Tool know-how dates fast. The real target is critical and ethical AI literacy, which is hard to teach: critical thinking is domain-specific and needs dispositions, the evidence base is thin, and fully ethical use may be impossible. Language learning itself is the most workable domain.
Five difficulties
- Critical thinking is domain-specific (Willingham 2007).
- It needs dispositions, not just skills (Facione 2000).
- Media literacy alone doesn’t solve misinformation (Bulger & Davidson 2018).
- Little evidence about what works.
- Ethical use is complicated by copyright and environmental costs.
What can work
“Try, then reflect” tasks: learners test a tool against their own learning goals and judge its contribution (Pegrum, Hockly & Dudeney 2022 Digital Literacies). Build this longitudinally, not as a one-off. Scaffold self-regulation: AI’s potential “is only likely to be realised if the learner is autonomous and self-regulating”.
Key reading
Edmett et al. (2024) AI and ELT (British Council); Menyhei & Szőke (2025) AI Literacy in the Language Classroom (DELTA Publishing); Watters (2021) Teaching Machines.
Apply in my context (Claude)
- AI-literacy reflection fits the metacognition routines in Make metacognition explicit in lessons: “Did the chatbot help you learn this, or just finish it?”
Connections
Does ChatGPT enhance learning - performance is not learning · AI guidelines for teachers - six principles