Adaptive AI Technologies for Vocabulary Development in Foreign Language Learning

This article looks at how artificial intelligence helps learners build vocabulary, and why programs designed for English do not work well for German. Only research from 2022 to 2026 was used. The evidence is positive: AI-supported vocabulary teaching gave a large average benefit (g = 1.309), and AI chatbots gave a moderate benefit (0.576), with vocabulary among the areas that gained most. Modern review-scheduling programs, trained on 220 million review records, can now predict very accurately when a learner will forget a word. The problem is how these programs store a word. They usually treat a word as one item with one score. In German this is not enough: to use a noun correctly a learner must know the word, its gender (der Tisch), its plural (Tische), and for verbs the particle or the case it governs. Each part can be learned separately, so the program should keep a separate record for each part. The paper also shows that difficulty prediction for German needs extra information, such as gender class, plural type, compound structure and false friends, and that for a long transparent compound like Krankenversicherungskarte the best response to an error is not more repetition but teaching the learner to split the word. An eight-week study with 100 A2 learners and 600 coded words is proposed. The conclusion is that results obtained for English cannot simply be assumed to hold for German.