Why AI in schools often looks like change without changing anything

Generative artificial intelligence may be forcing education to confront an uncomfortable pattern that researchers now call simulated innovation: schools and universities adopt new tools, dashboards, and platforms that visibly signal modernisation, while the underlying purposes of teaching, the logic of assessment, teacher judgement, and student agency remain largely untouched. In a new Perspective article published in Frontiers in Education, researchers from Universidad de La Sabana in Colombia and Universidad de Sevilla in Spain argue that generative AI creates a critical inflection point in the long-standing relationship between educational innovation and technological development, one that could finally allow educators to set the agenda for technology rather than perpetually chase it.

The authors, Diana Carolina Fajardo-Ramos, Andrés Chiappe, and Sandra Martínez-Pérez, describe a familiar sequence that has dominated educational technology for decades. A new technology emerges, institutions search for its possible pedagogical uses, teacher education programmes adapt belatedly, and innovation is framed as the capacity to incorporate what has already been designed elsewhere. This technology-led sequence, they contend, has produced a recurrent pattern in which educational change is measured by the presence of new tools rather than by the transformation of pedagogical purposes, learning relationships, assessment practices, or professional judgement. The result is the performance of transformation without its substance.

The article defines simulated innovation as a diagnostic construct with four defining characteristics. The first is symbolic modernisation, in which innovation is made visible through the adoption of new tools, platforms, automated feedback systems, or institutional narratives of digital transformation, even when the educational problem being addressed remains weakly specified. The second is pedagogical continuity: technologies are incorporated into educational settings, but the deeper organisation of teaching and learning remains largely unchanged. New systems may accelerate feedback, automate planning, or personalise content, while inherited assessment formats, teacher roles, and expectations of student performance stay intact.

The third characteristic is evidentiary displacement. In simulated innovation, the evidence used to justify change shifts from pedagogical transformation to indicators of implementation, efficiency, visibility, or technical performance. Success is demonstrated through adoption rates, frequency of use, speed of production, or dashboard indicators, rather than through evidence that learning, agency, equity, or professional judgement have been strengthened. The fourth characteristic is externalised directionality: the direction of educational change is defined primarily by technological availability, market narratives, policy urgency, or competitive pressure, rather than by educationally formulated problems. Schools begin with the question of how to incorporate a technology that has already arrived, instead of asking what kind of learning, equity, or democratic participation should guide its use.

The authors are careful to distinguish their concept from related ideas. Simulated innovation differs from technological determinism, which assumes technology drives educational change by its own internal logic; it examines instead how institutions, policies, and markets organise themselves around technological possibilities while presenting that alignment as innovation. It also differs from digital transformation, which refers broadly to how digital technologies reorganise institutional practices, and from simple technology adoption, which describes uptake rather than educational consequence. A school may adopt AI extensively and remain pedagogically unchanged, while a modest use of AI may avoid simulated innovation if it helps teachers and students make reasoning visible, improve feedback cycles, or expand access to meaningful support.

What makes generative AI different from earlier educational technologies, the researchers argue, is not merely its speed or sophistication. Large language models can produce fluent explanations, arguments, feedback, summaries, and written academic performances that resemble competent educational performance. Written answers, essays, lesson plans, and problem explanations have historically been treated as evidence that a learner or teacher has engaged in intellectual work. Because GenAI can generate artefacts that educational institutions have long used as proxies for thinking and understanding, it destabilises the evidential relationship between product, process, authorship, and learning. It compels education to decide what kinds of learning, evidence, authorship, judgement, and human agency should be preserved, strengthened, or redesigned in AI-mediated environments.

The authors stress that this does not mean GenAI thinks, understands, or learns the way students or teachers do. The point is more precise: the technology can produce artefacts that schools have historically accepted as proxies for intellectual work, and it can do so without transparent reasoning, sometimes citing sources inaccurately or producing texts whose surface quality exceeds the learner’s demonstrated understanding. In this context, simulated innovation becomes harder to detect: a classroom may appear more personalised, interactive, or efficient while learners’ responsibility for judgement and the evidential value of student work become less visible or less secure.

Against this backdrop, the article proposes pedagogical primacy as the counter-mechanism. Rather than searching for pedagogical uses for each new technological release, educational communities would first define the learning problems, evidentiary standards, ethical limits, and forms of agency they seek to strengthen, and only then determine whether and how generative AI should be designed, adopted, constrained, or refused. In a technology-led sequence, innovation begins with the artefact. In a pedagogy-led sequence, innovation begins with an educational horizon: what kind of learning should be strengthened, what inequities should be reduced, what forms of judgement should be protected, and what kinds of evidence should count as learning.

Teacher education emerges as a strategic site for this shift, though not because teachers alone can govern AI-mediated education. Students, school leaders, policymakers, curriculum designers, families, and platform developers all shape how AI is interpreted and used. But teacher education is where future and practising teachers can learn to translate educational purposes into design decisions, assessment criteria, ethical boundaries, and professional judgement. The authors draw on established frameworks, including the AI Competency Framework for Teachers, the European DigCompEdu model, and AI-TPACK, which incorporates AI-related knowledge into the relations among technological, pedagogical, and content knowledge. They argue these frameworks should be read not only as models of integration but as instruments of direction-setting: teachers need to know not only how AI can support content and pedagogy, but how disciplinary purposes and pedagogical principles can define what AI should and should not do.

In practical terms, the researchers outline four connected capacities that teacher education should cultivate. The first is diagnostic: identifying whether an AI use responds to a clearly defined educational problem or merely adds technological novelty. The second is design-oriented: constructing tasks and assessment conditions in which AI supports inquiry, feedback, and revision without replacing learner responsibility. The third is ethical and epistemic: addressing bias, transparency, authorship, validity, privacy, and accountability as pedagogical concerns rather than external compliance requirements. The fourth is ecological: working with students, colleagues, institutions, and communities to define shared norms for AI-mediated learning. The authors also caution that pedagogical primacy can fail, when institutional policy adopts AI mainly to signal modernisation, when teachers are trained only as tool users, or when assessment systems continue to reward final products that AI can generate on demand. The article’s claim is deliberately conditional: generative AI creates an opportunity for education to lead innovation, but that opportunity is realised only when educational communities can define the problems technology is expected to address. If AI is treated merely as another tool to be incorporated, education will remain at the end of the innovation chain; if teacher education and broader educational ecologies use it as an occasion to define their own horizons of transformation, technology can stop leading the agenda and begin responding to pedagogically defined innovation.

Subject of Research: Simulated innovation and pedagogical direction-setting in teacher education in the era of generative AI

Article Title: Can education set the agenda for AI? Simulated innovation and pedagogical direction-setting in teacher education

Article References: Fajardo-Ramos, D. C., Chiappe, A., & Martínez-Pérez, S. (2026). Can education set the agenda for AI? Simulated innovation and pedagogical direction-setting in teacher education. Frontiers in Education, 11, Article 1897627. https://doi.org/10.3389/feduc.2026.1897627

Image Credits: AI Generated

DOI: 10.3389/feduc.2026.1897627

Keywords: generative AI, teacher education, simulated innovation, pedagogical primacy, assessment redesign, AI literacy, educational technology, digital transformation, teacher judgement, student agency, educational innovation, education