AI and the college payoff, NHJ’s school in context, Sal Khan’s origin story: A Chalkbeat Ideas roundup

Is AI devaluing a college degree? An update.

For many years, a college degree has been a major advantage for those seeking well-paying jobs. This has been a key theme of this newsletter. Yet I’ve also noted that it’s reasonable to wonder whether artificial intelligence will wipe out some white-collar work and devalue a college education. 

A new study provides initial evidence for that theory: The college wage advantage has dropped sharply in the last several years. See the graph below. The researchers offer a rough estimate that a third of this dip is due to AI, though it’s hard to say for sure. Since the pandemic, the job market for people without college degrees has been unusually strong.

Keep in mind that the college premium remains high, and we’re still early in seeing how the economy responds to AI. This data will be worth watching closely.

College for all for who?

Prompted in part by AI-related concerns, some educators are rethinking the college-for-all mindset that some say used to dominate K-12 schools. I was curious how close U.S. students got to that aspiration. This Brookings Institution study from 2023 tracked students who started ninth grade in 2009. Less than half enrolled in a four-year college, and a third didn’t go to any college. Only students from the most economically advantaged families came close to “college for all.”

The education story that broke the internet

There is much to be said, and much that already has been said, about Nikole Hannah-Jones’ blockbuster piece in the New York Times, describing the poor education her daughter received in two New York City schools.

For now, I’ll offer one revealing graphic. It plots New York City elementary schools by student poverty rates and school performance. Performance is measured by the school’s impact on student learning, controlling for factors like poverty. A lower score implies worse learning gains, with the median being 0.5.

P.S. 307, where Hannah-Jones sent her daughter, is among the worst performing elementary schools in the city by this measure. The most recent data available is from 2025, years after Hannah-Jones’ daughter moved on to middle school. Alarmingly, the school seems to have continued to wallow.

In general, high-poverty schools perform worse than more advantaged schools, even on a measure that accounts for student need. This suggests that the concentration of disadvantage creates severe challenges for schools. 

Yet many high-poverty schools score at or above the city average. Poverty and segregation seem to matter, but so do instructional quality, school culture, and many other things (including demographic factors not captured by the city’s poverty measure). 

Notably, P.S. 307 does not appear representative of high-poverty elementary schools citywide. 

Revisiting Sal Khan’s origin story

Over the years, Sal Khan has often described how he stumbled into the field of education. The MIT graduate was tutoring his younger cousins and started making videos to explain math concepts. His cousins actually preferred the YouTube videos to live tutoring, Khan said in a 2011 Ted Talk. “The very first time that you're trying to get your brain around a new concept, the very last thing you need is another human being saying, ‘Do you understand this?’” Khan explained then.

But in a recent interview with me, Khan revisited this origin story. In hindsight, he now says, it was not just that he was good at explaining tricky math by video or otherwise. It was that he had a relationship with his cousins that encouraged them to work hard. “The real power of the one-on-one is the accountability and the pushing,” Khan said. That cannot easily be done by video or AI chatbot.

We were speaking about a recent study showing that students are not engaging much with Khan Academy’s AI bot. Khan says the company is working to better integrate the AI so students use it more. 

Mark Zuckerberg’s latest tech hope

Meta CEO Mark Zuckerberg, meanwhile, has more uncomplicated enthusiasm about AI’s potential to transform education. “Everyone will have a personalized tutor,” he wrote in a recent manifesto arguing against extensive AI regulation. “Students will have extra help in areas they need it that is currently only available to those whose parents can pay.”

Longtime ed-tech observers may be getting déjà vu. Less than a decade ago, Zuckerberg predicted that computer programs would serve as exceptional tutors. Those efforts fizzled and have been largely forgotten.

Zuckerberg’s new hope for education may well be genuine, but he’s also leveraging this idea for a political purpose: To make the case that AI’s value outweighs its potential harms and that regulation should be limited.

Americans aren’t so hyped about AI in schools

Zuckerberg’s pitch faces an uphill battle, though. Most Americans believe schools will better prepare students for the future by banning AI rather than integrating it into classes, according to a recent NBC poll. NBC also found that a majority of adults believe AI will do more harm than good in schools. Such skepticism may help explain why large school systems, including New York City and Los Angeles, recently announced sharp restrictions on AI.

The value (and limits) of virtual learning

I recently wrote about the growing number of students receiving their education entirely online. Chalkbeat Ideas reader and University of Chicago researcher Chris Campos pointed me to a recent study of his that helps explain why. Using data from Los Angeles, Campos and colleagues show that students report lower rates of bullying after opting into online education. But Campos’ paper, like many others, also finds that most students learn less in online schools. “Some families appear willing to forego short-run achievement gains in exchange for guaranteed improvements in bullying-related outcomes,” the study says.

The surprising case for outsourcing civil rights enforcement from the Education Department

The Trump administration is moving aspects of the Education Department’s Office for Civil Rights to the Department of Justice. Kenneth Marcus, who ran that office during the first Trump administration, says this is a good idea. “One of OCR’s persistent structural challenges has been that educational institutions have often had insufficient reason to fear it,” Marcus writes in Education Next. The DOJ, he argues, has a “bigger hammer” and more credible “enforcement authority.”

This may well be true, but it is a surprising case to make while the Trump administration says it wants less Washington meddling in schools. Marcus argues the move would give more power to the federal government in civil rights enforcement.

Around Chalkbeat

ICYMI from Chalkbeat Ideas

Reach me at [email protected]. Forwarded this email? Sign up here to receive the Chalkbeat Ideas newsletter.