AI in class: use with caution
Recent coverage urges teachers to treat classroom AI as a curated aid, not a shortcut — research warns ChatGPT‑style tools can act as a 'cognitive crutch' that weakens memory if students aren't actively engaged. Schools should combine open dialogue about AI limits, careful tool selection, and lessons in digital literacy rather than handing tasks outright to bots. ( )
A randomized controlled trial (N = 120) found students who used ChatGPT while learning scored 57.5% on a surprise retention test 45 days later versus 68.5% for students who used traditional study methods (t(83) = –3.19, p =.002, Cohen’s d = 0.68). (papers.ssrn.com) A mixed-methods study of 666 participants reported a significant negative correlation between frequent AI-tool use and critical thinking, with younger users showing higher dependence on AI. (mdpi.com) Reviews in cognitive science frame those outcomes as cognitive offloading amplified by generative AI and warn that offloading can trade short-term efficiency for weaker durable memory unless active retrieval is preserved. (frontiersin.org) TeachAI’s AI Guidance for Schools Toolkit recommends adopting clear principles, updating policies, and integrating AI literacy into professional development and curriculum as part of school-level implementation plans. (teachai.org) Practical classroom protocols tested in applied research include brief, low-stakes retrieval tasks before permitting external aids (for example, 2–3 minute written recall or two targeted questions), because spaced retrieval and testing have repeatedly improved retention in primary-school settings. (cultofpedagogy.com) Human‑centered models endorsed by state and international guidance (the “Human–AI–Human” scaffold and UNESCO’s age‑appropriate, human‑agent approach) recommend using AI for drafting feedback and formative scaffolds while reserving final product creation and assessment to student-generated work. (ospi.k12.wa.us) Simple monitoring strategies mirror the RCT design by checking delayed retention (4–7 weeks) on core objectives and using those data to gate or broaden AI access; TeachAI lists “Evaluation” as one of its core principles for iterative policy adjustments. (papers.ssrn.com)
