Amid the “techlash” tumult, there's no shortage of people saying "why not just focus on evidence?"

Turns out it's not so simple.

Plenty of products have real evidence behind them—including studies vetted against ESSA standards by researchers at respected institutions like Johns Hopkins University—and it doesn't seem to move the parent advocates and academics stoking outrage (often on the very platforms they deride).

Part of the problem is that the standard of proof depends on who's making the claim.

Serious journalists hold a high bar for the data that informs a story, and they should. But that bar can, at times, make it hard to tell both sides of the techlash story. An education reporter I admire told me recently that he'd write about the incongruity between parental concerns about a product and the outcomes achieved by students who use it—if only he had an academic study to prove the data. 

I appreciate the rigor and risk aversion. But some reporters seem to apply the no-vendor-data standard with vigor while simultaneously publishing parent claims grounded in little more than vibes. Anecdotes and outrage clear the editorial desk. Corporate data doesn't.

As I wrote last week, this asymmetry wouldn't be so problematic if it wasn’t driving state and district decision making. 

Leaders trying to craft edtech policy with precision face the same bind as reporters covering the debate: They feel like they can't trust (often over packaged) data from companies, but third-party research is slow, expensive, and often stale by the time it lands.

The good news is that plenty of companies are awash in outcomes data. (Those that aren't should build better reporting capability or get out of the business altogether.) My hypothesis is that journalists and policymakers would actually use it if it were presented consistently and objectively—with equal clarity about what it is and what it isn’t.

There's a model for this. In accounting, independent parties offer attestation services to help build trust, reduce risk, and comply with regulations. These come in three forms: An examination, where the practitioner independently tests the underlying evidence—recalculating, inspecting, confirming—and signs a positive opinion at the same level of assurance as a financial audit. A review, where the work is lighter: inquiry and analysis, concluding that nothing came to the reviewer's attention suggesting the claims are wrong. And an agreed-upon procedures report, which offers just factual findings, no opinion. 

In the edtech context, examinations may be a bridge too far—at least for now. But reviews, paired with a disciplined set of critical questions about what the evidence is and isn't, would be a leap forward. Company-specific reporting modeled on attestation standards, reviewed and signed by qualified academics and statisticians (like, say, Elizabeth Tipton) would be no substitute for formal research, but they might still offer a way to get accurate, actionable data faster. Even a procedures report that offers a credible articulation of what is and isn’t happening could be a valuable signal.

That hybrid would give us an honest view of the gaps, and help to surface critical questions. “Data is promising, but only in a subset of schools.” “Results might appear positive, but there is no evidence that the instrument is actually valid.” “Growth accelerates with usage, but product effects can't yet be isolated from other site-level variables.” Implement a product in eight schools of a 25-school district and get mixed-but-promising results and today, we have no honest way to talk about that. This would give us one.

Here's what it would take: an initial cohort of seven to 10 companies willing to open up their data. A distributed bench of researchers. Legal clarity to enable matching of outcomes to demographic data. I’d propose that this be industry funded (because they are the ones with both incentive and resources), but pooled to remove potential bias. With an advisory group of district and state leaders to poke holes in the first round of reports.

More, and more variable, reviews of outcomes data may also help both districts and vendors understand the impact of context on product outcomes. On paper, running is a perfect way to get in shape, unless you have bad knees. Context is often a difficult variable to account for, even in expensive, long-term studies, yet it can make or break real-world efficacy. Or as a superintendent once told me: “I'd rather have a bad intervention implemented well than a good intervention implemented badly.”

This doesn't solve everything. But it gets real data into the discourse—and given the pace that bans, policy, and technology are moving, speed matters.

This week, the U.S. Department of Education (ED) released its “Learning Agenda Playbook,” a resource which packages together tools states can use to build their own multi-year education research agendas, so they can make evidence-based decisions that improve student outcomes. The Playbook comes as part of the Trump administration's broader push to move education decisions out of Washington.

ED, through its Office of Elementary and Secondary Education and the Institute of Education Sciences, built four components with input from Arkansas, Georgia, Idaho, Kansas, New Hampshire, Tennessee, Utah, and Wisconsin:

R&D advocacy organizations have spent years making the case ED is now echoing. The Alliance for Learning Innovation, a coalition of more than 150 education nonprofits, funders, and researchers, has been pushing for the federal government to invest more seriously in education R&D infrastructure at the state and local level since 2023. However, ED’s Playbook doesn't come with any new funding. States that already have research shops and data systems will move quickly; the majority of states that don’t, however, will need to build capacity before they can take full advantage of the Playbook.

Most AI tools in K-12 classrooms are still judged by adoption, engagement, and how much time they save educators. K-12 needs a different yardstick, one grounded in learning science and close enough to everyday classroom practice to show whether AI is actually improving teaching and learning. 

Next week, join us for a webinar, where we’ll unpack a new evaluation framework and explore what districts can measure beyond simply adoption and time savings. Attendees will leave with a clearer understanding of how better measurement can lead to more meaningful uses of AI in classroom instruction as we kick off a new school year.

🗓️ September 16, 2026 at 11 a.m. ET

💡 Featured Panelists

The next generation of college students has greater mental health needs and different expectations as to when, where, and how they receive support, according to a new survey commissioned by Uwill. Less than half (41%) of high school seniors rate their mental health as “good” or “very good,” and one in four (28%) say their mental health has declined during the past three years. They are also less likely to trust institutions, including health care providers and schools, to act in their best interest—and are increasingly turning to alternative sources for help with relationship and emotional issues. A quarter of high school seniors surveyed shared that they use AI for this purpose at least weekly; 21% of those students said they “don’t have anyone else to talk to” and 11% said AI is more supportive than their friends. Colleges and universities now face mounting pressure to expand their mental health offerings to better serve their incoming students, who seek on-demand support wherever they are, whenever they need it.

As debates about whether technologies belong in the classroom heat up, a new report from the American Psychological Association (APA) provides 10 research-based recommendations to help guide decision making around edtech. In the report, APA calls on education leaders and policymakers to ground final decisions on edtech tools in scientific evidence, and to rigorously evaluate the design, implementation, and actual learning outcomes of edtech tools. APA also cautions against conflating learning with student engagement, enjoyment, or interest, and encourages decision-makers to carefully consider the impact of tools intended for students with disabilities or learning differences.

California Gov. Gavin Newsom signed a three-bill youth online safety package into law today, giving California the most aggressive rules in the country for how platforms and AI companions treat minors. The three bills cleared the Legislature without a single “no” vote:

Why it matters: Kids' online safety has stopped being a partisan fight, and California has written a template that other statehouses may copy. For AI and edtech companies serving young users, the compliance clock starts now, and questions about age assurance, third-party audits, and product design move out of the policy shop and onto product roadmaps. [KQED]

Last week, the U.S. Department of Education (ED) released updated findings from its Civil Rights Data Collection (CRDC), a federal survey of nearly every public school in America that aims to reveal disparities in the student experience based on factors like race, sex, disability, among others, and assess schools’ compliance with civil rights laws. This CRDC release, which covers the 2023-24 school year, notably withheld data pertaining to nonbinary students and gender identity. According to ED, these data were suppressed in compliance with President Trump’s January 2025 executive order, which ordered the federal government to recognize only two sexes, male and female; ED will not collect data on nonbinary students for the 2025-26 CRCD for the same reason. [PBS; K-12 Dive]

Check out W/A Jobs, which features 3,480 career opportunities from 328 organizations across the education industry. Hiring? For $250/month or $1,000/year, organizations can post an unlimited number of jobs with no per-job fees.