Eduthink4Ai: Bridging Educational Critical Thinking And Multi-Agent Llm Systems

Xinmeng Hou, Nanyang Technological University

Ziting Chang, Nanyang Technological University

Zhouquan Lu, Shanghai Jiao Tong University

Chen Wenli

Liang Wan, Tianjin University

Wei Feng, Tianjin University

Hai Hu, Hong Kong University of Science and Technology

Qing Guo, Nankai University

Date

01/2026

AI Overview

Why it matters

Large language models do not regulate their reasoning, leaving them susceptible to adversarial prompts while indiscriminate critique over-corrects sound answers.

Study at a glance

Question: Can an adaptive multi-agent metacognitive regulation framework improve language model reasoning and support student critical thinking?
Setting and data: 45 college students from one institution testing writing tasks, plus 8 standard AI benchmark datasets.
Method: Model benchmark evaluation and within-subject user study with preference surveys.
AI tools: MetaCrit, GPT-3.5-Turbo, GPT-4o, Claude-3.5-Sonnet, Claude-Opus-4.6, DeepSeek-v3

Key findings

MetaCrit raised GPT-3.5-Turbo accuracy to 94.12% on TruthfulQA and 84% on CIAR, while achieving zero toxicity on BOLD and HONEST benchmarks.
Indiscriminate critique degraded baseline performance, showing that intervention gains depend on selectively applying scrutiny rather than critiquing every response.
In the writing study, college students significantly preferred MetaCrit over multi-agent and single-agent baselines for critical thinking and instructiveness support.

Summary drafted with AI from the full paper and may contain errors. See the full paper for details.

Large language models (LLMs) have demonstrated significant potential as educational tutoring agents, capable of tailoring hints, orchestrating lessons, and grading with near-human finesse across various academic domains. However, current LLM-based educational systems exhibit critical limitations in promoting genuine critical thinking, failing on over one-third of multi-hop questions with counterfactual premises, and remaining vulnerable to adversarial prompts that trigger biased or factually incorrect responses. To address these gaps, we propose \textbf{EDU-Prompting}, a novel multi-agent framework that bridges established educational critical thinking theories with LLM agent design to generate critical, bias-aware explanations while fostering diverse perspectives. Our systematic evaluation across theoretical benchmarks and practical college-level critical writing scenarios demonstrates that EDU-Prompting significantly enhances both content truthfulness and logical soundness in AI-generated educational responses. The framework's modular design enables seamless integration into existing prompting frameworks and educational applications, allowing practitioners to directly incorporate critical thinking catalysts that promote analytical reasoning and introduce multiple perspectives without requiring extensive system modifications.