Toward Agent-Based Educational Science: Rethinking Educational Research in the Age of AI

Abstract

Educational science faces a structural mismatch between the pace of educational innovation and the methods used to evaluate its developmental impact. While new pedagogical approaches and AI-driven learning technologies are rapidly deployed, established classroom-based empirical research remains slow and fragmented, often generating evidence only after large-scale implementation has occurred. Here, we propose agent-based educational science (AES): a research agenda in which educational theories are formalized as interacting agents and environments, enabling in silico experimentation on developmental processes that are otherwise slow or infeasible to test empirically. By research agenda we mean not merely a list of open questions, but a coordinated invitation to the research community to reorient its priorities, workflows, data infrastructures, and validation practices around theory-constrained, dynamically updateable, and empirically examinable simulations of learner development. Recent advances in generative artificial intelligence bring this agenda within practical reach. As a concrete instantiation, we introduce Student Development Agents, computational agents designed to generate developmental features under counterfactual educational environments. Rather than replacing empirical research, agent-based educational science reconfigures its role, enabling predictive, ethical, and cumulative theory building in the science of learning.