Updated October 2026: This piece now focuses on Gert Biesta’s critique of evidence-based education. The earlier version leaned on Michael Apple for much of its argument, and that material is becoming a separate piece. For Biesta’s three domains of educational purpose, see What Is School For? Leading Conversations With Biesta.
The phrase shows up in a predictable spot. A district is deciding whether to adopt a reading program, a behavior framework, or a platform, and somewhere in the second half of the meeting someone says that the option on the table is research-based. What happens next is the part that interests me. The conversation doesn’t get more rigorous so much as shorter, and the questions that were still open a moment earlier, about whether this is the kind of thing the district wants to be doing at all, get treated as settled by a claim that never addressed them.
I’m describing this as a composite rather than a single meeting because it has happened often enough that no one version is the real one. I’ve been the person saying it. I’ve written it into proposals and put it on slides, and when I ask for source grounding before I draft an article, which I do as a matter of routine, I’m reaching for the same authority. So this isn’t a piece about other people’s bad habit. It’s a piece about a phrase I keep using and have started to distrust, and about what I think it’s doing when it works.
What does the phrase actually answer?
“Research-based” is a claim about means. It says that some intervention, delivered under some conditions, produced some measured effect on some outcome. That’s useful information. It’s also information of a narrow kind, and the narrowness is easy to miss because the phrase carries the tone of a full verdict. Effectiveness is never freestanding, in that something is effective at producing a particular result, and the result had to be chosen before anyone could measure progress toward it. A reading program that raises scores on a fluency assessment is effective if fluency, as that assessment defines it, is what you were after. Whether that’s what you were after isn’t a research question but a question about what you want reading instruction to be for, and the study can’t answer it because the study assumed an answer in order to get started.
Gert Biesta made this argument in a 2007 essay whose title I haven’t been able to get out of my head: “Why ‘What Works’ Won’t Work.” His target is evidence-based practice, which education imported largely from medicine. That model treats the ends of education as given and leaves open only the question of which means reach them most efficiently. Biesta’s point is that this narrows educational decisions in two ways at once. It limits what can be decided to questions of effectiveness, and it limits who gets to take part in deciding, since the people with standing in an effectiveness question are the people who can produce or interpret the evidence. He calls the second narrowing a democratic deficit. In education the ends aren’t given but contested, and arguing over them is work a democratic society is supposed to do out loud. When “what works” becomes the governing question, that work doesn’t disappear. It gets done quietly, by whoever chose the outcome measure, and then presented as though it were a finding rather than a choice.
Biesta later organized the question of ends into three domains of purpose (qualification, socialization, and subjectification), which I walk through in What Is School For? For this piece, one observation about them is enough: nearly everything that gets called research-based is research on qualification, and usually a thin slice of it. That would be fine if we treated it that way. Instead, evidence about one domain gets used as a verdict on the whole.
Why doesn’t “what worked” travel?
Biesta came back to the argument in 2010 with a follow-up whose title I like even more: “Why ‘What Works’ Still Won’t Work: From Evidence-Based Education to Value-Based Education.” By then evidence-based policy had spread further, and he pushed the critique past the question of ends. Even when everyone agrees on the goal, he asks, can evidence deliver what it promises? He names three deficits: 1) a knowledge deficit, in that research can tell us what has worked in the past and not what will work next time, 2) an efficacy deficit, in that education isn’t a closed causal system, since teachers and students act on their own interpretations of what happens to them and the link between an intervention and its effect is at most probabilistic, and 3) an application deficit, which is the one I find most useful for the work I do.
The application deficit rests on an example Biesta borrows from Bruno Latour’s account of Pasteur. Pasteur’s techniques worked on French farms, but only because the farms were first changed to look more like the laboratory where the techniques had been developed. What worked in the lab didn’t simply travel out into the world. The world was rebuilt to receive it.
Education has its own version of this, and it goes by the name fidelity. When a district adopts a research-based program, the adoption usually comes with pacing guides, scripted lessons, implementation checklists, and walkthroughs that measure how closely teachers follow the design. Some of that is reasonable, since a program can’t be judged if nobody actually ran it. It is also, in Biesta’s terms, the classroom being reshaped to resemble the conditions of the study. The things that get trimmed to make it fit (a teacher’s judgment about a particular group of kids, a detour the class needed, a purpose the community holds that the study never measured) are the things the evidence had no way to see in the first place.
This is where I think “research-based” does its quietest work. It doesn’t only settle the question of ends before anyone has argued about it. It then supplies a reason to reorganize the school around that answer, and the reorganization looks like diligence rather than a decision.
Where do I sit in this?
The machinery that certifies “research-based” is formal now. The 2015 Every Student Succeeds Act defines “evidence-based” in four tiers, from strong evidence down to interventions that only “demonstrate a rationale,” and the federal What Works Clearinghouse rates programs that districts can cite when a purchase needs justifying. The clearinghouse is named after the exact phrase Biesta’s essays argue with, which says something about how settled the assumption has become. Platforms add their own layer on top. A learning management system produces engagement reports, a behavior tool produces dashboards, and a curriculum product ships with its own efficacy studies.
Someone has to configure and read all of that, and I’m one of the people who does. I set up the reporting structures for a statewide LMS deployment, and I’ve watched how a completion rate, once it exists, becomes the thing the next conversation is about. Nobody decided that completion was what the professional learning was for. The number was simply the one that was available, and available numbers have a way of turning into purposes. That is Biesta’s democratic deficit in a small, ordinary form. The educators taking the courses never weighed in on whether completion was the point, and if I’m honest, neither did I. The field in the report made the decision, and I built the report.
Is this an argument against research?
No, and it isn’t one Biesta makes either. His alternative, value-based education, still gives evidence a role. It just refuses to let evidence set the direction. The history of education has plenty of examples of what happens when values operate without evidence, and the reading wars are the obvious one. Whole-language instruction was a values commitment about how children should relate to text, and a generation of kids paid for the fact that its advocates were slow to take seriously the evidence that it was failing many of them. Research-based reading instruction was a correction that was needed, and I’m glad it happened.
What I’m arguing is narrower than research versus values, then. It’s about sequence. Research can tell us how well a means reached an end somewhere else. It can’t choose the end, and it can’t promise that the means will work here without changing “here” to fit. When “research-based” is used to close the conversation about ends, it’s doing something the research never licensed. The purpose question has to be asked first, by the people whose purposes are at stake, and the evidence comes in afterward as a servant of that answer rather than a substitute for it.
Where does that leave me?
The habit I’m trying to break is reaching for “research-based” when I want to win. The question I’m trying to move earlier is “toward what,” asked before “does it work,” and asked in a way that leaves room for an answer I didn’t supply. When a district asks me about a platform configuration, I try to ask what they want their teachers’ professional learning to be for before I show them what the reporting can track, and the honest answer is often that they haven’t decided.
The harder part is my own work. The completion rate I set up didn’t stop existing when I started to distrust it. Numbers like that outlast the reasons they were built, and other people come to rely on them in ways I can’t fully see. What I can do is notice when one has started making decisions I didn’t intend it to make, and say so in the room. I don’t know yet whether that’s enough, or whether the person who builds the measures can also be the one who keeps the purpose question open.
If you’re working the same layer, running the reports and sitting in the meetings where a study gets cited and the conversation ends, I’d like to hear how you keep the harder question alive. Reach me at licht.education@gmail.com, and find more of my writing at bradylicht.com.
