challenge from AI.
—
None of this argues for complacency. Artificial intelligence is already shaking up the world of work, and educators need to attend closely to the shifts to prepare today’s students for the world they will inhabit. A clear-eyed view of the world, however, can help us make better decisions, and Messy Jobs is as clear a crystal ball as we’re likely to find today.
