Evidence-Based Education - From Causal Ascriptions to Relative Effectiveness Generalizations and Effectiveness Predictions - A Research Brief

Sep 15, 2026

Uploaded by Jasmine

Evidence-based education requires rigorous standards for identifying improvement in learning, moving beyond knowing whether an intervention works to determining whether it works better than current practice under controlled conditions and with minimized bias. The brief argues for “relative evidence” synthesized across studies and meta-analytic comparisons, and for effectiveness predictions that incorporate local contextual variables. It reviews the nature of educational evidence and presents the EBE3 framework connecting theory-building with validation to guide researchers, policymakers, and practitioners.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages16

Reading Time40 min

AccessFree to download

Introduction: “What Works” Is Not Enough

Problem 1: Higher Standards for Evidence of Improved Learning

Problem 2: The Need for Relative Evidence

Problem 3: Context-Responsive Effectiveness Predictions

Theory-Building and Theory-Testing in Educational Research: Divide, Compromise, or Synergy?

EVIDENCE-BASED EDUCATION: FROM CAUSAL ASCRIPTIONS TO RELATIVE EFFECTIVENESS GENERALIZATIONS AND EFFECTIVENESS PREDICTIONS A Research Brief by the International Science and Evidence-based Education Assessment An Initiative of UNESCO MGIEP* To more reliably achieve educational goals based on values and policies (Brighouse et al., 2018), quantitative and qualitative traditions should complement each other to strengthen the quality and impact of empirical research under the broad banner of evidence-based education (EBE) . And since human learning and development are the cornerstones of educational goals, the domains contributing to educational research rest on a vast number of areas, including in the following fields: learning sciences and cognitive science (psychology and neuroscience [behaviour and brain processes]); computer science (computer-based learning systems, learning analytics); economics; and social sciences (the learner and their broader context) . INTRODUCTION: “WHAT WORKS” IS NOT ENOUGH This research brief identifies three overarching problems in educational practice which a different approach to EBE can help solve. Problem 1: There is a need to apply a high minimum standard for what counts as evidence of improved learning. From the perspective of EBE, decisions about which practices to use in a given learning context should ideally be based on evidence (Slavin, 2020) . Evidence begins with a demonstration of the effect of a treatment on a defined outcome (Connolly et al., 2018) and, more broadly, empirical support that a policy works generally or in a specific context (Joyce & Cartwright, 2020) . However, it is not enough to simply know whether a given intervention works or not. As important as that is, we should expect a higher standard: that is, to know whether an intervention will work better than what we are already doing, compared to a control group, and after eliminating as many sources of bias as possible. 1 Director, NeuroLab, University of Quebec in Montreal, Montreal, Canada 2 Ph. D. student, University of Quebec in Montreal, Montreal, Canada * The first author is Co-Chair of the ISEE Assessment Data and Evidence Group. The second author is a lead author on a full chapter on the present topic, which is to be part of the ISEE Assessment. The authors thank Stephen Fraser, Anantha Duraiappah, Yoko Mochizuki and Clara Cardona for their reviews, comments and inputs on earlier drafts. This research brief is an interim output of the ISEE Assessment and has not been extensively peer-reviewed. Please read the ISEE Assessment Report for peer-reviewed content. The analysis, conclusions, and recommendations contained herein are solely a product of the individual authors involved in the ISEE Assessment and are not the policy or opinions of, nor do they represent an endorsement by, UNESCO or UNESCO MGIEP. Problem 2: Before the need for additional evidence in the form of new interventions, we argue that there is a need for what we call “relative evidence.” Relative evidence arises from the combined results of multiple studies, meta-analysis, and thorough comparisons of multiple extant interventions. Such comparisons enable an assessment of all pertinent interventions with respect to their effectiveness against specific outcomes, allowing practitioners to answer the question: Given all the possible interventions available to me, which is most likely to yield success in my specific context? The consistency or variability of effect sizes across studies of similar interventions is critical to support assertions regarding their general effectiveness. However, there is a lack of relative evidence in extant literature regarding a majority of educational issues: new interventions are being tested against a control group (business as usual) and welldocumented interventions are rarely compared and aggregated through proper metaanalytic approaches. Problem 3: There is a need for strong assertions about ho

FAQs

What higher standard does evidence-based education require beyond “what works”?

It requires knowing whether an intervention will work better than what learners are already experiencing, compared with a control group, while eliminating as many sources of bias as possible.

What is “relative evidence” and why is it needed?

Relative evidence comes from combining multiple studies and meta-analytic comparisons of existing interventions to judge effectiveness across outcomes and support decisions among available options in a specific context.

How does effectiveness prediction address local context in education?

Effectiveness prediction forecasts how an intervention—abstracted from causal claims and general effectiveness statements—will work within the specific constellation of variables in a given setting, rather than relying on biased or non-scientific reasoning.

Sep 15, 2026

Uploaded by Jasmine