Agent-based educational science (AES)

Research Method · All research methods
Evidence · 4 claims (4 for) · 3 studies (1 associational, 1 design, 1 theoretical), q2 · 0 of 3 report an effect size · 4 claims rest on one study

Description

Agent-based educational science (AES) is a research agenda in which educational theories are formalized as interacting agents and environments, enabling in silico experimentation on developmental processes. The authors frame it as "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." It targets developmental processes that are otherwise slow or infeasible to test empirically.

Accounts

• Agent-based educational science: formalizing educational theories as interacting agents and environments for in silico experimentation: Agent-based educational science (AES) is a research agenda in which educational theories are formalized as interacting agents and environments, enabling in silico experimentation on developmental processes. The authors frame it as "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." It targets developmental processes that are otherwise slow or infeasible to test empirically. (Zhang et al. (2026))
• LLM-based web agent that engages with an online learning experience like a student and generates a description of the student experience: The article creates a web agent built on LLMs that "engages with an online learning experience much like a student would – autonomously navigating the interface and analyzing the website content". It then "generates a comprehensive description of the student experience", which the authors use both to predict student outcomes and to give designers feedback on lesson designs. The agent is positioned as an efficient, low-cost stand-in for early-stage design feedback. (Wang et al. (2025))
• AI reliance as a complex adaptive system: a micro-founded agent-based model of population-level reliance dynamics: The article models AI reliance as a population process rather than a one-shot individual judgment. Agents repeatedly choose among solving alone, accepting AI output unverified, or verifying, updating Dirichlet beliefs about AI quality and, when networked, learning from peers through experience-based or opinion-dynamics channels. As the authors put it, "individual learning rules, peer observation, and feedback exposure jointly determine whether a population remains calibrated or enters a self-reinforcing state of unverified reliance." The model is used as a theoretical tool for isolating mechanisms and deriving qualitative comparative statics, not as a point forecast. (Ahana Biswas (2026))
• Student Development Agents: computational agents generating developmental features under counterfactual educational environments: Student Development Agents are the article's concrete instantiation of AES: "computational agents designed to generate developmental features under counterfactual educational environments." They serve as simulated learners whose developmental trajectories can be examined in silico, allowing educational innovations to be evaluated before large-scale implementation. The article presents them as enabled by recent advances in generative artificial intelligence. (Zhang et al. (2026))
• Student development agent framework: learning environment, endowment dimensions, developmental dimensions, actions, behaviors, and history: The article defines a design framework whose input includes "the learning environment, the endowment dimensions of students, the developmental dimensions of students, the actions defining how student agents can interact with the given learning environment," and whose output is simulated learning behaviors and changes in developmental dimensions. Developmental dimensions serve as "both outcomes for one period and inputs for the next period," enabling iterative simulation, while history records past trajectories. This framework targets prospective evaluation of AIED innovations without exposing real students. (Jianxiao Jiang (2025))
• Use LLM-based student development simulation as a risk-free pre-deployment evaluation of AIED innovations to avoid exposing real students to uncertain effects: The article argues that because educational interventions may carry irreversible effects, novel AIED designs should be prospectively evaluated before exposure to students. Its framework enables "a risk-free simulation process" that "directly avoids exposing actual students to uncertain effects of new teaching approaches, assessment methods, or adaptive feedback mechanisms," so that initial prototyping and experimentation occur without risking learners' academic or psychological well-being. (Jianxiao Jiang (2025))

Claims

• Agent-generated descriptions significantly improve prediction of student dropout in a lesson, outperforming all other measured approaches [+M]
• In the model, social learning creates consensus in trust without shifting aggregate reliance; connectivity alone does not produce collective overreliance [+M]
• In the model, visible unverified peer use (social proof) suppresses verification and tips the population into collective overreliance, via a smooth crossover with no hysteresis [+M]
• In the MAIC case study, student development agent predictions (Concept and Scales methods) outperformed the mean baseline across all five non-cognitive dimensions on RMSE and MAE [+M]

Related Research Methods

Key Sources

• Zhang, Y., Jiang, J. & Tang, X. (2026). Toward Agent-Based Educational Science: Rethinking Educational Research in the Age of AI. Educational Psychology Review, 38, 113. https://doi.org/10.1007/s10648-026-10209-5
• Wang, S., Mitchell, J., Piech, C. (2025). AI Web Agents Can Effectively Guide Lesson Design and Predict Student Outcomes. In: Cristea, A.I., Walker, E., Lu, Y., Santos, O.C., Isotani, S. (eds) Artificial Intelligence in Education. AIED 2025. Lecture Notes in Computer Science, vol 15878. Springer, Cham. https://doi.org/10.1007/978-3-031-98417-4_21
• Ahana Biswas. (2026). Modeling AI Overreliance as a Complex Adaptive System. arXiv. https://arxiv.org/abs/2608.19616
• Jianxiao Jiang, Yu Zhang. (2025). Student Development Agent: Risk-free Simulation for Evaluating AIED Innovations. Preprint. https://arxiv.org/abs/2510.09183

Page metadata

Field	Value
type	research-method
id	agent-based-educational-science-aes
aliases	student-development-agents
title	Agent-based educational science (AES)
description	Agent-based educational science (AES) is a research agenda in which educational theories are formalized as interacting agents and environments, enabling in silico experimentation on developmental processes.
status	review
generated	by process:settle-candidates on 2026-10-10
trust	unverified — no one has confirmed this page's content

Frontmatter is the page's machine-readable record — see the Schema & Guide.