AI-BASED ADAPTIVE LEARNING IN ONLINE FOREIGN LANGUAGE EDUCATION

Abstract

This paper explains how artificial intelligence can make online language courses adapt to each learner, and what this means for teaching German. Only studies published between 2022 and 2026 were reviewed. Recent research shows that adaptive systems help learners: across 45 studies the average benefit was g = 0.70, and across 41 studies of AI chatbots it was 0.576. New scheduling programs that decide when a learner should review a word have also become much more accurate, cutting prediction errors by up to 64% on very large datasets. However, most of this research does not say what an adaptive program should actually track in a specific language. This paper therefore does two things. First, it shows that adaptivity is not one decision but six: choosing the next task, ordering tasks, giving feedback, planning review, placing learners at a level, and defining what can be practised. Second, it turns eight common German problems — article gender, plurals, cases after two-way prepositions, separable verbs, adjective endings, the auxiliary in the Perfekt, word order in subordinate clauses, and long compound nouns — into clear rules that a computer program can follow, each with a signal that starts extra practice and a condition that ends it. A twelve-week study with 120 learners at A2–B1 level is proposed to test these rules. The main conclusion is simple: adaptive technology only helps if it knows what is difficult in the language being taught.