For several years, much of the discussion surrounding artificial intelligence in language education has concentrated on what the technology can do. Researchers are now increasingly asking a harder question: does it actually improve language learning?
Recent research published in ReCALL, Cambridge University Press’s journal devoted to technology and language learning, reflects this change in emphasis.
Researchers are examining areas including generative AI-supported self-regulated learning and the combination of corpora, generative AI and text-to-speech technology for speaking practice. The interest is moving beyond demonstrating that learners can interact with AI towards understanding what those interactions actually contribute to language development.
Generating an exercise, correcting a paragraph or conducting an artificial conversation is relatively easy to demonstrate. Establishing whether repeated use produces measurable improvements in proficiency is considerably more difficult.
The same problem applies to feedback.
Generative AI can provide learners with comments on grammar, vocabulary, organisation and style within seconds. Speed, however, tells us little about educational value. What matters is whether learners understand that feedback, act on it and subsequently become better writers or speakers.
Schools and teachers have spent the first years of the generative AI era experimenting. The next phase will require evidence.
Which applications improve learning? Which primarily save teachers time? Which encourage learner independence? And which simply make existing activities faster without producing better educational outcomes?
AI has already demonstrated that it can generate language-learning material remarkably quickly. The harder task now belongs to researchers, teachers and technology developers: demonstrating that students actually learn more because of it.
