NEPC Review: 'Measuring Artificial Intelligence in Education' (Bellwether, October 2025)
Saved in:
| Title: | NEPC Review: 'Measuring Artificial Intelligence in Education' (Bellwether, October 2025) |
|---|---|
| Language: | English |
| Authors: | Bradley Robinson, University of Colorado at Boulder, National Education Policy Center (NEPC) |
| Source: | National Education Policy Center. 2025. |
| Availability: | National Education Policy Center. School of Education 249 UCB University of Colorado, Boulder, CO 80309. Tel: 303-735-5290; e-mail: nepc@colorado.edu; Web site: http://nepc.colorado.edu |
| Peer Reviewed: | Y |
| Page Count: | 13 |
| Publication Date: | 2025 |
| Sponsoring Agency: | Great Lakes Center for Education Research and Practice |
| Intended Audience: | Policymakers |
| Document Type: | Reports - Evaluative Opinion Papers |
| Education Level: | Elementary Secondary Education |
| Descriptors: | Artificial Intelligence, Logical Thinking, Influence of Technology, Elementary Secondary Education, Measurement, Evaluation Methods, Models, Research Methodology |
| Abstract: | Because of the rapid proliferation of AI-powered technologies in education, educators and others face urgent challenges around evaluating the potential impacts of these technologies on teaching and learning. Bellwether's October 2025 report, "Measuring Artificial Intelligence in Education," aims to address such challenges by promoting logic models as a framework for moving beyond superficial metrics toward more robust evidence-based educational outcomes. Logic models involve determining four main components--inputs, activities, outputs, and outcomes--and they have a long history of use in program evaluation. The logic-model approach, however, has very real limitations that are not fully addressed in the report. Just as importantly, the report simply assumes that AI should indeed be integrated into education. That is, logic models function here as a methodological heuristic for ensuring AI fulfills its taken-for-granted potential. By positioning logic models as value-neutral, the report overlooks how such approaches ignore contextual complexity and the potential for unintended harms. Rather than offering critical guidance for assessing AI's role in education, the report provides methodological cover for predetermined conclusions about AI's inevitability and desirability. Policymakers seeking rigorous, evidence-based approaches will find little support in what is, at its core, a promotional document. |
| Abstractor: | As Provided |
| Entry Date: | 2026 |
| Accession Number: | ED681180 |
| Database: | ERIC |
| Abstract: | Because of the rapid proliferation of AI-powered technologies in education, educators and others face urgent challenges around evaluating the potential impacts of these technologies on teaching and learning. Bellwether's October 2025 report, "Measuring Artificial Intelligence in Education," aims to address such challenges by promoting logic models as a framework for moving beyond superficial metrics toward more robust evidence-based educational outcomes. Logic models involve determining four main components--inputs, activities, outputs, and outcomes--and they have a long history of use in program evaluation. The logic-model approach, however, has very real limitations that are not fully addressed in the report. Just as importantly, the report simply assumes that AI should indeed be integrated into education. That is, logic models function here as a methodological heuristic for ensuring AI fulfills its taken-for-granted potential. By positioning logic models as value-neutral, the report overlooks how such approaches ignore contextual complexity and the potential for unintended harms. Rather than offering critical guidance for assessing AI's role in education, the report provides methodological cover for predetermined conclusions about AI's inevitability and desirability. Policymakers seeking rigorous, evidence-based approaches will find little support in what is, at its core, a promotional document. |
|---|
