AI Education

LLM agents as tutors, learners, user simulators, and standardized patients for scalable, feedback-rich education.

Education agents and learning simulation
AI Education Socratic learning User simulation Self-regulated learning AI standardized patients
Socratic dialogue and education agent pipeline

AI Education studies how LLM agents can become scalable tutors, learners, simulators, and standardized patients. The focus is educational interaction: models should ask useful questions, provide learner-aware feedback, support repeated practice, and make learning outcomes easier to evaluate.

Background and Motivation

Education needs interaction, not just answers

A strong educational model should probe understanding, adapt to learners, guide reflection, and create practice loops rather than only deliver final responses.

Scalable practice is still scarce

Teachers, tutors, and standardized patients are difficult to scale across repeated sessions, diverse cases, and individualized feedback needs.

Evaluation should follow learning behavior

Educational AI should be judged by whether it improves question asking, self-regulation, communication, reasoning, and confidence under realistic interaction.

Research Storyline

Simulate
Use LLMs as interactive users and learners

User simulators create controllable multi-turn partners that can teach dialogue models, stress-test systems, and generate richer learning signals.

Question
Teach through Socratic dialogue

PlatoLM and Socratic-style data construction turn questioning, clarification, and guided reasoning into a training signal for multi-round dialogue.

Coach
Support self-regulated learning

SRLAgent explores how LLM assistance and gamified feedback can help learners plan, monitor, and reflect on their own study process.

Practice
Build AI standardized patients

EasyMED and SPBench make clinical communication training repeatable, controllable, and measurable for medical learners.

Representative Work

SRL
SRLAgent: Enhancing Self-Regulated Learning Skills through Gamification and LLM Assistance

Builds an LLM-assisted learning system around planning, monitoring, reflection, feedback, and engagement.

Paper
User
Large Language Model as a User Simulator

Frames LLMs as simulated users for interactive dialogue training and evaluation.

Paper
PlatoLM
PlatoLM: Teaching LLMs in Multi-Round Dialogue via a User Simulator

Uses Socratic-style questioning to improve multi-turn dialogue learning.

Paper
AI-SP
Human or LLM as Standardized Patients?

Introduces EasyMED and SPBench for comparing AI standardized patients with human standardized patients in medical education.

Paper
Co-design
It Talks Like a Patient, But Feels Different

Studies how medical learners experience AI standardized patients and what design requirements emerge from real use.

Paper

Project Clusters

Learning companion agents

LLM systems that scaffold planning, reflection, formative feedback, and learner engagement.

Socratic dialogue training

Question-driven user simulation and data construction for teaching models to handle multi-round educational dialogue.

Medical education simulation

AI standardized patients for history taking, communication, empathy, diagnostic reasoning, and structured feedback.

Evaluation-centered education AI

Benchmarks and rubrics that connect interaction quality to measurable learning behavior and outcomes.

Display Figures

Resource Map

PlatoLM

Open resources for Socratic-style dialogue training and user-simulator-driven model improvement.

Repository
SocraticChat

Dataset resource behind Socratic dialogue construction and educational multi-turn training.

Dataset
EasyMED

AI standardized patient framework and medical education evaluation resources.

Repository