Simulating Students with Large Language Models: A Review of Architecture, Mechanisms, and Role Modelling in Education with Generative AI

Marquez-Carpintero, Lopez-Sellers & Cazorla (2025) present a thematic review of empirical and methodological studies using LLMs to simulate student behavior in education. They synthesize evidence on how LLM-based agents emulate learner archetypes, respond to instructional inputs, and interact in multi-agent classroom scenarios, and examine implications for curriculum development, instructional evaluation, and teacher training — while flagging persistent concerns around algorithmic bias, evaluation reliability, and alignment with educational objectives.

The review frames simulated students as a valuable methodological tool for evaluating pedagogy and modeling diverse learner profiles — tasks that are hard to undertake systematically with real learners. LLM integration is highlighted as a particularly versatile and scalable paradigm because it affords linguistic realism and behavioral adaptability.

Scope and synthesis

Emulating learner archetypes: LLM agents can approximate a range of learning styles, cognitive development pathways, and social behaviors, capturing diversity that is difficult to assemble in a real cohort.

Responding to instruction: simulated students engage in contextually appropriate pedagogical dialogues, enabling the testing of instructional inputs.

Multi-agent classrooms: agents interact within simulated classroom scenarios, extending student simulation to social dynamics.

Applications

The review examines implications for curriculum development, instructional evaluation, and teacher training — using simulated learners to practice and assess instruction without real students.

Concerns and gaps

Algorithmic bias: simulated student populations may encode or amplify bias.

Evaluation reliability: how well a simulated student models a real learner is itself hard to validate.

Alignment with educational objectives: simulations must serve pedagogical goals, not just reproduce plausible dialogue.

The review identifies technological and methodological gaps and proposes directions for integrating generative AI into Adaptive Learning systems and Instructional Design.

Connected Concepts

Connected Articles

Valid Student Simulation LLM 2026 — Towards Valid Student Simulation

Simulating Students Diverse Cognitive Levels 2025 — Embracing Imperfection: Simulating Diverse Cognitive Levels

Agentschool Multi Agent Simulation Education 2026 — AgentSchool: Multi-Agent Simulation for Education

History Aware Student Simulation — History-Aware Profiles for Student Simulation

LLM Student Simulation Teacher Insights — Can LLMs Simulate Human Learners?

LLM Student Simulation Misconception Faithfulness — Simulating Students or Sycophantic Problem Solving?

Citation

Marquez-Carpintero, L., Lopez-Sellers, A., & Cazorla, M. (2025). Simulating students with large language models: A review of architecture, mechanisms, and role modelling in education with generative AI. Computer Science Review, 62, 101008. arXiv:2511.06078.