Midterm elections are often tough for the party that controls the White House. That’s especially true for Republicans this year, with inflation up and Trump’s poll numbers way down. Signs of a potential blue wave are evident in the generic ballot and projections for key races. Five weeks from Election Day, betting markets predict that Democrats will claim both houses of Congress. Last week, Nate Silver marveled, “Polls for Republicans are just getting worse by the day in a way that isn’t actually all that common in midterms.”

Republican senator John Kennedy recently visited Kansas, Nebraska, and Iowa to fundraise for GOP candidates, grumbling, “If you told me six months ago I was going to be having to spend my time in those three states, I wouldn’t have believed you. Those ought to be fat guy layups for us.” The New Republic has observed, “Trump’s super PAC and its cutout organizations are spending big money in at least 17 districts that Trump carried in 2024 by over five points.” Latino voters who helped fuel the GOP’s 2024 gains are exhibiting buyer’s remorse.

Some Republicans insist they’ll still prevail, pointing to the GOP’s money advantage or the shaky performance of polling in recent elections. It’s possible; time will tell. But odds favor a good November for the blue team, and I’ve been fielding a lot of queries about what that could mean for education. I figured I’d share a few thoughts.

Expect oversight with a vengeance. Remember how House Republicans put the screws to Hunter Biden and Claudine Gay? You ain’t seen nothing yet. The Trump White House has been flagrant in its norm-breaking and unapologetic about its self-dealing. Democrats are itching to launch investigations and start imposing some accountability-cum-payback.

With so many potential targets for Democratic investigations, the presumptive majority will have to pick and choose what deserves scrutiny. For those who don’t spend a lot of time noodling on such things, The Dispatch’s Kevin Williamson last week offered up a partial accounting of Trump’s second-term “excesses”:

The enrichment of the president and his family and his cronies; the campaign of wanton murder on the high seas; the illegal war against Iran; the illegal war against Venezuela; continued attempts to undermine confidence in our elections and to meddle with their administration; the illegal imposition of a new national system of taxation in the form of unconstitutional tariffs; . . . the attacks on allies undertaken while the president enabled actual literal enemies of the United States of America, including the Vladimir Putin junta in Moscow and the Kim Jong Un cult in North Korea; financial arrangements that run from the shady to the illegal and that stretch from Caracas to Doha.

You’ll notice what’s missing from the list. It’s unlikely much of the recent education turbulence will make the cut. Democratic leadership will want to focus on the things they judge most likely to resonate while avoiding narrative overload. What’s that mean for education? I’d guess that Democrats would take a hard look at DOGE’s cuts to education contracts and staff, the push to dismantle the Department of Education, the legal settlements with elite universities, and the way the administration has approached civil rights enforcement. If the Democrats take both chambers, they’ll have added bandwidth for more such investigations.

While the savvier hands in the Department of Ed have probably been careful about their paper trail, Democrats are likely to surface a treasure trove of incriminating missives from DOGE’s amateurs and Trump’s White House apparatchiks. Fielding oversight requests would become a huge time-suck and distraction for the executive branch. The pace of Trump 2.0’s education efforts would slow markedly.

The DSA isn’t likely to loom large for education. A lot of political conversation this year has focused on the rise of the Democratic Socialists of America. While their numbers will be small in the new Congress, DSA-aligned members will argue that they represent grassroots energy on the left.

Tensions between the DSA and establishment Democrats may loom large for healthcare or our relationship with Israel, but it’s less likely to shape education policy. It’s notable that DSA candidates have embraced hugely expensive measures like Medicare for All, free childcare, and a national rent freeze but have had little to say about schools. Indeed, the DSA platform’s education agenda is limited to universal pre-K, free college, and loan forgiveness. There’ll likely be some noise about Dems exploring a deal to boost early childhood funding, but it’s a good bet that the Democrats’ direction on education will be determined more by the familiar tug-and-pull of unions and centrists than by the priorities of the AOC-Mamdani camp.

Choice will dodge a bullet. The effects of a potential blue wave on Congress are obviously significant, but so are the potential consequences beyond the Beltway. DSA candidates who reject “privatization” and champion government-run healthcare have been oddly quiet about their hostility to school choice. It just hasn’t been a big target for them. Heck, New York City’s democratic socialist mayor hasn’t even really gone after charter schools.

What’s going on? Well, these programs are popular with core Democratic constituencies. Low-income and minority parents with lousy neighborhood schools like having the choice to send their children to better ones, and, in cities like Los Angeles or D.C., many now take choice programs for granted. In Wisconsin and Arizona, Democratic governors have been reluctant to square off against established choice programs. The easier path is to cap, regulate, and quietly stifle them. The Biden administration, for instance, would insist that it wasn’t “anti”-charter school even as it pursued federal funding rules that made it tougher for new charters to open.

While a blue wave would strengthen the hand of choice skeptics, the popularity of education options and the energy behind the “abundance” movement is likely to constrain them. This may be especially true in the case of the new Federal Scholarship Tax Credit. Even if Democrats capture the governor’s mansion in Ohio or Georgia, for instance, they may hesitate to overturn their predecessors’ decisions to opt in—choosing to treat the issue as a fait accompli.

Techlash will accelerate. Opposition to data centers, worries about social media, and support for school cell phone bans are bipartisan, so it’s not likely that a big blue wave would fundamentally alter the ed tech picture. That said, Trump’s embrace of TikTok, AI, and Big Tech has limited what Republicans are willing to say or do against the industry. A blue wave could change the dynamics in Congress while triggering hearings designed to surface plenty of embarrassing revelations about Big Tech and its relationship with the White House. That would only embolden tech skeptics in blue states and school districts. The AI pauses adopted by the school systems in New York City and L.A. could spread, especially if pushing back against AI becomes seen as a blow against the Trump agenda.

Higher ed will get a respite, but only a modest one. Colleges and universities have been in the crosshairs during Trump 2.0. As regular readers know, I think the scrutiny was richly deserved. But it has also featured a considerable amount of bullying and dubious conduct. If Democrats win big, much of that may abate. Congressional oversight will distract Trump’s higher ed team, and the change in power will provide allies and air cover for beleaguered campus leaders. The pressure on higher ed will ease (even if an aggrieved, isolated Trump tries to lash out with a new spate of directives).

But the relief won’t be as palpable as campus leaders might anticipate. There’s a good chance congressional Democrats will dig into the deals that some colleges struck with the Trump administration—putting college presidents in the hot seat and uncovering negotiations that they’d thought would stay secret. And then there are the new limits on lending and the new higher ed accountability system legislated by Congress last year. These measures are now the law of the land and enjoy some support on the left from officials concerned about college costs and completion. Even a massive Democratic win wouldn’t send things back to status quo ante.

Of course, nothing is permanent in American politics. Nowadays, we tend to approach every election as epochal. But the truth is that wins and losses are temporary. By Christmas, the pundits will be handicapping the 2028 primaries.

Frederick Hess is an executive editor of Education Next and the author of the blog “Old School with Rick Hess.”

In 2024, Laurence Holt of the XQ Institute wrote in this publication about “The 5 Percent Problem,” a term that quickly became common parlance within the ed-tech zeitgeist (see “The 5 Percent Problem,” features, Fall 2024). Holt argued that when it comes to measuring the student-learning outcomes of various ed-tech programs, the results often only apply to the 5 percent of students who “used the program as intended. The other 95 percent see minimal gains, if any.”

Holt’s primary focus was on online math programs, and the central concern running throughout his essay involves whether and how students might “get the recommended dosage” of using various ed-tech tools. Whether the problem arises from unmotivated students or unenthused teachers or disparate access to tech or some combination thereof, the challenge when framed this way is getting students to opt in to using a particular technology.

Generative AI does not suffer from this problem. Uniquely perhaps in the history of ed-tech, AI has been broadly embraced by students worldwide, to the point the phrase “AI is inevitable” has become commonplace in education discourse. There are pockets of resistance, of course, and over the last six months we’ve seen mounting opposition to AI across multiple vectors, most prominently with data centers. Nonetheless, we all know students are using AI frequently. “Dosage” is not an issue.

In the nearly four years since ChatGPT was commercially deployed, however, we have suffered from a lack of high-quality empirical research on the impact of students using generative AI. There are notable exceptions—such as these—but a recent research landscape analysis out of Stanford indicated that, out of more than 800 studies of AI in education, a mere 20 produced credible causal evidence. Meanwhile, at least one prominent meta-analysis purporting to show massive learning gains stemming from AI has been retracted, though not before being viewed 400,000 times. The field is a mess.

Perhaps this is because studying AI in “the real world” poses serious challenges. We know AI tools are free and widely available to students. We also know many are using them after school hours for various purposes, including to help with their schoolwork. But quantifying this is very difficult, because it’s hard to peer into the home lives of children. And it’s equally if not more challenging to connect out-of-school behavior to measurable learning outcomes; the data sets are not easy to link. Accordingly, research on AI’s education impact has tended toward lab studies or small-scale evaluations.

No longer.

Evidence of AI’s Impact on Learning

Recently, the study “The Generative AI Learning Penalty: Evidence from Chinese Secondary Education,” authored by David Strömberg, Victor Lei, and Yanhui Wu, went viral after The Economist published data visualizations from the research paper. The working paper came out in June 2026 and uses data from approximately 27,000 Chinese students in grades seven to 12 to explore a straightforward question that has preoccupied me for several years: “How does self-directed use of generative AI affect cumulative learning over time in ordinary school settings?”

We will get into the details shortly, but here’s the headline summary of the results:

• Between 2023 and 2025, approximately 80 percent of all students started using generative AI. After just five months of use, approximately 50 percent of these students engaged in “full homework outsourcing,” meaning, they essentially stopped doing their homework independently and used AI instead.
• Over time, this caused significant learning loss as measured both on monthly closed-book tests and via the comprehensive high-stakes exams that China uses to determine high school and college placement for students. By the end of June 2025, overall student performance fell by 24 percent on the high school entrance exam and 18 percent on the college entrance exam—a massive decline.
• There is no evidence—none—of any corresponding learning benefit arising from students using AI. As such, the researchers state plainly that “our findings show that generative AI, which is likely to become a prevalent technology for education, has a substantial negative impact on student learning.” (My emphasis)

Put simply, we now have rigorous, empirical evidence on the real-world, long-term impact of students using AI. This research indicates that AI substantially harms learning for half of all students, what the researchers call the “Generative AI penalty,” and provides no discernible benefit to the other half. Again, by the time this study concluded, 50 percent of students using AI with regularity had fully outsourced their homework to generative AI. For these students, when at home, they just stopped thinking about their schoolwork.

In homage to Laurence Holt’s 5 Percent Problem, I have taken to calling this finding the “50 Percent Problem,” a conservative measure of the harm stemming from widespread AI adoption by students, rather than their failure to use it. I say it’s conservative because the researchers who conducted the study go further and suggest that “the negative effects on learning outcomes appear to be mostly driven by the 81 percent of AI-using students, who spend less time on homework than even the fastest non-AI student, receive high homework scores matching the capability of generative AI tools they are using, and yet very low exam scores.” Whatever number you prefer, however, the point is that a massive number of children are learning less because of AI: at least half and almost surely more than that.

As I’ve said repeatedly for several years, generative AI is a tool of cognitive automation. It’s both predictable and tragic that students are learning less because of it. This new empirical research provides us with a clear measure of the degree of this harm.

What the Data Show

So, let’s turn to the data itself. In what follows, we’ll look at the performance of students who did and did not use AI over time across multiple learning activities.

The baseline results come from the 81 percent of students who eventually used generative AI for schoolwork but capture only the period before they began using the technology—these results are labelled “Pre AI,” and charted in green. Next, there are data from the admirable group of students, all 19 percent of them, who did not use generative AI at any point between 2023 and June 2025 when the study concluded—these are labelled “Never AI,” and charted in blue. The third set of results capture the performance of the first group of students after they commenced usage—these are labelled “Post AI,” and charted in red. It’s worth noting these students did not all start using AI at the same time, as adoption phased in over time. (For the data wonks, the researchers used “student-month” rather than student as their unit of analysis, meaning that each student produced multiple observations based on data collected each month. Remarkably, the researchers were able to capture the precise point in time students began using generative AI by having them record the registration dates for the AI tools on their devices.)

We’ll start by looking at the amount of time students spent on their homework. Then, we’ll look at how students scored on these same assignments. Finally, we’ll examine how student performance changed on monthly closed-book exams, as well as on China’s very high-stakes high school and college entrance exams.

1. What is the impact of AI on the amount of time that students spend on their homework?

The researchers were able to track homework time because students had to log in online to download their assignments and upload them upon completion. The results here are unsurprising—AI significantly reduces the time students spend doing homework. On average, both the Pre AI and Never AI students spent about an hour on homework, across a range of 50 to 80 minutes. In contrast, most Post AI students spent around 45 minutes, and some even cruised through in a mere 25 minutes (presumably the minimum amount of time it takes to copy-and-paste answers from a chatbot).

This is AI as tool of cognitive automation working exactly as intended.

2. What is the impact of AI on how students score on their homework?

Like politics globally, there’s a substantial shift to the right.

Here again we see that Pre AI students and Never AI students have near-identical results, as we’d expect. Not so with the Post AI students, who see a huge shift in purportedly “positive” outcomes, with scores far higher than even the most studious Never AI student managed to achieve. Note the scores here were normalized to make the average score a 100 (not the maximum), so a score of 130 essentially means “30 percent above the average.” To restate, the Post AI students studied less yet scored better than their peers. That’s a pretty sweet deal for them.

But it’s worth reflecting on the broader implications of this within schools. When I talk to students about AI, one thing I hear them say frequently is that they don’t want to be played for chumps (my term, not theirs). Meaning, if they know their classmates are using AI and getting better grades as a result, even those inclined to resist AI may feel pressured into using it, just to keep pace.

It’s a cognitive race to the bottom.

I’ll also add that homework scores are obviously a very imperfect measure of student learning. A great deal of education research, and certainly the studies often promoted by ed-tech vendors, use comparable “point in time” data of supposed learning akin to what we see here. This is understandable, to a degree—it’s very difficult to gather information on long-term learning outcomes. But what ultimately matters in education is building durable student knowledge. The seductive danger of AI that’s being exposed here is that, by using it, students may have falsely believed their schoolwork was proceeding swimmingly. In this sense, AI fosters a mental masquerade as scores go up while actual learning goes down.

Now, to remove the mask.

3. What is the impact of AI on how students perform on closed-book exams?

As Democrats are hoping for the upcoming midterms, there’s a massive shift to left.

Consistent with the two previous data sets, we again see near-perfect alignment between the Pre AI and Never AI students on monthly closed-book exams. But now the adverse impact of AI is laid bare: the Post AI students are scoring at levels far lower than their Never AI peers. Indeed, many score far lower than any student did prior to AI’s existence.

The steep downward shift evident from comparing the green and red distributions represents what I call the “Deadweight Learning Loss” caused by generative AI. What we’re seeing here is unambiguous evidence of a decline in overall student performance that grew over time as students adopted AI. The average decline was 20 percent with a 1.4 standard deviation (SD). Although far from an apples-to-apples comparison, this decline vastly exceeds the estimated learning loss in the U.S. after the pandemic (approximately .25 SD in math and .13 SD in reading).

What’s more, over time this Deadweight Learning Loss had significant adverse consequences for students on the comprehensive high-stakes high school and college entrance exams that China administers to determine student placement. For students who adopted generative AI two or more years prior to being tested, the estimated negative effects are 24 percent (1.5 SD) for the high school exam and 18 percent (1.3 SD) for the college exam.

If you are a parent with a child who is in junior high or high school and uses generative AI and has college on the horizon, this data should terrify you. This isn’t about academic integrity; it’s about a generation of kids being told AI is “inevitable” and “the future” and encouraged to act accordingly when we’ve yet to develop firm norms about what’s acceptable to do with these tools within education. Students are using AI in ways that are harming their cognitive development and foreclosing their life opportunities. These harms are not remediated easily, if at all.

Generative AI is cognitive cancer, and we are doing next to nothing to stop it from spreading.

Heading Off Objections

First, I want to credit The Economist for putting this research on my radar and for raising broader attention to its harrowing findings. Yet, remarkably, in the brief article accompanying the research data we’ve just covered, the anonymous magazine author suggests that “AI can boost learning productivity but only for those who use the technology intelligently.” Similarly, Blake Richards, a researcher at Google who works “on the intersection of machine learning and neuroscience,” argued on social media that AI is not really the problem here, because “if you control for how long students spend studying, then students using AI actually perform equal or better” than those who did not.

Motivated reasoning is a powerful force. Here, of course, “controlling” for how long students spend studying erases the major findings of this research. But let’s leave that aside, and probe whether this argument can be justified on its own terms. Does this study suggest that AI can boost learning if used “intelligently”?

As best I can tell, this claim is premised on this correlation between homework completion time and closed-book exam results:

If we ignore all that pesky data on the left side of this chart (which, of course, we shouldn’t) and just focus on the overlap between the Generative AI students who continued to study for the same duration as the No Generative AI students, it’s true there’s no major gap between them. It’s also worth noting there’s no discernible benefit to using AI either—apart, that is, from the spike at 75 minutes.

So what’s the deal with the spike? Might we cling to it as proof of the “promise” of AI? Well, here’s a good lesson on why one should always be careful eyeballing charts without access to the relevant underlying data. According to David Strömberg, the lead researcher, we’re talking about a vanishingly small number of students:

It is very rare for students who have adopted AI to spend 75 minutes on homework. This occurs for only 20 students, and for each of them only in a single month (0.2% of AI student-month observations). More than five months after AI adoption, we never observe students spending 75 minutes on homework. In fact, in this group, only four students spend more than 65 minutes on homework.

So let’s get real. Given this study involved almost 27,000 students, indexing on these 20 studious students who used AI isn’t just putting lipstick on a pig, it’s smearing its body in Revlon from snout to tail. We need to stop pretending there will be some massive benefit to learning if only we train kids properly on how to use AI. It’s not a force multiplier for learning; it’s fundamentally harmful.

A different and more sophisticated counterargument might proceed along the following lines: This study tracked student usage of AI stemming from its earliest days, when the tools were less capable than they are today. Perhaps relatedly, the researchers found evidence that it was the early AI adopters most harmed by generative AI, insofar as “the estimated AI learning penalty fell from around 25 percent in early 2023 to around 16 percent by June 2025.” As such, perhaps this study constitutes the “high-water mark” of AI-induced harm, and we might hope to see the AI learning penalty continue to drop.

But as the cliché goes, hope is not a strategy, and there are some problems with this counterclaim. For one thing, the rapid evolution of AI cuts both ways. We now have AI companies explicitly marketing “AI agents” to students to complete their tasks for them. Agents are even worse than chatbots from a cognitive development standpoint—at least the latter require some sort of interaction to produce output. For another, even if the magnitude of the learning penalty continues to diminish, there’s a question of volume. Recall that 20 percent of students managed to resist using AI prior to June 2025. Do you think that number has gone up or down since then? I know my bet.

Finally, some might argue I’m placing too much weight on this research. It’s just one study from China, after all. But surely China is the one country in the world that can match the U.S. for AI adoption and enthusiasm, though it may surprise you to learn that AI companies in China disable their products completely during the high-stakes testing periods. OpenAI, in revealing contrast, heavily promotes ChatGPT on college campuses during finals week in the U.S.

More importantly, if you are an AI-in-Education Enthusiast, I am confident you will not be able to produce research of comparable rigor that shows positive long-term education impact of generative AI in real-world conditions comparable to those here. And no, this single study from 2024 involving roughly 150 physics students at Harvard is not going to cut it.

Real Harm is Happening

The generative AI learning penalty will not abate of its own accord. The Edu-Cognoscenti continue to chatter about learning loss related to school closures during the pandemic. Well, the learning loss stemming from generative AI appears more substantial, is happening right now, and may endure far longer. So, to all the policymakers and philanthropists and so-called thought leaders who purport to take education evidence seriously, I ask you: What’s the plan to prevent ongoing education harms of generative AI? Are you doing anything at all?

I’ll close by echoing what cognitive scientist Dan Willingham recently said about when and how students should use AI as part of their education. His argument echoes what I’ve been arguing for several years and stems from our shared understanding of the basics of human cognition:

[T]he point of assignments is the mental processes required to complete them, and the point of the mental processes is learning. That seems to suggest a simple litmus test for the use of AI. Artificial Intelligence tools should not substitute for tasks wherein students would benefit from doing the mental work themselves. . . . Students should only use AI for things that they already know how to do well.

A tool that should only be used once you know something is not a learning tool. We must stop gesturing at AI’s imagined potential and focus our efforts instead on mitigating its undeniable harms. Our students need to stop paying the penalty for using technology that is designed to automate cognition, not foster human learning.

Benjamin Riley is the founder of Cognitive Resonance, dedicated to building human knowledge to halt AI hype. A version of this post originally appeared on the Cognitive Resonance Substack.

I just finished a new book with USC’s Pedro Noguera. (It’ll be out next April.) It’s got me reflecting on how odd the education discourse has gotten. But let me back up a bit.

In 2020, Pedro and I penned our book A Search for Common Ground. It consisted of a back-and-forth correspondence that played out while the nation was paralyzed by Covid-19 and rocked by the killing of George Floyd.

Reasoned, respectful debate was in short supply. Even before the pandemic, it had gotten damn hard to tackle thorny questions in good faith.

That’s what made the exercise so heartening. A right-wing think tanker and the progressive dean of a prominent ed school found that we could still talk our way through hard questions. We wrote candidly and trusted one another to respond with arguments rather than insults. Boy, was that a nice feeling.

We disagreed, of course, on topics like choice, school spending, for-profit education, SEL, and civics. But the dialogue helped clarify our disagreements. We repeatedly found some sizable patches of common ground on most issues, especially when it came to problems and principles. Where we tended to part ways was, predictably, on crucial questions of policy and practice.

Well, Pedro and I were convinced to take another swing at it. With the rush of post-pandemic controversies—from DEI to dismantling the Department of Education to “techlash”—there were a host of hot-button issues to debate. And that’s where things got odd.

This time around, I was struck by how often the two of us were on the same page. At times, it felt like we spent more time skewering the performative hobbyhorses of the left and right than arguing with one another. What was going on? Well, for all our differences, both Pedro and I are ultimately two guys who put a lot of stock in common sense. Pedro had no problem acknowledging that a lot of the DEI craziness got out of hand or that sprawling bureaucracies are a problem. And I had no problem conceding that the DOGE-driven cuts to the Department of Ed were feckless or that it’s hypocritical for Republicans to respond to campus speech codes with their own brand of censoriousness.

At many points, the conversation was less about left vs. right than about common sense vs. performative grandstanding. Take the debate over dismantling the Department of Education. It often devolves into aggressively dumb arguments that are more symbolic than substantive. Democrats bizarrely insist that shuttering a building with 4,000 bureaucrats will devastate student achievement and well-being, while Republicans bizarrely promise that doing so (while barely touching the Congressionally funded programs) will radically shrink the federal role and turbocharge learning. To my surprise, Pedro and I found little to argue about: We were supportive of streamlining bureaucracy and critical of Trump’s reckless, legally suspect approach.

On issue after issue, it seemed contemporary debates had been co-opted by performance artists on either side. That meant Pedro and I often found our practical differences overshadowed by mutual agreement that our disputes were more tractable than they can seem. We come at DEI, for example, from very different places. But we readily agreed that, while equality and inclusivity are worthy values, schools and colleges have done a number of indefensible things in service of them. When it comes to “censorship,” it really shouldn’t be that hard to distinguish between keeping pornography out of middle school libraries and banning Toni Morrison’s Beloved—even if it’s a distinction that PEN America and the activist community seem intent on ignoring.

For all the polarization that we see on social media or cable news, the data make clear that rumors of American division are greatly exaggerated. Most people aren’t fighting on X or Bluesky. In fact, the lion’s share of Americans never tweet at all. Most Americans aren’t watching the histrionics on cable news. And keep in mind that 80 to 90 percent of Republicans and Democrats agree that it’s “very” or “extremely” important to protect freedom of speech, freedom of religion, equal protection under the law, and the right to vote.

Given all that, why does the education discourse feel so nasty and aggressively performative? Well, shifts in media and technology have sorted us into information bubbles, while influencers and personalized feeds (assisted by Russian and Chinese bots) ensure that we’re bombarded with edge-case anecdotes designed to infuriate us. The result: We spend a day or two arguing bitterly about some teacher in Tennessee or a community college professor in Colorado, until algorithms amplify the next viral controversy and trigger the next rage cycle. Rinse and repeat.

This isn’t deliberation. It’s not even healthy debate. It’s an exercise in Pavlovian distraction and an assault on common sense.

Look, don’t get me wrong. Pedro and I still disagree about much that matters. But what’s striking is how modest those substantive differences now feel compared to our shared frustration with the partisans and provocateurs who just like to brawl.

In 2020, Pedro and I found that we could still reach across our divides with sufficient time, discipline, and intention. This time, after a half-decade of hysteria, distrust, and bad behavior, I found it a helluva lot easier to bridge our principled divides than it used to be. It’s not where I expected my reprise with Pedro to wind up. But I’ll take it.

Learning we could still talk through our differences was hugely reassuring amidst the chaos of 2020. Today, realizing we can make common cause in defense of common sense made for an equally bracing Round 2.

Frederick Hess is an executive editor of Education Next and the author of the blog “Old School with Rick Hess.”

Michael Kirst, professor emeritus of education at Stanford University, joins Paul E. Peterson to discuss a recently passed law that gives California’s governor more power over education in the state.

Follow The Education Exchange on Soundcloud, Apple Podcasts, Spotify, YouTube, or here on Education Next.

— Education Next

When I taught high school in Virginia in the early 2020s, I was only vaguely aware of the existence of something called “portrait of a graduate.” Whatever this portrait was, it didn’t seem to impact my daily work, so it didn’t capture my attention. On the other hand, I knew my IB courses and our state assessments backward and forward. Those shaped what we did in the classroom.

Since then, portraits of a graduate (or, sometimes, of a learner) have spread to at least 26 states, plus many more districts. The rationales vary—student disengagement, misalignment with evolving workforce demands, equity concerns—but the final product is typically a list of broad “competencies” that schools are meant to cultivate in each student. These range from the academic (“reason quantitatively”) to soft skills (“collaboration and teamwork”) to the warm and fuzzy (“embraces differences/diversity”).

I like to think schools were already doing the four to 15 items featured in each portrait. But as I experienced in Virginia, what actually affects students is not those specific items, but instead the practical requirements that change—or don’t—as a result of implementation.

As I’ve already written, I’m deeply skeptical of New York’s approach. To borrow from the immortal words of Cher Horowitz, the Empire State’s portrait is a “full-on Monet”: “It’s like the paintings, see? From far away, it’s OK, but up close, it’s a big old mess.” That is, seemingly unobjectionable competencies like “creative innovator” and “critical thinker” give way, upon inspection, to an apparent disregard for rigor, consistency, and transparency.

And unlike impressionist art, education policy can’t afford to brush over the details.

Fortunately, though, not every state portrait is trending Monet. To the contrary, the most recent portrait-of-a-learner proposal from Massachusetts is promising and could offer a model for other states heading in this direction.

Admittedly, the Bay State’s portrait puts forth its own hokey competencies, like “self-aware navigators.” But behind the jargon is a proposed system that specifies four sets of requirements for high school graduation: end-of-course (EOC) exams, a culminating experience, postsecondary planning, and, like 46 other states, required coursework.

The state EOC exam requirement is arguably where Massachusetts’s proposal differs most from New York’s. Two years ago, Massachusetts voters eliminated the state’s much-lauded MCAS test as a graduation requirement—much like the Board of Regents has done to New York’s erstwhile exit exams—but the Bay State hasn’t abandoned the concept of a consistent, externally validated measure of student learning. According to the plan, the new EOC assessments will be aligned with specific courses in English language arts, math, science, and potentially social studies, and they will be “designed, administered, and scored” by the state. Of course, the value of those exams will depend on how high Massachusetts sets the bar, and what the state does with the results.

Still, compare this to New York, where as-of-yet-undefined “assessments” are just one of many suggested options for gauging students’ graduation readiness, alongside performances, presentations, portfolios, and for all I know, baking soda volcanoes and shoebox dioramas—all to be designed and scored by the students’ own teachers. Even with a statewide rubric and the best of intentions, I can’t fathom how teachers could ensure consistency in scoring for “effective communicator” across a research paper on the Cold War, a dramatic recitation of a Shakespearean soliloquy, and a presentation on the works of Sabrina Carpenter.

With EOC assessments baked into the graduation requirement, Massachusetts will be far better equipped to identify where students and schools need more support, which is crucial for educational equity. And, as long as standards are set high enough, its proposed system will also ensure that all students are held to a high standard.

The exams aren’t the whole story, either. Even Massachusetts’s most flexible graduation requirements offer more consistency than anything in New York’s plans to date. Bay State students will need to complete a “culminating experience”: either a capstone project or a portfolio of assessments. Districts (not individual teachers) will define and score students’ culminating experiences based on a statewide rubric.

This is potentially a deft compromise: Stakeholders wanted students to demonstrate learning beyond traditional testing. The culminating experience allows some local discretion while ensuring consistency within districts. Most importantly, it comes in addition to—not instead of—statewide exam requirements that enable education leaders to compare results across districts, schools, and student subgroups. In short, Massachusetts hasn’t thrown the baby out with the bathwater; it’s found a way to honor the valid concern that a test can’t fully capture a student’s knowledge without effectively letting every school create its own set of graduation requirements.

Finally, the Bay State’s “postsecondary planning” recommendation includes several substantive components that address oft-cited concerns about students’ readiness for life after high school: demonstration of financial and digital literacy, federal and state financial aid application completion (or opt-out), and work-based learning experiences that may include internships, community service, and the like.

Of these, New York’s current proposal mentions only the potential for work-based learning, suggesting that districts could choose to “increase opportunities for career-connected learning.” It’s not clear whether any more substantive postsecondary preparation will become part of New York’s requirements.

Lest readers see New York’s haziness as excusable, since full implementation won’t occur until 2029, I’ll observe that Massachusetts has until 2031. That gives the Bay State a full five years to develop materials, train educators, and fine-tune the details. There’s just no excuse for New York to be building the plane while flying it when 860,000 high schoolers’ educations are at stake every year.

In sum, the aspirational nature of portraits of a graduate needn’t come at the expense of substantive graduation requirements—the stuff that actually shapes instruction and learning. States adopting portraits would do well to be a Massachusetts, not a Monet.

Meredith Coffey is the senior policy and operations associate at the Thomas B. Fordham Institute.

This post originally appeared on the Fordham Institute’s Flypaper blog.

For the better part of a century, most public school teachers in the United States were paid according to rigid “step and lane” systems, with pay and tenure based on education and years of experience. Meanwhile, performance evaluations were infrequent, did not include measures of student learning, and rated virtually every teacher as good or great. In the 2010s, education leaders and policymakers began to shift their focus from uniform longevity to individual effectiveness—most notably after the 2009 federal Race to the Top grant program incentivized states to adopt new performance-based systems for teacher evaluation, retention, and compensation.

The changes were swift and sweeping: By 2015, 43 states had revamped their teacher-evaluation systems. By 2020, 22 states considered teacher performance in tenure decisions; by 2022, 23 states at least partially tied teacher pay to performance (see “The Teacher Evaluation Revamp, In Hindsight,” feature, Spring 2017). Research has found positive impacts from new evaluation systems, which typically feature robust classroom observations, feedback, and student test-score data (see “Can Teacher Evaluation Improve Teaching?” research, Fall 2012). Other studies have found positive impacts of performance pay on student learning, including gains due to higher rates of turnover among low-performers (see “A Lasting Impact,” research, Fall 2017), and reassigning strong teachers to higher-needs schools (see “The Power of Performance Pay,” research, Spring 2025).

What about performance-based tenure rules? Traditionally, teachers have earned tenure after three consecutive years on the job—simply showing up to work on the first day of year 4 has enacted employment protections for life. But some states are changing those rules, including by requiring more years in the classroom and by setting minimum performance standards linked to multiple measures of effectiveness. If such conditions incentivize teachers to improve, that could be especially meaningful for early-career teachers, who experience dramatic growth and high rates of turnover in their first few years on the job.

I investigate this scenario in Tennessee, which implemented a new multiple-measures evaluation and tenure system in the 2011–12 school year. Under the new rules, teacher performance is assessed annually on three measures of effectiveness, including robust classroom observations and a value-added estimate based on student test-score growth. Teachers can qualify for tenure after five years in the classroom, but only if they earn “above expectations” effectiveness scores in years 4 and 5.

I look at changes in teacher value-added from 2008 through 2015 to identify the effects of the new evaluation and tenure system over time. I look at two groups of teachers: novices in years 1–3 of their careers who anticipate and then experience new tenure requirements along with new evaluations, and teachers in years 4–7, who are already tenured under the former service-time requirement but experience the new evaluations and associated feedback on their performance. I compare changes in these teachers’ value-added to that of similarly experienced teachers before the 2012 reforms.

Across the board, teacher performance improves as soon as the new evaluation system begins. But the improvement gains are twice as large for pre-tenure teachers: Value-added increases by 4.7 percent of a standard deviation of student test scores over and above average early-career growth before the reforms. On its own, the new evaluation system boosts value-added by 2.4 percent of a standard deviation for both pre- and post-tenure teachers, and the new tenure incentives increase value-added by another 2.3 percent of a standard deviation for teachers held to those rules. These differences translate to up to two additional weeks of instructional time and $2,000 per student in net present earnings.

I find that these improvements are largely experienced during teachers’ first three years on the job, when they are anticipating the future tenure incentive but before formal tenure rules apply in years 4 and 5. The gains are also lasting: Value-added does not decline after a teacher earns tenure despite the fact that the incentive to earn a high score goes away. The cumulative effect of the performance-based tenure rules for new teachers is 7.9 percent of a standard deviation in value-added growth in the first six years of their careers.

In summary, teacher performance improves before any scores count for earning tenure, and performance remains at the new higher level after teachers earned tenure. These anticipatory and persistent effects are consistent with skill improvements caused by the evaluation program and its performance incentives.

A New Teacher Evaluation System

Tennessee’s teacher evaluation and tenure reforms began at the start of the 2011–12 school year following the state’s 2010 First to the Top law and $501 million federal Race to the Top grant. The law requires that all teachers and principals be evaluated annually and permits student achievement data to be used in tenure decisions, among other innovations.

Under the new rules, each teacher’s evaluation includes three performance measures: a classroom observation rating, a value-added score, and an additional student achievement measure selected by the teacher. All three measures make use of a five-point expectations scale: (1) “significantly below expectations,” (2) “below expectations,” (3) “at expectations,” (4) “above expectations,” and (5) “significantly above expectations.” At the end of a school year, the three measures are weighted and averaged to determine the teacher’s “level of effectiveness” score on the same five-point scale.

First, classroom observations account for half of a teacher’s overall effectiveness score. A school principal or other administrator conducts observations using a detailed rubric that describes specific teacher behaviors and associated scores. Every teacher is observed up to three times each year and scored in categories like presenting instructional content, managing student behavior, and academic feedback. Observers provide feedback after each visit, and the task-specific scores are averaged for the final observation rating.Previously, observers had used a less robust rubric, and after being observed two to three times annually through year 3, a teacher was not formally observed again until year 8.

Second, value-added scores account for 35 percent of their overall effectiveness score. Known locally as Tennessee Value-Added Assessment System, or TVAAS, the scores are estimated by the SAS Institute and often referred to as the “student growth score.” When describing TVAAS to teachers, Tennessee emphasizes the growth characteristic and that students are compared to peers who scored similarly in prior years. While Tennessee principals and teachers have had access to TVAAS reports and value-added scores since the early 1990s, they were not formally used for personnel decisions prior to the new evaluation rules in 2012.

Finally, the student achievement score accounts for 15 percent of a teacher’s rating. This measure is also usually based on student test scores but typically focuses on the level of student achievement as opposed to growth. Each teacher defines this measure for herself, in collaboration with her school principal, choosing from a state-approved list of student assessments. The teacher and principal set the criteria that will map from student scores onto the five-point expectations scale, such as a class or grade-level pass rate on state tests, or some other commercial assessment.

In this research, I study only teachers who teach math and English language arts to students in grades 4–8. The description of Tennessee’s new evaluation system in the preceding paragraphs applies to these teachers, for whom value-added scores can be calculated. Evaluations for teachers in non-tested grades and subjects are different, notably teacher value-added scores are replaced with alternatives like student growth portfolios or school- or district-level data.

Tennessee’s Tenure Reform

Also starting with the 2011–12 school year, teachers’ annual level of effectiveness scores determine whether they earn tenure. Previously, teachers had been eligible for tenure based on how long they worked in a district. At the end of a three-year probationary period, a district had two options: either fire the teacher or grant them tenure. Some 90 percent of eligible teachers received tenure under the old system.

Under the new system, a teacher is first eligible for tenure after teaching for five school years, not three. Tenure is granted based on teacher performance in years 4 and 5 on the job: Only teachers who earn effectiveness scores of 4 or 5 in both years qualify. Teachers whose scores fall short can continue working on a probationary contract in year 6 and beyond, but they can only earn tenure after scoring 4 or 5 two years in a row.

These new rules are a real constraint on tenure. I look at the level of effectiveness scores of teachers in years 1 through 7 of their careers. Only about two-thirds of teachers earn scores of 4 or 5 during that time (see Figure 1). In any two consecutive years, just 57 percent of teachers score 4 or 5 in both years. However, teachers in their fourth and fifth years on the job earn higher scores, with 63 percent meeting the tenure requirement.

Tenure is revoked when a teacher earns two successive scores of 2 or 1, which rates their performance as below or significantly below expectations. In practice, however, this is rare: Fewer than 5 percent of teachers earn disqualifying low scores in two consecutive years. Schools may continue to employ teachers who lose tenure, if they choose, and teachers can regain tenure by earning scores of 4 or 5 twice in a row again.

Notably, the new tenure rules apply only to teachers who began working in 2009–10 or later—teachers hired before then were already tenured by the time the new rules took effect.

Assessing Impacts

One way to summarize these many details is to think in terms of a pair of experimental treatments. The first treatment is the change in performance measures for all teachers. The new evaluation system expands annual evaluations to every teacher, not just pre-tenure teachers. Further, classroom observations are based on a more comprehensive rubric, and teachers’ value-added scores are formally used in their performance evaluations. The second treatment is the change in performance incentives attached to the new measures: new tenure rules that require teachers to score “above” or “significantly above” expectations in both their fourth and fifth years on the job.

I estimate the impacts of these treatments separately and in combination, based on Tennessee data from 2007–08 to 2014–15. I focus on math and English language arts teachers in years 1 through 7 of their careers with students in grades 4–8, who are tested annually in those subjects. In all, my study includes data for 11,000 teachers and 720,000 students.

In my analysis, only teachers hired in 2009–10 or 2010–11 receive both the evaluation and tenure treatments. For these novice teachers, the tenure incentive treatment begins in year 4 and ends after year 5 if they successfully meet score requirements. By contrast, teachers who are in years 4–7 when the new rules begin never experience the tenure incentive treatment. These early-career teachers are subject to new performance evaluations but are already tenured under the old rules.

I focus on changes in teacher value-added in both of these teacher groups. For the novice teachers who receive both treatments, these first estimates capture the additional growth between their first and second, or second and third, years in the classroom. This analysis captures the anticipation effects of the new tenure incentives, because teacher performance only counts toward tenure awards in years 4 and 5 of their career. I then estimate changes in value-added between year 5 and 6, after incentives end, to identify persistent effects of performance incentives. These estimates show the additional growth in value-added, caused by the reforms, on top of the typical growth new teachers experience in their early years on the job.

For the tenured early-career teachers who only receive the evaluation treatment, the estimates describe growth between years 4 and 5, or 5 and 6, or 6 and 7—years when many teachers have completed their most dramatic growth but are continuing to improve. Still, again, these estimates capture the additional growth in value-added net of the average growth observed for teachers in years before the tenure reforms in 2012.

Novice teacher value-added improves by 4.7 percent of a standard deviation in the first year of Tennessee’s reforms (see Figure 2). That gain is in addition to the typical performance growth at the outset of a new teacher’s career. While both math teachers and English language arts teachers experience gains in performance, value-added for math teachers grows by more than twice as much: 6.5 percent of a standard deviation, compared to 2.8 percent of a standard deviation for English language arts teachers.

How much of the improvement in value-added would have occurred without future incentives? Robust new evaluations and feedback can reduce the costs of a teacher’s investment in her skills, but future incentives, like tenure, may increase the returns on a teacher’s investment in her skills. Those future returns alone could generate skill investments and performance gains.

Indeed, I find that value-added for tenured early-career teachers, who do not anticipate or experience the incentive, improves about half as rapidly, by 2.4 percent of a standard deviation. Math teachers improve by 3.6 percent of a standard deviation, while English language arts teacher performance grows by 1.3 percent of a standard deviation.

And what happens when novice teachers’ performance evaluations begin to count toward tenure, starting in year 4? Given that new incentive, we might expect a meaningful improvement in performance; however, I find only a small, estimated increase of 1.3 percent of a standard deviation in teacher value-added between years 3 and 4. This impact is relatively imprecisely estimated, and thus I cannot reject the possibility of no change in their value-added. In other words, the gains to performance overwhelmingly accrue in a teacher’s earliest years when they anticipate the future tenure incentives, but when their performance scores do not yet count for earning tenure.

Nearly two-thirds of novice teachers earn tenure after year 5 of employment, and their scores in year 6 no longer count for tenure. The conventional view of performance incentives would predict a decline in value-added after year 5 when tenure incentives end—but that doesn’t happen. I find suggestive evidence that teachers continue to improve and can rule out any notable decline in performance after the incentive is removed. The persistence of higher value-added is consistent with teachers having improved their skills in response to the new evaluation program’s tenure incentives.

What about the teachers who did not earn tenure? Roughly one-third of teachers do not earn consecutive “above expectations” scores by the end of year 5. These teachers are allowed to continue working in year 6 and beyond under a probationary contract and can eventually earn tenure if they score above the cutoff any two years in a row.

I compare attrition among these low-scoring novice teachers to tenured early-career teachers with similarly low effectiveness scores—in other words, teachers who would not have earned tenure had they been subject to the new rules. Some 4.6 percent of low-scoring teachers who don’t earn tenure on time quit or change jobs compared to 5.4 percent of low-scoring teachers who were tenured under the old, performance-blind system. Both groups have higher rates of attrition than the average Tennessee teacher in the years before the 2012 reforms. But the tenure incentives do not affect attrition. In that regard, the new tenure rules are not successful as a selection policy—at least not selection out of teaching. The reforms may have influenced selection into teaching, among newly hired teachers, but that is outside the scope of this study. I also find that value-added essentially holds steady for teachers who continue teaching on a probationary basis in year 6 who remain subject to the tenure incentive.

Implications

Teachers, like most workers, respond to robust performance evaluations and employment incentives—and in the case of Tennessee, these responses boost their effectiveness in the classroom and, therefore, student achievement. The impacts of these measures are best understood by considering their influence on teacher behavior over time. I find evidence of both anticipation and persistent effects following Tennessee’s evaluation and tenure reforms. First, novices’ value-added improves by 4.7 percent of a standard deviation as a result of robust new evaluations and the future prospect of tenure decisions. Further, value-added does not decline after a teacher is granted tenure and the incentive ends.

Ignoring these anticipation and persistent effects substantially understates the benefits of  teacher evaluation and tenure reforms. Consider a research design which assumes incentives only affect performance by raising effort when scores are linked to explicit incentives. The average teacher’s value-added improves by just 1.5 percent of a standard deviation in years 4 and 5, when scores count for earning tenure. Looking at the cumulative effect of the new tenure requirements across years 1 through 6, however, value-added grows by 7.9 percent of a standard deviation.

This pattern of effects—especially the anticipation effects and persistent effects—is consistent with teachers having invested in skill improvement as a response to the evaluation program’s performance incentives. An alternative argument, sometimes raised in education policy discussions, is that evaluation on its own and without any extrinsic incentives can improve teacher performance, because teachers are motivated professionals who will use the individualized feedback from evaluation to improve. The evidence presented here does not necessarily contradict that hypothesis; however, the gains of 2.4 percent of a standard deviation among already-tenured but still early-career teachers, who only experienced the evaluation reforms, are half as large as those for novice teachers, who were subject both to new evaluations and new tenure rules. If “motivation” plus “feedback” alone were sufficient for evaluation-induced skill improvements, then both groups would have experienced similar value-added gains.

The most direct application of these results is in understanding the effects of similar teacher evaluation programs. Many states, like Tennessee, have policies that link teacher employment security to multiple-measures evaluation systems, including classroom observations, student surveys, value-added, and other evidence of student learning. My analysis shows that such programs can improve early-career teacher performance, making their students better off.

Eric S. Taylor is a professor at the Harvard Graduate School of Education and a research associate at the National Bureau of Economic Research. This article is based on “Employee Evaluation and Skill Investments: Evidence from Public School Teachers” (NBER Working Paper No. 30687, 2024). The Tennessee Department of Education provided the data for this study but does not guarantee its accuracy or endorse the findings.

Suggested citation format:

Taylor, E. S. (2026). “The Lasting Benefits of Tougher Tenure Rules: Evidence from Tennessee.” Education Next, 26(3), 15 September 2026.

The pushback on AI and data centers is widespread and real, but parents seem a lot less worried about the devices that have overrun their kids’ classrooms. For all the attention to the recent Meta litigation and the push to ban cell phones in schools, parents haven’t urged educators to resist the cascade of ed tech represented by Chromebooks, iPads, Canvas, i-Ready, and Kahoot. If anything, they have welcomed these devices and programs into their children’s education and daily lives.

As I noted a few weeks back at Law & Liberty:

Earlier this summer, EdChoice reported that 86 percent of parents support using laptops in class, 84 percent support the use of online learning platforms, 80 percent support the use of online learning games, and 60 percent support the use of AI in class. Hart Research has found that 74 percent of parents describe ed tech as positive for their child’s learning and 54 percent say they want their school to use more technology going forward.

Why the continued warm feelings towards “educational” tech, even as polls evince a broader turn against Big Tech? I suspect two things are going on. One is the eternal allure of “future-ready” schools. Parents don’t want their kids to be left behind. Since schools are struggling and the world is changing, they hope a classroom full of cool tech means their kids will be up to speed. There’s also a deep-seated impulse to trust that educators know what they’re doing—that the devices and software are being put to sensible use.

But parents may yet awaken to the harm that a device-driven culture of corner-cutting and distraction is visiting upon their kids. If they do, I suspect efforts like the Fordham Institute’s new report, High Tech, Low Standards: High School in the Age of Distraction, will have helped.

The report, authored by Fordham’s Brian Fitzpatrick, shares the findings from a national survey of 1,000 high schoolers and another 1,000 recent high school grads. The results are sobering, if unsurprising. Three-quarters of respondents reported using some kind of “illicit” technology daily during school hours—50 percent of respondents said they use social media (like TikTok or Instagram) at least once a day in school, 33 percent said the same about watching YouTube videos, and 17 percent about shopping online.

The distraction problem isn’t just about sneaking and peeking, though. Sixty-two percent of respondents said they spend more than half their class time on school-issued devices. Nearly one-in-three students said they spend more than three-quarters of class time on screens. In short: Most students are spending most of the school day on screens before we ever factor in the time they illicitly spend on their personal devices.

Moreover, for all the suggestions that AI tutoring will bolster instruction, students are clear that they mostly use it to evade the rigors of learning. Forty-eight percent of students reported using AI to solve their math problems for them, 36 percent to summarize books or texts to avoid reading them, and 29 percent to pen the first drafts of their essays or reports. Twenty-six percent reported using AI for help during quizzes and tests. (That’s what us Gen X types used to call “cheating.”)

When asked to respond to statements about the effects of AI on school, 58 percent of students agreed that it’s made homework meaningless and 57 percent that it lets students get good grades without doing any real work. Technology is fostering a classroom culture where students give the impression of learning while the school day becomes a shrinking circle of videos, memes, and chatbot-supplied shortcuts.

Studies documenting the adverse effects of devices can be complicated and abstract. But survey results from students themselves admitting that most of them spend most of the school day on devices (with regular forays into illicit distraction) make the implications clear and concrete.

Here’s one very practical consequence: Fitzpatrick reports that students who spend more class time on devices are more likely to say that missing school won’t affect their grades. This tracks with recent speculation that ed tech is fueling chronic absenteeism, as students conclude it doesn’t much matter whether they stare at screens at school or at home.

Am I arguing that ed tech is “bad” for kids? No. I mean, I’ve written at length about the ways in which it can be a powerful tool of learning. The problem is that, in practice, 21st-century technologies have too rarely been used that way. I’ve long thought that parents should be angrier about this than they’ve shown themselves to be.

It’s never been clear if the prevalent parental lassitude towards classroom tech is because caregivers are unaware what kids actually do all day in “21st century” schools or because they’re reluctant to stand athwart “progress” when they’ve long been told that one-to-one devices are a key to their children’s future. Either way, enough research like Fordham’s might finally help get parents to appreciate the real-world consequences of how all that cutting-edge classroom tech is getting used.

Frederick Hess is an executive editor of Education Next and the author of the blog “Old School with Rick Hess.”

Angela R. Watson, a senior research fellow at the Johns Hopkins Institute for Education Policy, joins Paul E. Peterson to discuss how regulations with the new Federal Scholarship Tax Credit may exclude home schools and microschools.

The Emerging School Models: Broadening the Landscape for Choice conference will be held by the Harvard Kennedy School’s Program on Education Policy and Governance in Cambridge, Mass., on Sept. 17 and 18, 2026.

Watson’s brief, “The Trouble with Defining Homeschooling: Why Strict Definitions of Homeschooling Need a Reality Check,” co-written with Christy Batts, is available now.

Follow The Education Exchange on Soundcloud, Apple Podcasts, Spotify, YouTube, or here on Education Next.

— Education Next

Since it was published two years ago, Jonathan Haidt’s The Anxious Generation has made, we believe, a huge contribution by sounding the alarm on teen screentime.

Haidt’s been active endorsing solutions: parents shouldn’t permit smartphones until age 14; schools should ban phones from 8 a.m. to 3 p.m.; and policymakers should limit social media consumption out of school—during what we call “3 p.m. to 3 a.m.”

Haidt’s work has led to action. While kids overwhelmingly continue to get phones before age 14, the school bans have taken hold in large numbers, and some governments have enacted laws. Australia took the lead in December 2025, launching a law restricting the use of social media by children under 16. Indonesia and Brazil soon followed with their own. In the U.S., 12 states have some version of a restriction, mostly enjoined by the courts, though Florida’s is somewhat like Australia’s and “sort of” operational.

What does Australia hope to accomplish with its law?

As Prime Minister Anthony Albanese puts it, “Australians pride ourselves on our beaches and parks, our love of playing sports and socializing with friends. Less doom-scrolling means more chances to spend our precious hours relishing the remarkable world around us.”

Haidt says, “[Kids] will get countless hours of their childhoods back to sleep, study, or socialize, and we will all be better off for it.”

As authors, let’s reveal our priors. We’re cheering for the Australian law to work, because we would like to see teens experiencing less screentime and more “real life.” But we start with some skepticism, because some older national laws trying to limit screentime, namely in China and South Korea, have failed.

To its credit, the Australian government set up an evaluation of its law, done by Stanford’s Social Media Lab, involving more than 4,000 children aged 10 to 15, plus one parent for each. The first report came out in July, comparing kids’ answers and behaviors before the law took effect with their answers and behaviors three months later. The researchers will return at six months, one year, and two years.

So what were the July results? Let’s dig in.

A Quick Quiz

First, dear reader, we invite you to take a quick quiz about Australia’s new law.

1. What does the law do?

a. Bans children under 16 from using social media

b. Requires platforms like TikTok to prevent children under 16 from having accounts

c. Requires parent permission for children under 16 to use social media

Answer: b

The Australian law says that Meta, TikTok, and other social media platforms must put steps in place, such as “facial age estimation,” to keep anyone under 16 from having an account. U.S. states do something similar. Some bar accounts for kids under 14 (Florida), some do so for kids under 16 (Georgia, Louisiana, and Ohio, while Virginia caps usage at an hour a day), and some set the limit at under 18 (Arkansas, Utah, Texas, Nebraska, Mississippi, and Tennessee). Generally, the platform must try to verify the user’s age or obtain parental consent before letting a minor open or keep an account. Noncompliance brings possible civil penalties or fines. But unlike Australia, most of these state laws are currently blocked or partially enjoined by federal courts on First Amendment grounds, so they aren’t actually being enforced right now.

2. Imagine the Australian law worked perfectly, with zero workarounds. Could a 14-year-old Australian still use social media?

a. Yes

b. No

Answer: a

That teenager can still watch YouTube videos without having an account, for example. He couldn’t comment. YouTube couldn’t personalize his feed (though sister Google might be able to, through his browser).

3. What would qualify as a “win” for Australia’s law?

a. A decline in teen-operated social media accounts. Individual, personalized accounts are more addictive than using platforms anonymously, so if the number of accounts falls, that’s a win.

b. A decline in the number of hours spent on social media per day. If the average teen consumes four hours of social media per day before the law, we should see that amount decline after the law, which would be a win.

c. A decline in the total amount of screentime. A reduction in time spent on social media will be meaningless if it’s substituted with other screen activity like texting and gaming. If total screentime goes down, that would be a win.

d. An increase in the amount of time tweens and teens are active outside the home. “Less doom-scrolling means more chances to spend our precious hours relishing the remarkable world around us” was the bar set by PM Albanese. If that’s happening more, it will be a win.

e. An improvement in the mental health of Australian children. Nothing else matters if Haidt’s underlying claim that predicates social media bans isn’t satisfied. We need to see the mental health of youth get better for this to be a win.

Answer: A judgment call.

We anchor on “c”—for us to back the law, we’d want to see declines in teen screen consumption. We’d hope to see screentime substituted with positive, real-life stuff (d), and over time we of course ultimately want to see mental health improve (e), but we’d start with “c” as the plausible direct consequence of the law after only three months. That would be a “win” in our eyes.

Results from the Stanford Study

Here is what researchers found in their first study, three months after Australia’s law took effect.

1. Are social media companies actually keeping kids from having their own social media accounts? We see a clear drop.

But whether this decline constitutes a “win” depends on your answer to quiz question 3 above.

Writing in Time, Jonathan Haidt, Ravi Iyer, and Zachary Rausch point out that platforms removed 4.7 million under-16 accounts belonging to an estimated 2.5 million Australian children. For Haidt, this shows that “the law made immediate progress and compelled platforms to take action on a massive scale.” Moreover, they argue that account ownership should continue to decline from the 42 percent shown above as enforcement tightens and “no account before 16” becomes the norm.

2. Did social media use—interaction with the platforms, not the number of accounts—decline? Not by much.

Sadly, the authors do not have baseline data on screentime or social media time per day. They did capture it at the three-month mark (3 hours, 58 minutes on Saturdays and Sundays; 2 hours, 9 minutes on weekdays), so we will be able to see what happens going forward.

3. Do we see an increase in positive, real-life activities—the alternatives to screentime? Here we look at three metrics from the Stanford report, none of which shows meaningful change. Uh-oh.

4. Finally, is mental health better? Given the lack of change in offline activities, it would seem implausible we’d see an improvement, and that’s precisely what Stanford found.

Where Does This Leave Us?

How can we square the three-month Stanford result with Haidt’s claims that the Australian law is a smashing success?

We’re sympathetic to his position. His book has swept the world, even as it has inevitably gotten some pushback—some in good faith, some in bad faith. Often this causes people to dig in, to never admit they’re wrong.

We hope our critique is of the good-faith variety. We’d say to Haidt: You have laudably identified a key problem: massive screentime among our youth.

Your earlier goals were and are the right ones for tweens and teens: less screentime, more real-life living.

Your current effort to shift the goal posts—using the Australia example to focus on the technicality of a decline in social media accounts—is counterproductive to your own goals!

These early indicators from Down Under should be troubling to you. What if these findings hold at six months, a year, two years? You’ll be constantly hammered with headlines that you were wrong. You risk losing the baby (your correct diagnosis of how smartphones worsened teen lives) with the bathwater (your well-intended prescription of laws that seek but fail to reduce 3 p.m. to 3 a.m. screentime).

Big picture: You’re winning. You’ve persuaded many people of the harm. You’ve got reasonable 8 a.m. to 3 p.m. school phone bans in place. Attorneys general are chasing Meta for their harms to teen mental health.

When you’re winning, you have room to correct course when you stumble! Lean in.

First, acknowledge that the Australia results are not what you hoped for but you think we need time, and articulate how much. It is reasonable to allow time for culture to change and the law to work.

Create a scorecard of what you’d hope to see continue in order to endorse the law. Here is an example we created for you.

If that comes to pass—and there’s still time—then your idea will be persuasive. You can then make the case for similar laws you have cooking across the U.S. and around the world. If not, walk it back.

Remember Chicago’s soda tax passed in 2017? It was well-intended! They thought it would reduce consumption on something unhealthy. But the tax didn’t work, and the city repealed it four months later. Speed matters here.

Finally, you don’t have to be the owner of the Problem and the Solution. Invite more tribes into your tent—like us and many others—who share your goal of more real-life living for teens but have different mechanisms besides direct screen bans to achieve them. Work with PM Albanese to invite other experiments in Australia, where the measurement structure is already in place, to see if those might reduce actual screentime and increase hours spent each week doing “the good stuff.” Bring new people into the tent!

Sean Geraghty and Mike Goldstein are the co-founders of the Center for Teen Flourishing.

This summer, the Trump administration announced plans to overhaul the $12 billion federal Head Start program. The plan has sparked fierce debate, with fiery critiques from Democrats and early-childhood advocates. What is the Trump administration seeking to do? What motivated this effort? Is this ultimately about trying to abolish the program? For the administration’s take, I talked to Alex Adams, the guy steering this effort. Adams is the assistant secretary for family support at the U.S. Department of Health and Human Services. He came to this role after more than a decade in Idaho state government, including as budget director and head of the Idaho health and welfare department. Here’s what he had to say.

—Rick Hess

Rick Hess: You’ve recently proposed overhauling the regulations that govern Head Start. What’s the story?

Alex Adams: Head Start is a program I believe in deeply. Since 1965, it has supported more than 40 million children and families, promoting school readiness for low-income children from birth to age 5 through education, health, and family services. But over time, Washington has layered on rules that make it harder and costlier to serve families. Our 1,600-plus grantees operate in very different communities, and a one-size-fits-all rulebook can’t account for every circumstance. Our proposal preserves Head Start’s federal-to-local structure and the protections Congress set when it reauthorized Head Start in 2007, while giving local leaders and parents more room on operational decisions. It also strengthens the focus on nutrition, health, and physical activity. The principle is simple: Focus federal oversight on outcomes, safety, and statutory duties and stop micromanaging decisions local programs can make themselves.

Hess: What motivated these changes?

Adams: The most immediate reason is access. Regulations adopted in 2016 and 2024 contributed to rising per-student costs, which resulted in the loss of 277,658 slots for students from 2015 to 2025, according to a recent analysis by the assistant secretary for planning and evaluation at HHS. That’s a 29 percent decline in a decade. Supporters of those rules argued that fewer slots would be offset by higher quality for the families still served. But that same analysis found no substantial gains in quality, despite federal spending on the program outpacing inflation by double digits. Higher costs, no improvement in quality, and fewer children served: Our proposal is designed to reverse this dynamic.

Hess: How will loosening these requirements help boost the number of kids served?

Adams: A separate regulatory analysis by HHS estimates these reforms could preserve or expand as many as 236,000 slots nationwide by 2031 . That’s not a promise every program will add seats immediately—implementation would vary by local decisions—but it signals the scale of the opportunity. The analysis projects roughly $2.2 billion in annual savings at full implementation. To be clear, this isn’t a cut to federal spending; the federal appropriation won’t be affected. It’s an estimate of how much less grantees would need to spend on administrative costs, staffing, paperwork, and other expenses to be in compliance with the new rules. Because Head Start is grant-funded, programs can reinvest those savings in additional slots and services. The goal isn’t to spend less on children; it’s to make sure more of the money Congress spends actually reaches them.

Hess: As you know, critics have warned loosened requirements will lead to unsafe environments or problematic conduct. What’s your response to these concerns?

Adams: The Head Start Act’s statutory requirements remain fully in force, including services for children with disabilities, civil rights protections, and parent involvement. Programs also remain subject to applicable state and local law and statutorily required governance structures. What we’re reconsidering is whether Washington needs to prescribe every operational detail on top of those protections. Flexibility is permission, not a mandate. If a provider believes its current model is best for its children, it can keep it.

Hess: You propose cutting Head Start’s administrative budget cap for grantees from 15 percent to 5 percent. I get the desire to spend less on bureaucracy, but what do you say to critics who say this will make it hard for programs to retain staff or cover overhead costs like buses and maintenance?

Adams: The concern is understandable, but the 5 percent cap applies only to costs categorized as administrative and is in line with similar grants. Teacher salaries, classroom expenses, family services, and other costs that directly support services are not administrative. We’re also removing many of the federal requirements that generate administrative work in the first place. And because programs have different cost structures, the proposal includes a waiver process for those that anticipate undue implementation challenges, subject to HHS review.

Hess: You’ve proposed to require that classroom instruction be conducted in English. Critics argue that this doesn’t make sense, given that perhaps one third of participating children are dual-language learners. What’s your take?

Adams: Head Start is a school-readiness program. The goal is to help children enter kindergarten able to participate fully in English-language classrooms and the broader community. The proposal allows programs to request waivers from this requirement and Tribal Head Start programs are exempt when a tribal language is used to further tribal heritage. This is the kind of provision where public comment matters. We want to hear from programs and families about how to achieve strong English acquisition while serving children well.

Hess: Of the changes you’ve proposed, which ones address previous Obama- or Biden-era regulations—and which modify the 2007 law as written by Congress?

Adams: There are really two categories here. First, some of the regulatory provisions we’re taking on were added or expanded in 2016 and 2024. Our proposal gives programs flexibility to keep adhering to those provisions if they so choose, or try approaches better suited to local conditions. Second, some provisions introduced by previous administrations simply duplicated obligations already written into the Head Start Act or other federal law. Removing duplicative language doesn’t remove the underlying obligation. We are not changing the 2007 Head Start law—only Congress can do that.

Hess: Why do you think there’s been so much pushback against the proposed changes?

Adams: I understand why people care so intensely. Head Start has a proud history, dedicated employees, and parents who have built their lives around it. Any change to its rulebook is bound to raise questions. But some concerns falsely equate federal regulation with program quality. I’ve visited programs nationwide and met extraordinarily capable local leaders. Our approach starts from trusting them rather than assuming Washington knows best. It’s also worth remembering that this is a proposed rule. We released it for comment precisely because we want providers, parents, and experts to tell us where we got it right and where they think changes are needed.

Hess: OK, so what’s the biggest thing that you think critics are getting wrong?

Adams: The biggest misconception is that flexibility is a mandate. If a local program likes its current operating model, this rule doesn’t require it to change. Some criticism also implicitly assumes that local Head Start directors, teachers, parents, and state regulators can’t be trusted to make responsible decisions without Washington prescribing every detail. These are some of the most capable and mission-driven people I’ve worked with. Our job is to hold programs accountable for results while giving them room to solve problems in ways that fit their communities.

Hess: Over the past two years, various Trump officials and allies have talked about eliminating Head Start. Should observers view this as a half-step in that direction?

Adams: No. This is an effort to make Head Start more sustainable and accessible. As stated earlier, the proposed reforms are estimated to allow us to reach hundreds of thousands of more children over the next five years. I have said publicly that I believe in this program. The question is whether we want a Head Start system that becomes progressively more expensive and harder to access or one that trusts parents and local communities, focuses federal oversight on the essentials, and serves more children. We are choosing the latter.

This conversation has been edited for length and clarity.

Frederick Hess is director of education policy studies at the American Enterprise Institute and an executive editor of Education Next.

This post originally appeared on Rick Hess Straight Up.