We keep talking about the impact of AI in the classroom, but even more pressing is whether children have the security to learn to begin with. The future workforce needs human beings who are healthy, stable, and capable of imagining a life beyond survival.
If AI in the classroom isn’t on your radar yet, congratulations — you’re the last one in America. Because everyone else is asking: How much belongs in the classroom? How young is too young? Should kids learn to prompt AI, work with it, compete with it? Maybe all three?
Perhaps these questions are putting the robot ahead of the road. A more basic query, decidedly low-tech, is this: What exactly does it take for a child to actually learn?
As we race to crack the code on education in the age of AI, there’s an elephant in the room no classroom plan can prompt away: educational attainment is deeply tied to the socioeconomic conditions in which children are raised. So, if we want kids equipped to navigate an uncertain future, we’d better get cracking on building a society where every child has a genuine shot at success.
When it comes to education in America, the writing is on the wall even before the child even sets foot in a classroom.
It wasn’t supposed to be this way.
The Stalled Escalator
America built its public schools on the rather audacious idea that where you start out in life doesn’t have to determine where you’re headed. Instead of a luxury or a charitable endeavor, schooling, to many reformers, was envisioned as a great escalator by which children could ascend to better lives while making the nation more prosperous and democratic.
The Common School movement of the 1840s and ’50s greatly expanded access to free public elementary education, particularly in the North and Midwest. Then, from about 1910 to 1940, another revolution made high school — a proud American invention — a mass institution. The Morrill Acts of 1862 and 1890, followed decades later by the G.I. Bill, extended the ideal of mass education into colleges and universities, opening their doors to millions more Americans. In The Race Between Education and Technology, economists Claudia Goldin and Lawrence Katz argue that these movements, focused on building the nation’s human capital stock, helped turn education into a roaring engine of American economic growth and narrowed economic inequality, especially from roughly 1915 through 1980.
So how are we doing on that promise in the 21st century? Not so great, unfortunately.
A window is offered by the National Assessment of Educational Progress, which began in the early 1970s. For a time, the trend seemed encouraging: from 1971 to 2020, the average reading score for 9-year-olds rose from 208 to 220, while math scores climbed from 219 in 1973 to 241 in 2020. But underneath, the progress had begun to slow. Reading and math scores for 9-year-olds had largely leveled off by 2012, while scores for 13-year-olds were already slipping before the pandemic hit. With COVID, the bottom fell out: scores plunged after 2020, with some of those losses –especially for lower-performing and lower income kids — still not recovered. The pandemic didn’t create the underlying problems, but it appears to have magnified them amid the strains schools and families went through.
The story of who benefited from the gains, and who didn’t, is especially concerning. From the 1970s through the early 2000s, many American students were doing better, but that came with a troubling divide: a child’s chances of succeeding in school were increasingly tied to the circumstances into which they were born. As average scores rose, so did the gap between affluent and low-income students.
In a landmark 2011 study, Stanford sociologist Sean Reardon found out just how far that divide had grown. Among children born in 2001, the gap in math and reading scores between high- and low-income students was about 30 to 40 percent larger than it had been among children born 25 years earlier. The racial achievement gap had narrowed, but the class gap had grown, including within racial groups. Strikingly, the divide was already visible when children entered kindergarten. And the worst part? It hardly budged as they moved through school.
The educational escalator that whisked generations of Americans upward from the mid-19th century through the mid-20th century appears to have lurched to a halt. Today, what we see is more like a sputtering machine that locks children into their socioeconomic situation rather than raising them up.
In Our Kids, political scientist Robert Putnam argues that much of the inequality in educational outcomes is forged in things outside of school, like family life, neighborhoods, health conditions, and access to the extracurriculars that teach kids how to navigate the world. Basically, all the things where money makes a tremendous difference.
Putnam stresses that in schools serving families under financial strain, teachers and administrators spend much of their time dealing with the fallout of economic precarity: chronic absenteeism, drug use, disorder, and other disruptions that leave less time for teaching. His comparison of two Orange County high schools, affluent Troy High in Fullerton and Santa Ana High, makes the point starkly. The two schools spent roughly the same amount per student and had teachers with similar experience and qualifications. Yet their students had wildly different outcomes. The gap, he argues, couldn’t be explained simply by what schools did or how much they spent. He points instead to the advantages that affluent children often bring with them from home, like stronger study habits, organizational skills, parental engagement, and expectations about school. In Santa Ana, he found families contending with limited education, limited English, poverty, and the daily pressures of making ends meet. The school inherited some of the disadvantages that had piled up outside its walls.
Affluent children increasingly grow up wrapped in golden threads of enrichment—sports teams, music lessons, tutoring, fancy summer camps, and countless other experiences that build knowledge, confidence, relationships, and the habits schools reward.
Meanwhile, public schools have been feeling a financial squeeze for decades. It started back in the ‘80s when the Reagan administration cut federal school funding and dumped the financial burden onto states and local communities. Years later, the Great Recession hit, bringing a wave of budget cuts that tore apart the fabric of American schools and started putting basic opportunities behind a paywall. Suddenly, “pay-to-play” fees sidelined kids from lower-income families. While some parents could easily cough up the cash for a uniform or a tuba, for others, those price tags became major barriers to opportunity.
The result is a kind of two-track education system. The classroom may be public, but much of what happens around it is increasingly private. Some children arrive with years of accumulated advantages, like books, organized activities, stable routines, educated parents, and a network of adults who can help steer them through the system. Others show up weighed down by the accumulated burdens of scarcity. The school experience often simply cements the divide.
This mirrors the “dual economy” economists Peter Temin, Lance Taylor, and Servaas Storm all argue have taken shape in America. We have become a country increasingly split between a secure, a high-wage sector and a larger low-wage sector where work is precarious, opportunity is thinner, and the path upward is harder to find. As Temin put it, we’ve created, through polices ranging from tax cuts for the wealthy to eroded labor protections, “a structure that predetermines winners and losers.” He sees education as one of the only routes by which children can possibly break into the economy’s more prosperous side, but as we’ve seen, schools absorb the inequalities and divisions of the broader society.
That puts a question on the table: can schools help repair the damage of America’s runaway inequality? Some believe that a technology as powerful as AI could change the game, giving those who have the least a powerful new tool for catching up.
Machine Dreams
Before we dive into the AI question, it pays to revisit a little recent history.
When the internet hit schools, enthusiasts hailed it as the great equalizer: a world of information at every child’s fingertips. For its part, the Clinton/Gore White House waxed rhapsodic on the possibilities, explicitly framing internet access as a way to close the gap between wealthy and disadvantaged schools, giving inner city kids “the same universe of knowledge” as their suburban counterparts. A 1996 White House account quoted a teacher at a school serving a significant low-income population saying: “The Internet erases boundaries of age and class.”
The results have been far less transformative.
Technology spread rapidly, but simply handing students computers and internet access didn’t reliably improve learning. The OECD found little relationship between greater access to computers in school and student performance, while later research has documented a new problem: digital devices can turn into hard-to-resist distractions. As most parents know, young people on devices will multitask, check messages, browse unrelated sites, and divide their attention. Research indicates that technology — in schools and in homes — can amplify inequality, too: a report from Common Sense Media found that the children best equipped to use the technology are often the same ones who already have ample support and resources.
The city of Newark offers a cautionary tale of what can happen when Silicon Valley sets its sights on the classroom. In 2010, Cory Booker, then the mayor of Newark, and Mark Zuckerberg, founder and CEO of Facebook, unveiled a $200 million experiment to remake the city’s troubled schools into a shining national model, bringing the tech gospel of disruption, data, measurement, and rapid results to public education. Reforms included expanding charter schools, using data to evaluate teachers and students, and experimenting with cutting-edge approaches to technology and personalized learning.
The money poured in, and so did the consultants — more than $20 million from Zuckerberg’s gift and matching donations went to outside firms in the first two years, with some consultants bagging $1,000 a day. What didn’t come was the promised transformation. The reforms initially sent achievement growth downward. Five years in, independent researchers found gains in English but no significant improvement in math, with much of the English improvement attributable to shifts in enrollment rather than gains within existing schools. Zuckerberg eventually acknowledged that he had approached education with too little understanding of the communities involved.
In spite of this debacle, the Zuckerberg playbook remains familiar: when there’s an education problem, reach for a technological fix — more data, better software, personalized learning. But the causes of inequality aren’t hiding in the classroom waiting to be discovered by tech CEOs or their favorite politicians. They live in plain sight in the form of suppressed wages and stagnant paychecks, reduced bargaining power, unfair taxation, and inadequate public investment. Zuckerberg has pledged enormous sums to “promote equality,” but philanthropy is not the same as redistribution, and it has a funny way of leaving the underlying problem wealth and power largely untouched.
Solutions to inequality depend on the hard, unglamorous work of pushing policies like expanded Medicare and increased parental leave — and taxes on billionaires and their companies.
None of this is to say that the classroom doesn’t matter. It most certainly does. There is good evidence that what happens inside it can help narrow achievement gaps: Roland Fryer and his colleagues, for example, found that disadvantaged teenagers made substantial academic gains when tutors –the human variety — worked with them in small groups. Quelle surprise: Give a kid sustained attention from a skilled adult and the kid learns more. The challenge is asking teachers to perform this trick at industrial scale. In schools serving low-income communities, every minute spent calming a classroom or trying to get everyone back on task is a minute that can’t be spent tutoring.
Reports from Mississippi and Louisiana offer another useful reminder that instruction matters. Mississippi’s gains followed a wholesale improvement in literacy instruction, curriculum, and teacher training. Louisiana has taken a similar approach, combining research-backed curricula with teacher training and increasingly intensive tutoring. These cases show that when schools give teachers better tools, better training, and struggling students more targeted attention, achievement can move even in places where poverty remains stubbornly high. But there’s a caveat: these experiments haven’t made the underlying disparities in opportunity disappear.
This leaves many people wondering if AI can finally give schools a way to finally address those gaps at scale by putting individualized instruction within every student’s reach. The buzz suggests that new technology can tailor education to every child, mitigating, if not erasing, the disadvantages they bring with them into the classroom.
There is some evidence that AI tutors can help students learn. A UC Berkeley study found that learners who received ChatGPT-generated hints on math problems showed gains similar to those produced by human-written hints — although the study involved 274 adults recruited online rather than children sitting in classrooms.
A more direct 2025 study conducted in Turkey of nearly 1,000 high-school students found that AI could improve performance while students were using it. But again, a major catch: students who relied on an unrestricted AI system later performed worse when they had to work without it. A tutor designed to offer hints rather than answers avoided much of that decline, but the takeaway is that personalized AI for students is not a magic wand.
So far, it appears that the benefits of AI in the classroom are neither automatic nor evenly distributed, as a 2025 review by educational researcher Aziz Mimoudi and an OECD analysis of AI and educational equity both warn. Mimoudi, co-founder of the Institute of Education Sciences, sounds the alarm that “without safeguards, the same systems that promise to close learning gaps may reinforce them.” (And we don’t even know which versions of the technology may work, as RAND researchers note).
What we do know: a child with enough food and sleep, supportive adults, and time to think is in a very different position from one who is hungry, exhausted, distracted, or worried about what is happening at home. AI may give a rich and a poor child the same tutor, but it can never be a substitute for the social and economic infrastructure children need, and there’s plenty of reason to be cautious about running an uncontrolled experiment in the classroom when the technology is still new, the evidence is thin, and early warning signs — from the potential dangers of AI overreliance to diminished capacities — are already flashing.
Yet another reason for caution is the fact that there are mountains of money at stake. Big Tech is moving aggressively into classrooms, turning schools into a battleground for AI adoption. Google, Microsoft, OpenAI, and Anthropic are all pushing into education, offering schools, teachers, and students free or subsidized access to AI tools and training.
Microsoft, for example, offers Copilot Chat at no additional cost to eligible education users; OpenAI provides free ChatGPT access to verified U.S. K–12 educators and has expanded free access through district partnerships; and Anthropic has offered free or subsidized access through education partnerships and its Claude for Teachers program.
The companies say they want to help educators and prepare students for an AI-powered economy, but there is an obvious business logic: get a generation of students and teachers accustomed to your tools, and you may win customers for years to come. Just as Google and Microsoft helped make their software the infrastructure of American schools, they now have a chance to make AI a routine part of how students learn, teachers teach, and schools operate.
But they can’t make the elephant go away. An AI tutor can explain quadratic equations to a kid in the voice of a cartoon character, but it can’t give their parents jobs.
We tend to ask whether schools have enough money, whether teachers are well trained, or whether students have access to the latest gadgets. But education is a social institution embedded in an economic system, and schools can’t repair inequalities that are often being produced somewhere else.
The AI debate is really a debate about political economy. If AI is going to make things like human judgment, adaptability, collaboration, and the ability to deal with uncertainty more valuable, then we really have to look beyond the technology. Kids need the conditions to develop them in the form of stability, relationships, and adults who have the bandwidth to be there for them. They need affordable housing, decent healthcare, and caregivers with enough economic breathing room to actually be present in their lives.
In other words, America’s educational systems requires a strong middle class.
The narrative that AI will finally let us give every child a great education is, at best, only a modest part of the equation. And there’s a dangerous clause in the fine print: if AI’s gains flow disproportionately to capital owners and big corporate shareholders while everyone else faces greater insecurity, the technology will end up deepening the very inequalities schools are being asked to overcome.
Such problems require politics, not a chatbot.
