This is Naked Capitalism fundraising week. 999 donors have already supported our initiatives to fight against corruption and harmful practices, particularly within the financial sector. We invite you to contribute through our donation page. Here, you can find various ways to donate, including by check, credit card, debit card, PayPal, Clover, or Wise. For more details, read about the purpose of this fundraiser, what we’ve achieved in the past year, and our current target: providing bonuses for our exceptional writers.
Welcome, readers. This article addresses a crucial issue: whether the integration of AI in education truly enhances learning and skill development. Unfortunately, it fails to adequately tackle this topic, not only due to its complexity but also because it adopts an overly accommodating stance towards AI’s role in schools.
For example, it references a study claiming that children who received hints from AI while solving math problems outperformed those who received guidance from human teachers.
What kind of misguided study is that?!?
Since when has completing a task with assistance been deemed an appropriate measure of performance? Students need to acquire problem-solving skills, not merely rely on prompts. The idea that finishing work with a form of crutch is sufficient is absurd. Such reliance could leave students helpless in situations where technology is unavailable—what will they do if the power goes out or there’s no internet access?
Furthermore, we lack information about the qualifications and teaching styles of the instructors evaluated against the AI.
Lastly, we have no assurance that the study wasn’t designed in a biased manner, akin to a “chocolate helps you lose weight” type of investigation, where students using AI had their performance assessed on a variety of metrics, many of which could ensure one would yield favorable results. As noted in a CBS article:
“The world is just drowning in all this pseudoscience” surrounding diet and nutrition, said science journalist John Bohannon, a collaborator on the project. He and the filmmakers devised a scheme to demonstrate how easily misleading science can create sensational headlines. They constructed a fictitious organization, the Institute of Diet and Health, recruited a doctor and analyst, and compensated participants in a modest trial to test chocolate’s effects. Bohannon utilized his media expertise to get the results published and promoted.
The experiment was conducted like a genuine clinical trial: one-third of subjects were assigned a low-carb diet, another third followed the same diet yet included a daily 1.5-ounce dark chocolate bar, while a control group retained their normal diets.
Bohannon anticipates your thoughts: “The study did reveal accelerated weight loss in the chocolate group—shouldn’t we trust it? Isn’t that how science works?”
Yet he explains:
“Here’s a dirty little science secret: If you measure a large number of things about a small number of people, you’re almost guaranteed a ‘statistically significant’ result. Our study involved 18 different measurements—weight, cholesterol, sodium, blood protein levels, sleep quality, well-being, etc.—across just 15 participants. (One was eliminated.) That design is a recipe for false positives… We didn’t know what might emerge—the headline could have claimed that chocolate improves sleep or lowers blood pressure—but we knew the odds of at least one ‘statistically significant’ result were quite high.”
The fact that Silicon Valley millionaires send their children to schools with strict limits on tech usage until their teenage years is telling. They understand that excessive screen time can hurt attention spans and memory, among other issues.
Increasingly, evidence reveals that widespread AI use in schools leads to inadequate mastery of core skills. Numerous studies have shown that students who performed well with AI assistance struggled in its absence.
Moreover, AI tends to demotivate students. Many narratives exist about students refusing to engage in studying, arguing that AI can provide the answers, so why should they put in the effort?
This attitude is also evident among medical students. A recent account from a physician, known as IM Doc, illustrates this challenge. Each summer, he mentors students from a prestigious graduate program:
“So far this summer—four in July, three in August—not one has been able to read through a paper and synthesize the information. They rely entirely on AI on their phones for answers. I’ve given up.”
“Among the seven students, the two worst performers are the two Ivy League students—who shouldn’t be anywhere near patient care. It’s disheartening—they are good kids and likely quite bright—yet they have been poorly served by their academic environments.”
“I was astounded when one informed me they’ll graduate in May with over $690,000 in student loan debt. The mind boggles. There’s so much debt sloshing around the country—it’s hard to believe it will ever be repaid. We are in trouble.”
Lynn Parramore highlights a persistent issue in education: children from disadvantaged families tend to perform worse academically, and this gap continues to grow. She, along with others, seems to hope—perhaps too optimistically—that AI can help reverse this trend.
While there is a class issue here, it differs from the one she emphasizes. The elite are actively seeking to utilize AI to either lower skill requirements or completely replace numerous white-collar jobs, from call center operators to more advanced roles. Parents from middle and upper-middle-class backgrounds often worry quietly about how to ensure their children secure viable education and training for stable employment. AI, at the very least, continues to diminish the number of entry-level opportunities that allow aspiring professionals to learn their trades. Reports from IM Doc indicate that even in formerly stable and rewarding careers, students’ heavy reliance on AI makes them easily replaceable.
This trajectory implies a significant shrinkage of the technocratic elite, leading to a reduced group of skilled individuals possessing critical thinking and analytical skills, whether formally recognized or simply occupying key positions in major organizations.
This notion assumes the high error rates associated with AI do not precipitate broader systemic failures.
By Lynn Parramore, Senior Research Analyst at the Institute for New Economic Thinking. Originally published at the Institute for New Economic Thinking website
If you haven’t yet considered the role of AI in education, congratulations—you’re practically the only one left in America. Everyone else is debating: How much AI should be integrated into classrooms? At what age is too early? Should children learn to engage with AI, collaborate with it, or compete against it? Perhaps all three?
But perhaps these discussions are jumping ahead. A more fundamental question, decidedly not tech-driven, is this: What does it actually take for a child to learn?
As we strive to decode the educational landscape in the age of AI, we cannot overlook a significant reality: educational success is inseparably linked to the socioeconomic conditions in which a child is raised. If we truly want to equip children to navigate an uncertain future, we must work diligently towards creating a society where every child has a fair chance at success.
In the context of American education, the signs of struggle are evident long before a child steps into a classroom.
It wasn’t meant to be this way.
The Stalled Escalator
America’s public education system was founded on the radical idea that one’s starting point in life should not dictate their future. For many reformers, schooling was envisioned not as a privilege or charitable act but as a societal escalator for children to rise to better lives while also contributing to the nation’s prosperity and democracy.
The Common School movement of the 1840s and ’50s significantly expanded free public elementary education access, particularly in the North and Midwest. Subsequently, between 1910 and 1940, another revolution institutionalized high school education across the nation. The Morrill Acts of 1862 and 1890, followed much later by the G.I. Bill, further democratized education by opening colleges and universities to millions of Americans. Economists Claudia Goldin and Lawrence Katz argue in The Race Between Education and Technology that these movements, focused on enhancing the nation’s human capital, transformed education into a catalyst for American economic growth and helped reduce inequality, especially from around 1915 to 1980.
So, how are we faring in relation to that promise in the 21st century? The results are disappointing, to say the least.
Insights from the National Assessment of Educational Progress, established in the early 1970s, offer a window into these outcomes. For a time, the trend appeared hopeful: between 1971 and 2020, the average reading scores for 9-year-olds rose from 208 to 220, while math scores increased from 219 in 1973 to 241 in 2020. However, the pace of progress began to stall, with reading and math scores for 9-year-olds plateauing by 2012 and declines noted for 13-year-olds even before the pandemic struck. The global COVID pandemic exacerbated these underlying issues, resulting in significant declines in academic performance, especially among lower-performing and lower-income children, with many yet to recover from these setbacks. While the pandemic did not instigate these problems, it extensively highlighted the challenges faced by schools and families.
Equally concerning is the reality of who has benefitted from the gains—and who hasn’t. From the 1970s through the early 2000s, many American students made strides, yet this progress was accompanied by a troubling divide: a child’s chances of academic success increasingly correlated with their socioeconomic background. As average scores improved, the gap between affluent and low-income students widened.
A landmark 2011 study by Stanford sociologist Sean Reardon quantified this divide. Among children born in 2001, the gap in math and reading scores between the highest- and lowest-income students was about 30 to 40 percent larger than among children born 25 years earlier. The racial achievement gap had contracted, but the class gap had widened, even within racial groups. Alarmingly, this divide was already apparent when children entered kindergarten and barely changed as they continued their education.
The educational escalator that once lifted generations of Americans from the mid-19th to mid-20th centuries seems to have come to a grinding halt. Currently, we face an educational system more akin to a stuttering mechanism that keeps children confined to their socioeconomic status rather than elevating them.
Political scientist Robert Putnam argues in Our Kids that much of the inequality in educational outcomes is established outside the classroom—through family dynamics, neighborhood influences, health conditions, and access to extracurricular opportunities that teach children essential life skills. Essentially, all the factors where financial resources play a critical role.
According to Putnam, in schools serving economically disadvantaged families, teachers and administrators often spend their time addressing the consequences of economic instability: chronic absenteeism, substance abuse, disorder, and other disruptions that ultimately detract from teaching. He starkly contrasts two high schools in Orange County: affluent Troy High in Fullerton and Santa Ana High. Despite both facilities spending roughly the same per pupil and employing equally experienced and qualified teachers, their student outcomes diverge significantly. He asserts that the discrepancy cannot be attributed solely to what schools implement or their financial investment. Instead, he highlights the advantages that affluent students often bring from home—such as strong study habits, organizational skills, engaged parents, and high expectations for educational achievement. In contrast, families in Santa Ana face educational limitations, language barriers, poverty, and constant struggles to make ends meet. The school itself bears some of the repercussions from these accumulated external challenges.
Wealthier students frequently benefit from a wealth of enriching experiences—sports teams, music lessons, tutoring, and elaborate summer camps—that enhance their knowledge, instill confidence, foster relationships, and cultivate habits that schools reward.
Meanwhile, public education has experienced longstanding financial pressures. Beginning in the 1980s, federal budget cuts initiated by the Reagan administration shifted the financial burden to states and local communities. The subsequent Great Recession brought further cuts, dismantling the foundation of American schools and transforming basic opportunities into exclusive offerings. The emergence of “pay-to-play” fees sidelined many lower-income children. While some families can effortlessly absorb costs for uniforms or musical instruments, such expenses become significant barriers for others.
This scenario results in a dual education system. Despite classrooms being public, many additional experiences increasingly occur in private settings. Some children arrive equipped with years of advantages—books, structured activities, stable routines, educated parents, and a network of caring adults guiding them through the system. Others enter burdened by the weight of scarcity. The school experience often serves only to entrench this divide.
This reflects what economists Peter Temin, Lance Taylor, and Servaas Storm identify as America’s “dual economy.” The country has increasingly divided between a secure, high-wage sector and a more extensive low-wage sector characterized by precarious work and limited opportunities for advancement. Temin emphasizes that we’ve constructed a system that determines who benefits and who doesn’t through policies like tax breaks for the wealthy and diminishing labor protections. He views education as one of the few avenues through which children might access the prosperous sector, yet, as we observe, schools reflect the inequalities within the wider society.
This raises an essential question: can schools effectively mitigate the damage caused by rising inequality in America? Some argue that a transformative tool like AI could provide the necessary resources for those at a disadvantage to catch up.
Machine Dreams
Before diving into the AI discourse, it’s prudent to reflect on some recent historical context.
When the internet was introduced into education, proponents heralded it as the great equalizer, providing all children with access to a wealth of information. The Clinton/Gore administration championed this vision, explicitly framing internet access as a means to bridge the gap between affluent and disadvantaged schools, thereby granting inner-city children “the same universe of knowledge” as their suburban peers. A 1996 White House report quoted a teacher from a school serving low-income students, confidently declaring: “The Internet erases boundaries of age and class.”
The actual outcomes have proven far less groundbreaking.
Despite the rapid proliferation of technology, simply equipping students with computers and internet access did not reliably enhance learning. The OECD found little correlation between increased computer access in schools and student performance. Later research revealed a new concern: digital devices often serve as irresistible distractions. As most parents recognize, young people using devices will multitask, check messages, browse unrelated sites, and not adequately focus. Studies have also shown that technology—both at home and in schools—can amplify inequality; a report from Common Sense Media found that the students most capable of using technology often come from families with ample resources and support.
The city of Newark serves as a cautionary tale of what can happen when Silicon Valley targets schools. In 2010, then-mayor Cory Booker and Facebook’s Mark Zuckerberg launched a $200 million initiative aimed at transforming the city’s failing schools into a national model. This effort introduced a tech-driven approach powered by data, measurement, and prompt results to public education. The reforms included expanding charter schools, utilizing data for teacher and student evaluation, and exploring cutting-edge technology for personalized learning.
However, despite the influx of funds, the anticipated change never materialized. Initial implementations resulted in a decline in achievement. Five years into the project, independent researchers found improvements in English scores, but no notable gains in math, with much of the English progress owed to shifts in student enrollment rather than actual enhancements within the existing schools. Zuckerberg eventually acknowledged his limited understanding of the community’s needs and dynamics.
Yet despite this setback, the #Zuckerberg approach prevails: when facing an educational issue, reach for a technological solution—more data, improved software, personalized learning opportunities. However, the root causes of inequality don’t dwell within classrooms, awaiting revelation from tech moguls or well-meaning politicians. They openly reside amidst stagnant wages, declining purchasing power, inequitable taxation, and insufficient public investment. Although Zuckerberg has committed significant resources towards promoting equality, philanthropy does not equate to true redistribution, and it often leaves the fundamental issues of wealth and power unaddressed.
To tackle inequality, we require the persistent, unglamorous efforts to implement policies like expanded Medicare, enhanced parental leave, and taxes targeting billionaires and their corporations.
This is not to diminish the importance of the classroom—it is undeniably vital. Robust evidence suggests that effective teaching can help narrow achievement gaps. For instance, Roland Fryer and his research team discovered significant academic gains for disadvantaged teenagers when they received individualized attention from tutors (the human kind) in small groups. Quelle surprise: Providing sustained attention from a proficient adult leads to better learning outcomes. The real challenge lies in asking teachers to deliver this level of support at a large scale. In schools catering to low-income communities, any time spent managing a class or redirecting attention detracts from valuable instructional time.
Reports from Mississippi and Louisiana serve as powerful reminders of the significance of effective instruction. Mississippi’s educational successes arose from comprehensive improvements in literacy instruction, curriculum, and teacher training. Louisiana has undertaken a similar initiative, emphasizing research-driven curricula alongside teacher training and targeted tutoring. These cases illustrate that when schools equip educators with improved tools, effective training, and provide struggling students with tailored support, academic performance can improve even in areas grappling with persistent poverty. However, it is essential to note: these successful initiatives have not eradicated the underlying disparities in opportunity.
This raises a prevailing question: can AI provide schools with a means to finally tackle these gaps at scale by offering individualized education to every student? The conversation suggests that innovative technology might personalize learning experiences, helping students overcome the challenges they encounter entering the classroom.
Some evidence supports this notion: a study conducted by UC Berkeley found that students receiving ChatGPT-generated hints on math problems achieved similar improvements to those receiving hints from human tutors—though this study involved online participants rather than children in traditional classroom settings.
A more direct 2025 study in Turkey involving nearly 1,000 high school students demonstrated that AI could enhance performance while being utilized. However, a notable downside emerged: students who had become reliant on an unrestricted AI system performed worse when required to work independently. Encouragingly, a tutor programmed to provide hints rather than outright answers mitigated much of this decline. The overall lesson here is clear: while personalized AI can assist students, it is not a miraculous solution.
Thus far, it appears that the advantages of AI in education are neither guaranteed nor fairly allocated. A 2025 review by educational researcher Aziz Mimoudi and an analysis from the OECD on AI and educational equity both raise alarms. Mimoudi, a co-founder of the Institute of Education Sciences, warns that “without safeguards, the very systems that claim to close learning gaps may inadvertently reinforce them.” (Moreover, we lack clarity on which versions of the technology are effective, as RAND researchers point out.)
What is apparent is this: a child who is well-rested, well-nourished, supported by adults, and afforded time to contemplate is in an entirely different position compared to another who is hungry, fatigued, distracted, or anxious about home circumstances. AI might deliver identical tutoring to both a privileged and a less fortunate child, but it cannot replace the essential social and economic frameworks necessary for child development. Furthermore, there are significant risks in conducting an uncontrolled experiment in classrooms with a technology still in its infancy, limited evidence to support its efficacy, and indications—such as potential overreliance on AI leading to diminished skills—already raising concerns.
Additionally, compounding the urgency for caution is the staggering amount of money involved. Tech giants are aggressively entering the educational arena, leading schools to become battlegrounds for AI implementation. Companies like Google, Microsoft, OpenAI, and Anthropic are launching campaigns to integrate AI into education, providing free or heavily subsidized access to their tools for schools, teachers, and students.
For instance, Microsoft offers Copilot Chat at no extra cost to eligible educational users; OpenAI grants free ChatGPT access to validated K-12 educators in the U.S. and has broadened free access through district partnerships; Anthropic, too, has made offers for free or subsidized access via education initiatives and its Claude for Teachers program.
While these companies express their desire to aid educators and prepare students for an AI-driven future, an underlying business strategy exists: by familiarizing a generation of students and teachers with their tools, they secure potential customers for years ahead. Just as Google and Microsoft established their software as the underpinning of American schools, they now aim to embed AI into the core of student learning, teacher instruction, and school operations.
Nevertheless, they cannot eliminate the pressing issues at hand. An AI tutor may illustrate how to solve quadratic equations in an engaging voice, yet it cannot provide stability and jobs for their families.
We often question whether schools have sufficient funds, if educators are well-trained, or if students possess the latest tools. Yet, education functions within a broader social and economic framework, one where schools cannot rectify inequalities that originate elsewhere.
The debate surrounding AI is fundamentally a discussion about political economy. If AI is set to elevate the significance of skills like human judgment, adaptability, and collaboration, we must deepen our analysis beyond the technology. Children require the social infrastructure to enhance these skills—namely, stability, nurturing relationships, and adults capable of being present in their lives. They need access to affordable housing, quality healthcare, and caregivers with enough financial flexibility to engage meaningfully with them.
Simply put, America’s educational systems rely on a robust middle class.
The narrative positing that AI will finally afford every child a top-notch education is, at best, a minor part of the greater issue. A cautionary note reminds us: if the advantages of AI primarily benefit capital owners and corporate shareholders while leaving others with escalating insecurities, this technology will ultimately exacerbate the very inequalities that schools aim to address.
Addressing these challenges calls for political solutions, rather than mere technological fixes.
