The Education Gadfly Weekly

The backlash against the “college for all” movement was overdue. For too long, high schools treated a four-year degree as the default destination for too many students, without paying enough attention to how well students were prepared to embark on that pathway, the quality and costs of the colleges they might apply to, or the availability of alternatives that might be a better fit. But today we risk overcorrecting. The danger is replacing that overly broad default assumption with an overly confident effort to direct more students away from the path of a traditional four-year degree.

The growth of high-quality career and technical education programs, experiential learning opportunities, and other career-focused pathways has been an important and much-needed development. And there is good evidence behind the enthusiasm. Rigorous studies of Career Academies, technical high schools, and P-TECH have all found gains in earnings, graduation, and postsecondary attainment for participating students. But corrections can swing the pendulum too far in the other direction—and, as the history of K–12 education reform has shown, they often do.

As momentum gathers toward “multiple pathways,” we should be careful not to interpret modest grades, humble test scores, or other lackluster measures of college readiness as ironclad evidence that students would be better off pursuing something other than a four-year degree. A student’s academic record can help predict whether she will complete college. But the actual choice confronting that student is different: Would attending college leave her better off than pursuing her best available alternative?

One useful way to think about the problem is through the possibility of making mistakes in two different directions. Any rule we use to guide students toward different post-secondary paths will inherently produce some false positives and some false negatives. The first type of mistake is familiar. We encourage a student to attend a four-year college, but he struggles and ultimately drops out, incurring both financial and opportunity costs. This person may well have been better served by another pathway. That is a false positive: We recommended college when another option would have produced a better outcome. The other type of mistake runs in the opposite direction: we discourage a student whose academic record raises questions about their college readiness but who, with access to the right college and sufficient supports, would have completed a degree and benefited substantially. That is the false negative.

Much of the post-secondary debate in recent years has focused on the first error. However, recent research suggests we should take the second type of error more seriously.

A study of public four-year universities in Texas compared students just above and below admissions cutoffs, allowing researchers to isolate the impact of enrolling in college for students near the threshold. Those just above the admissions cutoff, and thereby eligible to enroll, increased their bachelor’s degree attainment by about 12 percentage points and eventually raised their earnings by roughly 8 percent compared to those students who fell just below the line and were not eligible to enroll. What is particularly important to keep in mind is that these were students near the margin of admission rather than the strongest applicants—meaning we would expect them to be at greater risk of dropping out.

A similar study in Georgia reached a similar conclusion. Enrolling in one of the state’s public four-year universities increased bachelor’s degree attainment and was estimated to raise household income around age 30 by roughly 17 percent, with larger benefits observed for students coming from lower-income high schools.

A note of caution: These studies apply to students who were just above and below admissions eligibility thresholds, so the findings should not be generalized to all students whose academic records might raise questions about their readiness to succeed in college-level courses. Still, they provide a useful reminder that students with a lower predicted probability of completing college do not necessarily face lower expected returns from attending.

For example, imagine a student whose academic record suggests only a 50 percent probability of completing a bachelor’s degree. That estimate is useful but, in isolation, tells us very little about whether college is the right choice. The decision also depends on a host of other factors, including which colleges are realistically accessible, eligibility for financial aid packages, out-of-pocket costs, the availability of wrap-around supports, the long-term payoff from completing, the opportunity costs of enrollment, and the alternative pathways actually available.

As important, the costs of the two errors may differ substantially in the long run. Sending a poorly matched student to college can mean debt, lost time, and foregone earnings. Steering away a student who would have completed can mean giving up a lifetime of higher earnings and other benefits generally associated with attaining a degree.

This is where the distinction between false positives and false negatives becomes especially useful. Any benchmark used to classify students as more or less likely to complete college changes the balance between the two errors. A more restrictive threshold may reduce the number of students who enroll and struggle, while increasing the number who are discouraged from attending even though they would have benefited. A more relaxed threshold shifts those risks in the other direction. That should make us cautious about replacing the old “college for all” model with another simplistic approach. If prediction cannot usefully tell us which path would be better for any given individual, then we should focus instead on helping students more carefully weigh the tradeoffs involved.

Research suggests that high school advising can be an important part of that process. Two recent studies are instructive. Bottom Line, an intensive advising program, pairs lower-income students with professional advisers who help them develop college lists, navigate the application and financial aid processes, compare options, and decide where to enroll. Advising continues into college for students who attend participating institutions. An evaluation found the program increased bachelor’s degree attainment, with much of the effect appearing to come from students enrolling at colleges with stronger average outcomes.

A second study examined a very different advising model. Across 62 Michigan high schools, seniors were offered an 18-week college-planning curriculum delivered largely by current teachers and school staff. The course addressed two- and four-year options, the costs and benefits of college, navigating application and financial aid processes, and broader career planning. The course did not increase college enrollment overall. Instead, it influenced who enrolled. Enrollment increased among students with stronger academic records and decreased among some students with weaker records. Ultimately, the program was found to increase longer-term persistence and associate-degree attainment among more academically prepared students.

That pattern is especially relevant to the current debate. The objective was not simply to push more—or fewer—students toward college. Better information and structured support appear to have helped some students enroll who otherwise might not have, while leading others to reconsider enrollment when it wasn’t the best available path for them.

But better advising should not simply mean giving high school advisors or teachers more discretion to decide who is “college material.” In Noise, Daniel Kahneman, Olivier Sibony, and Cass Sunstein describe how professional judgments can vary significantly across decision-makers confronting similar cases. Their proposed solution is what they term “decision hygiene,” which involves systematically assessing the relevant components of a particular judgment using the best available information before forming a conclusion.

Neither study noted above describes its intervention in those terms, but both embody a similar idea. They break a complicated decision about choosing a post-secondary pathway into a more manageable set of questions rather than relying on a single predictor or a general impression of whether a student is “college ready.” Applied to high school advising, even modest improvements to the process can improve the odds that a student will select the pathway that leaves them better off than the alternatives. Those judgments will still be imperfect. But with a more intentional approach, advisers can reduce reliance on incomplete information or overly broad assumptions.

Ultimately, the evidence points to a critical question that is often overlooked: How can we help students choose the post-secondary option that improves their long-term outcomes the most, while reducing the costs of errors in both directions? Answering that question depends on preparing students for multiple possible pathways and ensuring that valuable and accessible alternatives to a four-year college degree exist. Advising should help students evaluate those options using the best available information about the likely benefits, costs, supports, and fit, customized to their specific circumstances.

It also means preserving valuable options for as long as possible. When uncertainty is unavoidable—which will always be the case when discussing teenagers—keeping multiple viable pathways open may be more valuable than sorting students earlier.

“The turbulent AI era is here. The choices we make now are critical.” So begins Bill Gates, calling for the world to wise up to the risks we face as the technology rapidly approaches capabilities beyond human imagination and control.

It is not hyperbole to say that the stakes are high. The Pope has warned about the erosion of human ability to think. An Anthropic researcher publicly quit, citing concerns that the AI industry is acting recklessly given the “civilizational stakes” of its work. Leaders of the nation’s top AI companies appear to agree that the technology is potentially catastrophic, warranting a slowdown.

As a parent who watched tech leaders defend their predatory practices in a series of lawsuits over the past few months, I find this new call for safeguards among competing tech titans almost too hard to process.

It took well over a decade to see the alarming impact of smartphones, social media, and school-issued devices on children, and longer still to demand action to protect students from harmful effects. We clearly can’t wait that long on AI.

Alongside device bans and restrictions, school districts are racing to develop AI use policies—first piecemeal and now expansive. In New York City, the mayor announced a moratorium on student-facing AI use through eighth grade. District administrators in Los Angeles went further, apparently surprising even school board members with their total ban.

But AI use is about much more than whether a child can access Gemini on their Chromebook at school. In addition to increasingly clear impacts on learning, AI use raises complex questions about honesty, judgment, and whether struggling through difficult work has inherent value separate from the final product. Parents who believe a child’s moral education starts at home need to engage in conversations about when and how—or if—the technology should be used.

—

My eighth-grade daughter is taking “Humanities” this fall, an elective that apparently used to involve watching a lot of movies before a language arts teacher filled a last-minute vacancy and was given free rein to adjust the class as he saw fit. At back-to-school night, he discussed how our children are constantly inundated with media and shared his desire to help them learn how to view content with appropriate skepticism. (Amen!)

The teacher described a shift among students and families; it’s a change I’ve observed as well. Parents are more cautious about screen time, or at least more aware. Kids themselves know that social media is harmful (though they don’t stop using it when left to their own devices). My oldest child, born just as smartphones entered the chat and now a college freshman, went through the hollowed-out, tech-filled, high school experience described in Fordham’s report on high school in the age of distraction. For my youngest child, in her last year of elementary school, teachers seem noticeably more cautious about technology in their classrooms. Instead of seeking out ways to incorporate tech, they ask whether the task works without it.

This “techlash” is taking hold across the country. In Lower Merion Township, outside Philadelphia, parents are suing to retain the right to opt their children out of mandatory tech use in the classroom. Los Angeles preceded its new AI-ban with a sweeping screen-time policy passed over the summer, developed largely in response to parent demands. After years of heated conflict over partisan-coded issues like book bans and preferred pronouns, school leaders are hearing from parents across the political spectrum who don’t want their children sitting in front of screens all day, doing who knows what.

Action on screen time, driven by consensus, is a welcome development. Yet limiting the number of minutes a child spends on their school-issued device does not resolve the difficult questions—more moral than technical—that come with an AI future.  

Toward the end of our seven-minute block together at back-to-school night, my daughter’s teacher turned to AI and his enthusiasm disappeared. He was visibly disturbed by the dual reality that our children will live with AI tools and thus should know how to use them, while it is clear that the technology erodes cognitive development in ways we are only beginning to understand.

AI increases academic efficiency on certain assignments. Students may earn better grades for less work, but a recent analysis of closed-book test scores confirms that students using AI are not obtaining knowledge along the way. Meanwhile, some teachers are producing AI slop lesson plans and feedback, undermining student effort in exactly the moment when kids need to understand that effort is at the center of real learning, not the production of a polished final product.

The technology is changing faster than any state, district, school, or teacher can update their policies and procedures, which makes it exceptionally difficult to define what type of AI use constitutes cheating. Is it really that different from spell-check or a calculator? (Yes!) And even if kids want to avoid AI, their perceptions of peer use mean that every assignment brings a moral reckoning over whether the cost to not use AI is too high.

Few kids are prepared to fully grapple with these ethical concerns, especially when few adults can even begin to develop and articulate shared norms over how and when AI should be used.

As parents, our job is to support the development of our child’s moral compass, cognitive growth, and intellectual curiosity. At a minimum, we can ask our children about AI use and stay aware of their schools’ and teachers’ policies. We can push our kids to think about how and when technology should be used, and what happens when their moral convictions or family rules come in conflict with what others consider appropriate use. We can stress how shortcuts are limiting, even if peers take them to get ahead in the short term. Better—we can do all these things in partnership with the educators who guide their learning for many hours each day.

Children need to learn how to navigate technology with the confidence that they can reason, appreciate, converse, and serve their communities in distinctly human ways. Building this confidence requires teaching kids to work hard and think independently, without reliance on powerful tools that adults don’t fully understand and that their creators may not be able to control. Pressing “pause” on AI in schools is necessary but insufficient. Today’s students will not know a world without this new technology, and it’s up to the adults in their lives to equip them with the human capacity to debate and decide how it should be used. 

Private-school choice advocates, particularly in red states, have a lot to be excited about right now. Since 2020, more than two dozen states have either expanded or created state-level choice programs, which means that 18 states now have choice programs that are either universal or near universal and 27 have already elected to participate in the new federal scholarship tax credit, which will take effect in 2027. Taken together, these policy developments have created an unprecedented opportunity to tap into the demand for and increase access to a private school education, particularly among families who previously were unable to afford one. The question we face now is: How do we ensure that the families that want a good private or parochial school for their children can find one?

Until now, much national attention has focused on creating new schools. Venture philanthropies like the Drexel and VELA funds are both primarily (though not exclusively) focused on seeding growth among new schools, scaling existing schools and networks, or investing in low-cost innovation. This is important work, and over the long run the new school start-ups they support will add importantly to the supply of places for students to learn. But to ramp up quickly enough to meet today’s fast-rising demand, seeding and scaling growth should be two elements of a three-tiered effort to boost supply. Those tiers should include a coordinated push to fill the excess capacity in already existing schools; an investment strategy aimed at helping thriving existing schools and networks grow (i.e. scale); along with investments in new-school start-up and innovation to seed new schools and networks, particularly in areas where limited supply exists.

For those looking to ramp up quickly, we must acknowledge that the first tier provides the best and most cost-effective way to grow private schools quickly.

The fact is that excess private school capacity exists in communities all over the country, most heavily concentrated in the Catholic sector. For more than two decades, overall Catholic school enrollment, particularly urban Catholic school enrollment, has shrunk. A number of internal and external factors have contributed to the decline, including a dramatic increase in tuition-free competition from the charter school sector; an internal shift in staffing structure from one that relied heavily on religious men and women to one that required a far greater number of lay (and salaried) faculty, staff, and leadership; and a policy and philanthropic environment that has, for decades, funneled billions to seed growth and operations in the district and charter sectors, while the private and parochial school sectors were largely left out.

In short, the combination of internal financial challenges combined with an investment strategy that starved parochial schools while ramping up investments in competitors has constrained Catholic school enrollment across the country.

While recent private school choice policy victories haven’t fully leveled the fiscal playing field (parents choosing private schools will still get a fraction of the funds that public schools receive), they are an important step towards allowing private schools to compete on the same terms as their district and charter counterparts. If we add to that intentional investments in Tier 1 and 2 enrollment and capacity-building support, many communities will be able to quickly increase overall private school enrollment, which is critical to demonstrating the demand that does exist for more options.

To understand what’s possible, it’s worth looking more closely at a state that is a trailblazer in expanding access to private schools: Florida.

In August 2022, Florida’s Step Up For Students surveyed more than 2,000 private schools and found that participating schools were operating at 79 percent of capacity, with room to add another 84,442 students. Between 2021 and 2024, thanks to a combination of targeted philanthropic support, local efforts to attract more families, and strong local leadership, much of that existing capacity has been filled. Specifically, while the overall population of school-aged children (ages 5–17) increased by 4 percent between 2021 and 2024, private school enrollment increased by nearly 20 percent (more than 80,000 students). And these gains were driven largely by intentional efforts to fill vacant Catholic school seats.

Catholic dioceses across Florida invested heavily in family outreach. The Diocese of Venice alone launched a marketing campaign that sent more than a million texts and emails, helping families understand that a private education was now in reach. The Archdiocese of Miami invested in both strengthening existing academic programs and launching new dual-language programs, classical programs, and programs aimed directly at supporting students with special needs. They also expanded capacity by re-opening four schools that had recently closed.

These efforts paid off and helped drive a 19 percent increase in Florida Catholic school enrollment between 2022 and 2025. Jim Rigg, Secretary of Catholic Schools in the Archdiocese of Miami said, “at this point, there are areas of South Florida where there are simply no open seats in a Catholic school.”

Florida isn’t alone. Texas recently passed the largest Education Savings Account program in the nation, and the Texas Conference of Catholic Bishops worked quickly to ramp up Tier 1 enrollment and parent outreach support to Catholic schools across the state with the goal of increasing their enrollment by at least 20,000. Early evidence suggests that this targeted outreach and support is having an impact. As the new school year began, the Archdiocese of Dallas has projected it will add roughly 2,500 new students using Texas Education Freedom Accounts to attend Dallas Catholic schools in 2026–27. The Diocese of Beaumont projects a 15 percent overall increase in enrollment, and Corpus Christi estimates that enrollment will increase by 8 percent.

These early examples suggest how existing schools, supported by organized outreach and practical assistance, can respond quickly when new public funding makes tuition affordable. And the good news is that, unlike new school startups, growth in existing schools requires relatively modest, time-limited investments that focus on infrastructure improvement, systems development and support, and targeted outreach and culture building.

So if we want the supply of places in private schools to keep pace with the new opportunities, we need a national strategy that arms existing under-enrolled schools with the strategic investment and targeted technical assistance they need. And we need to understand that that’s far less costly than starting schools from scratch. There’s no land acquisition. There are no complex leases to negotiate. There’s little or no new bricks-and-mortar required, just some retrofitting and modernizing work that can be done relatively quickly.

Of course, investment in filling existing seats doesn’t preclude the need to start new schools in targeted markets. But if the aim is to make choice real for more families now, the first question in every state should be simpler: Where are the good schools with room to grow, and what will it take to help them find the families newly empowered to choose them—and vice versa? 

The National Center for Education Statistics (NCES) regularly publishes reports using results from NAEP testing to compare the proficiency standards that states set for their students. Since standards vary across states, the results of state assessments can’t be used to directly compare students’ progress from place to place. But by placing a state standard or cut score onto the NAEP scale, which is a common metric, a NAEP-equivalent score of that standard can be generated and compared across states. NCES’s latest Mapping State Proficiency Standards report, the ninth in the series, puts this method to work on student achievement results from state and NAEP assessments administered in 2022. Because states are required to participate in NAEP reading and math in grades 4 and 8, the assessment also serves as an external audit of state expectations—one that sits beyond any individual state’s control.

The latest iteration maps state expectations in fourth- and eighth-grade reading and math onto NAEP’s 0–500 scale, showing whether they line up with NAEP Basic or NAEP Proficient levels. There is some fluctuation between findings of the previous report in 2019 and these 2022 data, but not a great deal. More on that in a moment.

Let’s look at the most recent levels first. Only three states were equivalent to NAEP Proficient in both subjects and both grades in 2022—Illinois, Massachusetts, and Rhode Island. In grade 4 reading, 48 states had standards at or above NAEP Basic. Of those, six states had standards at the NAEP Proficient level—Alaska, District of Columbia, Illinois, Massachusetts, Oklahoma, and Rhode Island. Two states—Iowa and Virginia—had standards below NAEP Basic. (Make no mistake that the Old Dominion’s unacceptable standing on this metric is motivating serious work from those of us on the state board!)

In grade 8 reading, 48 states had standards at or above NAEP Basic and six of them were at NAEP Proficient—Illinois, Kansas, Massachusetts, Rhode Island, Tennessee, and Wisconsin—while two states—Iowa and Virginia—had standards below NAEP Basic.

In grade 4 math, 50 states had standards at or above NAEP Basic level. Thirteen had standards at NAEP Proficient—Alaska, Colorado, District of Columbia, Illinois, Kansas, Maryland, Massachusetts, Nebraska, New Jersey, North Carolina, North Dakota, Pennsylvania, and Rhode Island.

For grade 8 math, all states had standards at or above NAEP Basic. Ten of them reached the NAEP Proficient level—Alaska, Arizona, Illinois, Indiana, Kansas, Massachusetts, Oklahoma, Pennsylvania, Rhode Island, and Wisconsin. The authors also looked at the Smarter Balanced Assessment Consortium states as a whole and found they mapped to NAEP Basic at each grade and subject.

Now for those short-term changes. Since 2019, for reading in both grades, movement was mostly upward (e.g., D.C. and Oklahoma went from Basic to Proficient), but a few states slipped from Basic to below Basic, including Virginia (sigh).

Grade 4 math saw modest upward movement, including Maryland and Colorado, which went up from Basic to Proficient. There was more churn observed in grade 8 math as, for example, Indiana and Oklahoma rose from Basic to Proficient, while Massachusetts and Arizona fell from Proficient to Basic.

What to make of all this? The National Assessment Governing Board, which oversees NAEP administration, sets a high bar for NAEP Proficient at each grade and subject tested. States, of course, are not compelled to set their own standards that high—and, as you can see, most don’t. Reasonable people can debate, and many have, whether NAEP Proficiency is best understood as a college-ready benchmark and, similarly, whether NAEP Basic is the more appropriate target for all students. But the useful takeaway from this crosswalk is not that every state should immediately try to leap to NAEP Proficient (phased approach, anyone?). It is that states can easily compare where their standards sit on a common scale and see how much room there is to keep raising expectations. Here’s hoping the tenth edition shows more states doing just that!

SOURCE: Darrick Shen-Wei Yee and Brian Cramer, “Mapping State Proficiency Standards Onto the NAEP Scales Results From the 2022 NAEP Reading and Mathematics Assessments,” Institute of Education Sciences, National Center for Education Statistics.

The importance of childhood investments to the trajectory of life is commonly accepted, but less is known about the size and sources of disparities in these investments. A recent study in Nature Communications sought to assess differences in childhood human capital investments across different racial and socioeconomic demographics.

To expand the scope of investments across multiple areas of children’s lives, the researchers combine data from ten nationally representative surveys from 2010 to 2023 and convert public spending, private expenditures, and family time into a common dollar scale. The ten domains include informal education, nutrition, housing, formal schooling, health care, exercise, clothing, childcare, transportation, and college prep. For example, time-based investments, such as informal education and transportation, are converted using hourly wages, while expenditure-based investments—food, clothing, housing, childcare, and college preparation—are measured through costs.

The researchers then evaluate trends across socioeconomic quartiles and racial groups and identify three key findings: (1) disparities between groups are substantial but smaller than estimates reported in previous research; (2) the gaps are largest before children enter formal schooling; and (3) those disparities vary by race, ethnicity, and income.

For finding one, the average investment per child from birth to age 18 (counting all public and private spending) is $502,152. Total investments are approximately $86,000 more for children in the top quartile of income compared to the bottom quartile, and between $55,000 and $75,000 more for White families compared to Black and Hispanic families. These racial and socioeconomic investment gaps, which range from 6 percent to 15 percent, are much smaller than found in previous studies.

Investment gaps are widest between birth and age five and converge as children enter formal education. Notably, the investment gap between the highest and lowest socioeconomic quartiles is cut from 24 percent to 12 percent as children transition from early childhood (birth through age four) to older childhood through twelfth grade. Researchers identify that publicly funded investments in K–12 teacher salaries are comparable across groups, which contributes to the investment convergence in tandem with increased support for groups with lower investments during early childhood once they enter K–12 schooling (e.g., tutoring and special education funding).

Finally, investment gaps present differently across various racial and income groups. Because housing is the largest contributor to investment disparities for Hispanic families, the Hispanic-White gap is larger for expenditure-based investments than for parent- and family-time investments. By contrast, because Black children on average receive less family time during meals, the Black-White gap is larger for parent- and family-time investments than for expenditure-based investments. The contributors to disparities across household income are similar to, but not identical to, those across racial groups, further reflecting variability in how disparities present.

Put together, the evidence paints a vivid picture of the investment disparities that exist among children. Through examining a more comprehensive set of investments, researchers more accurately identify which disparities are greatest among groups, and which investments increase or decrease disparities.

In all, this study offers a more comprehensive picture of childhood investments and their disparities. While the gaps are smaller than those found in previous research, they remain significant before children enter formal education. These findings could help policymakers target support more effectively and examine how early investment gaps shape students’ readiness for school.

SOURCE: David Blazar, Michel Boudreaux, Steven Klees, Jennifer King Rice, Marvin Titus, Jiehui Zhao, Disparities in childhood human capital investments in the United States (Nature Communications, 2026).

Mike Petrilli is back on the mic solo this week to weigh the case for and against banning AI in schools. He considers growing concerns about cheating, an erosion of grading standards, and the broader effects of technology on student learning, as well as arguments that well-designed AI tools could still have a useful place in the classroom. Are blanket bans too extreme, or is it time for schools to press pause?

Then, on the Research Minute, Brian Fitzpatrick examines a new systematic review of state and locally funded pre-K programs and what it reveals about the strength of the evidence behind their short- and long-term effects.

Recommended content:

• Let’s debate the AI bans —Michael J. Petrilli, SCHOOLED
• High Tech, Low Standards: High School in the Age of Distraction —Brian R. Fitzpatrick, Thomas B. Fordham Institute
• Schools Are Urged to Embrace AI—and Ban Phones. Can We Resolve the Tension? —Rick Hess, Education Week
• Hot Takes on NYC AI Ban —Ben Wallerstein, Whiteboard Advisors
• A Systematic Review of Research Designs Used to Evaluate State and Locally Funded Pre-K Programs —Anamarie A. Whitaker, Cara L. Kelly, Tyler W. Watts, Mark Lipsey, Juan Camilo Cristancho, Jade M. Jenkins, Caroline M. Botvin, Drew H. Bailey, Greg J. Duncan, Tania Maria Cruz Cordero, Melissa Stoffers, and Mindy Rosengarten, EdWorkingPapers (August 2026)

Feedback Welcome: Have ideas for improving our show? We would love to hear them. Send them to [email protected]

---

Transcript

This transcript was generated with the assistance of AI and lightly edited for clarity and readability.

Introduction

Mike Petrilli [00:00]

Welcome to The Education Gadfly Show.

I’m your host, Mike Petrilli of the Thomas B. Fordham Institute.

This week, I go solo to talk about how schools should approach artificial intelligence and whether student-facing AI belongs in the classroom at all.

Then on the Research Minute, Brian Fitzpatrick reviews a new systematic review of state and locally funded pre-K programs and what it reveals about the strength of the evidence behind their short-term and long-term effects.

All that and more this week on The Education Gadfly Show.

Hello, this is your host, Mike Petrilli of the Thomas B. Fordham Institute, here at The Education Gadfly Show and online at FordhamInstitute.org.

It is my monthly solo edition, and today we are going to talk about AI. Not so much whether it’s going to kill us all, which, in the last few days, sounds like it very much might. But whether, in the meantime, we should embrace AI bans for our schools.

Let’s do that on Ed Reform Update.

Ed Reform Update: Should schools ban AI?

Mike Petrilli [01:20]

So, first I should say this thing about the AIs killing us all. My younger son came home the other day really nervous about everything that has been going around in the news. And it is sobering, I’ve got to say, the wall-to-wall coverage of these people building the AI saying that there’s a greater than 10 percent chance of it killing all of us within ten years.

All I could say to him as a father was, look, when I was your age, we were all very worried about nuclear annihilation and thought that was very much a real possibility. This is in the 80s. We would do these silly things, as many of you can remember, where we would practice hiding under our desks, as if that would have helped us at all had World War III started somehow or another.

And maybe it was partly luck, and partly it was the skill of our leaders. So far, knock on wood, that has not happened. And so I guess we have to hope for the same when it comes to AI and hope and pray that people who know more about this stuff than I do, and who are in positions of responsibility, are able to figure out this problem.

And look, I’m an optimist. I’ve got to believe that we will, just like other big challenges that we humans have faced. We are a pretty clever species. But we understand why people are nervous.

But that’s not what we’re going to talk about, really.

We really are going to talk about this question of the AI bans. And it’s remarkable. It was only back in April that my good friend Rick Hess interviewed another good friend, Michael Horn, for his Ed Week blog. He gave the post a very provocative title. It said, “Schools are urged to embrace AI and ban phones. Can we resolve the tension?”

And it was a good question, right? Because that was happening not so long ago.

And yet we have resolved the tension. We’re resolving it by banning the AI also. At least that is what New York City has done for grades K–8 and Los Angeles for grades K–12.

I understand from some folks like Erin Mote that New York’s ban is a little more nuanced. They left some room for some experimentation and some learning. And again, it’s only K–8. Los Angeles, it’s more really shutting it down pretty much absolutely, though again with some carve-outs in some cases, maybe tools for kids with disabilities, for example.

But Michael argued back then, he said, we should be very wary about promoting blanket bans. Not only is it out of tune with the reality that some parents are intentionally opting for schools that leverage AI, but it will also limit the use cases where ed tech makes sense in traditional schooling models.

He points, for example, to Coursemojo, now led by Dacia Toll, who was the founder of Achievement First out of New Haven, Connecticut, I believe. Some amazing no-excuses charter schools back in the day.

So, these are people that care a lot about learning, student achievement, high expectations. And Michael writes that Coursemojo is a curriculum-aligned teaching assistant for English language arts and that Dacia is doing this work with a clear instructional design, guardrails, clarity around the importance of knowledge and human teachers setting the rules around good instruction, good questions, and actual student misconceptions.

We’d be foolish to dismiss the opportunities that someone like Dacia sees in these tools.

And of course, others with close ties to the ed tech world, the AI world, like Ben Wallerstein at Whiteboard Advisors or Erin Mote herself, have been making the case that basically we shouldn’t throw out the baby with the bathwater.

I’ve been writing about this and getting some feedback.

Listen to, for example, Lindsay Dworkin, who is a senior vice president for public policy at the textbook giant HMH. She says a student asking a chatbot to write an essay for them and a teacher using an AI-enabled tool to understand where students are struggling and what support they need may both be technically using AI, but they’re doing fundamentally different things.

My concern, even with governing by broad categories, is that those rules can’t distinguish between technology that pulls students away from learning and technology that helps teachers deliver stronger instruction or help students engage more deeply. We should be differentiating. This is not a monolith.

James Cantonwine wrote in. He’s the director of research and assessment at a small school district in Washington State doing some cool stuff.

He and I have been going back and forth about basically an app he built where teachers can videotape themselves and then get feedback on their teaching aligned with some of the best teaching evaluation tools out there. Again, homegrown. Very cool.

He says he’s not necessarily defending AI, especially student-facing AI, but he does ask some good questions. We know that generative AI encompasses more than the chatbots that we’re used to, Claude and ChatGPT. Where do we draw the line?

Vendors of all types have been racing to integrate AI into their products, and they don’t necessarily let schools opt out of these new features that may be AI-embedded. What happens? Do schools have to cancel all of those contracts if a ban goes into place? What if this AI stuff is pretty tangential?

And then just the basic question of, look, we want to keep experimenting, right, to try to understand what works and what doesn’t and how to keep this stuff safe. And so who’s going to do that?

Even the American Psychological Association, which tends to be pretty small-c conservative on a lot of this tech stuff, says that AI bans are too extreme.

All right. So there’s an argument out there that says we really need to be wary of these blanket bans.

But I’ve got to say, I think it’s really hard right now for these groups to be out there saying, hey, we should trust the system. Right?

I mean, especially with all these concerns about runaway AI killing us all. Okay, but not just that.

We know that parents are leading this backlash. I think that is very much a grassroots thing. I think a lot of these parents are frustrated with the invasion of technology into their schools.

Some studies show the typical school district has literally thousands of ed tech products being used inside. Who knows what’s happening out there? A lot of it questionable and a lot of it doing some pretty dangerous stuff.

So, for example, Betsy Tao, who chairs the Technology Committee for the PTSA in my home district of Montgomery County, Maryland. She wrote in to say this.

She said AI moratoriums are a necessary step for all educational institutions for one simple reason: “must first do no harm.” Looking back at the last twenty years, at the unimaginable amount of damage that social media and smartphones have wrought, especially to our children, and the complete and utter failure of ed tech on educational outcomes, to take any other position is unconscionable. Naive.

And she says, look, we have to contend with the fact that it doesn’t matter whether school districts officially approve or don’t approve the product, right? Kids can access all of it. This is crazy.

She says, “I can still play the game Five Nights on Epstein’s Island with my fourth grader’s school account. And last week, I found a link easily on the first page of a search that worked.”

Okay, I don’t know much about this game, but Five Nights on Epstein’s Island does not sound like a game you want your kid playing. Right?

And this is after she’s been trying to get the school district to work on their firewalls and lock this stuff down. Common Sense Media just said that Google’s AI Overview has an unacceptable risk to children, and yet there it is. It’s in all of our schools.

She says a Baltimore dad I know has reams of data showing his son used his school Chromebook to do all of his math assignments over a period of months in between searching for entertainment and, of course, porn on a school device, and that he used AI for those assignments.

So she asked, what are we doing here?

I mean, this is the thing, right? We see this as parents. We see that the kids are using AI to cheat. They’re using it to get to all kinds of crazy materials. The Chromebooks are just basically big phones, distraction devices. Right?

And we talked about this last week on the podcast when my colleague Brian was here talking about his new report on a survey we did of high school students. Even the students themselves are admitting, screaming from the rooftop, saying, hey, all these screens, the AI, the erosion in grading standards, all of this has just cratered academic expectations in their schools.

The bullshitification of high schools, especially, is on full blast.

So I look at all this, and I think I start to say, I don’t know, maybe I’m a Luddite. Let’s just throw it all out. Throw out the baby, the bathwater, the tub, the whole shebang. Go back to paper and pencil, human thinking, human relationships. Make our schools tech-free zones. Right?

I don’t know. Maybe that’s too extreme.

Another great perspective in the last few weeks was from Helena Bennett, writing at the Slow Boring Substack. That’s Matt Yglesias’s Substack. She was reading a new MIT report that she thought was really thoughtful, trying to figure out, as it says, AI use in teaching, learning, and research training.

And it kind of walks through the various purposes that you might use AI. And then also, what is it that a given educational institution is trying to achieve that’s different for higher ed maybe than K–12? It’s different for, say, a career and technical training program working on career stuff than it might be on liberal arts.

She says that public K–12 systems are designed to produce in students the basic cognitive capacity to read, think, reason, and work through problems independently. That is the exact thing that unsupervised AI use is best at overriding in a young person’s development. It takes away the struggle necessary to grow both intelligence and confidence.

So New York’s response with this ban is an attempt to preserve the identified purpose of a child’s foundational education. Seems like it makes sense.

And of course, we’re talking about all of this on the heels of the PISA results coming out last week. Andreas Schleicher, for one, certainly raised the concern that AI could be depressing student achievement scores around the world.

When you dig into the OECD report, it says that students who never used AI scored 509 on the science test versus 481 for students who use it very frequently. That’s a big difference, though it did look like students who used AI once or twice a week outperformed those who used it less often.

So look, correlation is not causation, but this is yet another reason to be worried.

So this has been rather rambling. And I hope you can tell that I have mixed feelings about this. I’m not quite sure where I fall.

I like the basic notion that, hey, let’s just stay focused on student achievement. Let’s look for tools that boost student achievement. Let’s get rid of the ones that don’t. Let’s worry, of course, about distraction. Let’s worry about cheating because obviously kids aren’t going to learn anything if they don’t have to put in any effort.

Right?

But it does seem to me that the burden, which is growing heavier by the day, does belong to these ed tech companies, to the other people selling these products with AI embedded in them, and to the people who are the AI evangelists.

They need to show us those AI tools, especially when it comes to student-facing tools, that really are adding value to learning without replacing thinking and convince us that ordinary American schools, right? Not in some dream world, but in the real world, that they can enforce and notice and spot the difference between the good stuff and the bad stuff and have some mechanism of keeping the bad stuff out.

If they can’t, then I have a hard time not saying that other districts should do what New York City and Los Angeles did and press pause on all of this until we figure it out.

That’s where I’m at. Let me know what you think. I’d love to hear your perspective.

All right, people. Now it’s time for everyone’s favorite, Amber’s Research Minute.

Research Minute: What does the evidence say about publicly funded pre-K?

Mike Petrilli [14:36]

Brian, welcome back to the show.

Brian Fitzpatrick

Thanks so much for having me, Mike.

Mike Petrilli

Brian Fitzpatrick, who is filling in for Amber. You, of course, were here last week, so people are used to you. But this time you get to do the Research Minute.

And joining us, as always, David Griffith, my co-host. Welcome, David.

David Griffith

Hey, thanks for having me.

Mike Petrilli

Yeah, you guys missed my solo podcast, which was me rambling on about AI in schools and whether we should ban AI.

I am really torn. I’ve got to say, though, we’re having a very interesting internal debate at Fordham where it just feels overwhelmingly like people are mad at the robots. And not just because they’re going to kill us, but because they’re messing up our schools already.

I mean, that seems to be the sense I’m getting, is people saying, what the heck are we doing with this AI? Get it out. Get it out of the schools.

David Griffith

I think with regard to education, that’s probably the vibe right now, Mike. I don’t know. I have to say I’m a little mystified more broadly by the backlash. I’ve yet to find a physical ailment that it doesn’t help me diagnose.

But people are really angry about it. So I guess we should take their concerns seriously.

Mike Petrilli

Yeah, it’s just hard because it’s not easy to actually get rid of it at this point because, as in our normal daily lives, it’s kind of wormed its way into everything.

It wouldn’t even be easy to go back to textbooks at this point. I mean, we’ve got this whole infrastructure we’ve built that is based on Chromebooks and iPads and learning management systems and these various digital tools, and they’ve embraced AI.

David Griffith

Yeah. Yes, it’s difficult. Certainly, it’s coupled with one-to-ones in practice.

Mike Petrilli

Yeah, yeah.

All right. Well, I don’t want to take more time away from Brian’s minute. Brian, you’re down to nineteen seconds at this point.

No, just kidding. It’s been a long time since anybody has actually hit the minute mark. I think in the early days, Amber made an effort until we decided it was ridiculous.

We’ll give you more time, Brian. What you got for us this week?

Brian Fitzpatrick [16:44]

Well, this week’s study is relatively straightforward, though from the looks of it, it took a monumental amount of effort.

So this is a twelve-author study that was just posted as a working paper, led by Anna Markowitz of the University of Delaware. And the twelve names are big names, including Greg Duncan.

The paper itself reviews essentially the work that exists now on publicly funded pre-K and attempts to do a synthesis of all the research methods being used. It is not really a classic review piece where it’s trying to come up with a single consolidated number for the average impact of pre-K, but it really goes through the history of research on pre-K.

And I’ve got to tell you, I found it very convincing.

It is, I think, making the case that on pre-K, like with many other topics, there is almost like a pyramid of research studies, where at the top you have a small number of extremely high-quality studies, so RCTs or something like them. And then kind of as you go down, you have studies of lower and lower quality, but they are more common.

You have dozens and dozens of studies doing regression models with some good controls and even more doing regression models with no adequate controls. And the lower the quality of the design, as is often the case for a number of different topics, the more of a purported benefit there is to the intervention being studied.

And I think I would summarize my kind of aha moment while reading this piece as: A week ago, I think I would have summarized the research on publicly funded pre-K as good but not great evidence for some sort of benefit for students. I think as of today, I would summarize it as the best evidence we have is mixed, and there’s not a clear case that the next one of these programs to open will benefit students.

And really, there are four gold-standard-style studies. Two of them found no results for students in the long run, or mid-tier run, I should say. Not long run, nothing like post-high school. One found positive benefits. That would be in Boston. And one found a profoundly negative impact on students. That would be in Tennessee.

And I think they make a solid case that, A, that’s the best way to summarize the work we have on pre-K as of right now.

And then they go into the short-term versus long-term stuff, which is really important when looking at pre-K. But let’s pause here and talk about their main point before we move on.

Mike Petrilli

Well, take that, David.

David Griffith

Why do I have a question coming my way, Mike?

Mike Petrilli

How can you live with yourself?

No, just kidding.

No, that’s really interesting, Brian. And kudos to these folks, including some people that have been pre-K advocates. Right? So this is very impressive.

I am curious, what about for low-income kids? Can we see any kind of particular impact on spending money on pre-K for them?

Brian Fitzpatrick

So this is hundreds of studies they’re kind of mashing together. They didn’t try to do any breakouts by type of students impacted.

They did kind of dig in a little bit to, like they say pretty explicitly, they think that the difference between the Boston wonderful outcomes for students and the Tennessee truly negative outcomes for students is not a measurement issue. It’s that there are differences in the programs. But they don’t really say why.

They just say that we think Boston just nailed it, and Tennessee dropped the ball.

Yeah, I couldn’t speak to an actual group of students targeted.

Mike Petrilli

Of course, that universal versus targeted question is a big one.

Look, and this is a tough one for us here at Fordham. To the extent that we have engaged on pre-K, Checker has done a lot of this. He has been making the case for a real school readiness, education-focused pre-K.

And yet it’s been many years now. I remember Russ Whitehurst, who is no slouch from a research perspective or high-expectations perspective, saying, look, the evidence is not great. And therefore, we should really focus on child care.

That there are benefits to the labor market, to parents being able to work and bring in income, to having child care. But we shouldn’t worry so much if that child care is all that educational or not, as long as the kids are being well taken care of. And it can be a lot less expensive if you just try to focus on child care.

And maybe that’s the takeaway here. I don’t know.

But David, get in here. Tell us why this group of twelve experts got it wrong.

David Griffith [21:42]

I don’t think this group of twelve experts did get it wrong, Mike.

I mean, for starters, I think it would be folly for me to take a direct run at them in the next thirty seconds.

Let me just make a couple of points.

My understanding is that the Tennessee pre-K program we’re evaluating that is sort of driving the headline negative effects, you know, we’re evaluating the program based on very early performance, like the first couple of years, the first couple of cohorts of kids who went through.

And if we had stopped there with charter schools, we would have essentially no charter schools. Right?

I mean, it is well established that it takes five years for a school to, you know, three to five years, approximately that long, for a school to really get up and running. And so I’m sure it was a mess when it first got up and running.

So it’s not to offer excuses, but it is a little bit to say I am skeptical of dismissing any program, even the Tennessee program, based on the first couple of years before people have had the chance to evaluate it, before there’s been a cycle of improvement feedback. Staff are still learning their jobs. I really think that is a little premature.

Is it true that RCTs are going to show us smaller effects? I think that’s almost certainly true.

And I think it’s probably the case that, in the long run, the early literature on this is overstating the effects. Parents are getting, I think, the counterfactual is getting better. Right?

But I will say, I don’t agree, as much respect as I have for Russ and my brief interactions with him, I don’t agree with the child care argument.

And so anyway, those are my thoughts. I don’t reject the study. We absolutely need more evidence.

Mike Petrilli

Yeah.

All right. And then you were going to get to the long-term stuff, Brian, right?

And here’s the point: The effects fade, but that’s the case for everything, right? I mean, don’t we think it’s not just pre-K? Any intervention, probably at least the achievement effects fade, and what we hope is that there is some residue.

What was it we talked about once on the show? Social science dark matter. That there’s some residue that remains that maybe has a positive long-term impact on graduation rates or some kind of, you know, what we call non-cognitive skills, or there’s some benefit, even if it gets sort of pulled back down to whatever they were going to be anyways based on the kid’s background, yada, yada, yada.

I don’t know. I mean, is that basically what they’re saying here? It’s a fade-out story?

Brian Fitzpatrick [24:54]

So they don’t really delve.

Because of this, the studies past the end of pre-K span the gamut from end of kindergarten as a time point to measure outcomes to, you know, 22 years old in some cases. And they have to kind of match them together for the sort of methods they’re doing.

They don’t really speak to the specifics of fade-out.

One specific point that they make, however, is the downside of using end-of-pre-K scores specifically. Because I think, and I agree with them here, that if you’re going to look at, like imagine a hypothetical scenario where Timmy and Tommy are twins. Timmy goes to pre-K for some reason, Tommy doesn’t.

If Timmy’s far ahead of Tommy at the end of pre-K, but at the end of kindergarten, they’re at the exact same place, I don’t think we should summarize that as, oh, the evidence is mixed. We should summarize that as there’s no evidence for an effect here.

If by the end of kindergarten you’ve seen that fade-out, I don’t think that’s really fade-out. I think that’s evidence of no effect.

And I don’t think that’s a controversial claim. But I would encourage anyone listening who’s going to do a study of pre-K, just make your short-term research outcomes end of kindergarten.

And then if it fades out by the end of third grade, I think that’s a different conversation. But I don’t think we should be using end-of-pre-K outcomes here because I don’t think they’re a reasonable outcome.

David Griffith

I guess I’ll also say, Mike, I think what you said is broadly speaking correct. That’s the right way to talk about it, right? Like the universe is just sucking people back toward their predetermined social outcomes, unfortunately.

But I will say, when you dig in, there are a couple of really mysterious movements in this incredibly small handful of studies that we’re actually talking about here.

You’ve got the Boston study where essentially there’s no effect on middle school test scores, I think, but then you have these long-term big effects on non-academic outcomes that appear out of nowhere.

In the Tennessee study, you see basically the opposite thing, where you have these negative effects. But I just think the authors are right. We need more evidence. We don’t really understand.

My reading of what they’re saying is that there is fairly consistent evidence for short-term effects on academics, but there is a very concerning mix of effects, or no effects, when it comes to longer-term outcomes. And it’s concerning.

So anyway, I think I agree with their call for more evidence.

Mike Petrilli

Yep. And look, again, the question is, well, what else could we be spending this money on if it’s not having an impact? Lots of things that you could put on that list.

But then I can also imagine, like I said at the top, there is this work issue. I mean, there are a lot of reasons, and people on the right will make this argument, that having families increase their market income, making money not from welfare programs but from work, seems to be a really good thing for those families and their kids, and ending the dependency and et cetera, et cetera.

And so if making it possible for especially mom to be able to leave the house and work and have Timmy and Tommy taken care of by caring adults, all of that has some benefit as well that we’re not looking at here based on the impact directly on Timmy and Tommy educationally.

It’s complicated.

David Griffith

Timmy and Tommy, they’re featuring prominently in this conversation.

I can only agree, Mike. I’m probably showing my D.C. bias here, right? I can barely conceive of the counterfactual to universal pre-K. It’s like, I don’t even know what that would mean for, you know, I guess it would mean we’d be paying for it ourselves, right?

Which, from the perspective of the public purse, there’s a case for that. But if your point is education in these years doesn’t matter, it’s not the same point. Right?

So, yeah, it seems like a lot of people are working in these years. And so that calls for something.

Mike Petrilli

All right, gang, good discussion.

Really interesting, Brian. Thanks for bringing us that review.

That is all the time we’ve got for this week. So until next week.

David Griffith

I’m David Griffith.

Mike Petrilli

And I’m Mike Petrilli of the Thomas B. Fordham Institute, signing off.

The Education Gadfly Show is a production of the Thomas B. Fordham Institute, located in Washington, D.C. For more information, visit us online at FordhamInstitute.org.