Large language models are reshaping educational judgment, not just automating tasks.

Schools often treat large language models, or LLMs, as simple tools to automate chores like grading, tutoring, and writing feedback. The authors argue this narrow view focuses too much on speed and test scores. It ignores a bigger issue: how artificial intelligence shapes important educational choices. Instead of just doing tasks, AI now mediates educational judgment across teaching, learning, curriculum, and quality assurance. This happens across four linked dimensions: diagnostic, relational, normative, and actionable. Even if an AI gives a clear explanation, that choice might not be educationally sound. To handle this, the authors suggest practical steps: spot hidden AI judgments, demand sound educational reasons for them, test them with human experts, and let teachers and students question, change, and contextualize them.

Why it matters

If educators treat AI only as a time-saver, they risk outsourcing critical decisions about what students learn and how they are evaluated. This paper argues that LLMs mediate educational judgment—shaping what counts as correct, fair, or valuable. Understanding this helps instructors, students, and institutions design AI use that supports, rather than undermines, educational integrity.

Questions answered