From Hallucination to Critical Thinking: Developing an AI-MIL Framework in Korean School Libraries

Abstract

When students accept ChatGPT's answers uncritically, AI hallucination silently erodes the foundation of learning. But what if that very hallucination could become a powerful teaching tool for critical thinking? As generative AI reshapes information seeking, information literacy has shifted from "finding information" to "designing questions and verifying AI-generated answers." Yet existing instructional models have not caught up. Six school librarians from six Korean high schools collaboratively developed the AI-MIL (Artificial Intelligence–Media and Information Literacy) Framework, redesigning the traditional Big6 model by integrating UNESCO MIL competencies and Computational Thinking. In school library sessions, students progressively refined their prompts through a Question Continuum model and directly detected hallucinations and biases in AI outputs using the SIFT strategy—turning verification itself into a vehicle for critical thinking. Where empirical evidence of generative AI literacy education in K–12 school libraries remains globally scarce, a pilot study with 33 students (32 completing both surveys) using paired t-tests confirmed significant improvements in search strategy (p=.015) and evaluation skills (p=.021). Students demonstrated habitual cross-verification and evidence-demanding behaviours. Designed to ensure equitable AI literacy regardless of socioeconomic context, this framework offers a scalable, ready-to-implement model for school libraries worldwide.