Landzettel, Timo (2025) A Data-Driven Approach to Feedback Optimization in Language Learning Apps. Master thesis, Psychology.
|
Text
Final-Thesis-Timo-Landzettel.pdf Download (1MB) | Preview |
Abstract
Effective feedback helps learners understand what went wrong and how to improve. In language-learning apps, feedback is often created by comparing a learner’s answer with a target answer, but this works best when we know the typical errors learners make. This enables creating automated feedback messages tailored to explain specific mistakes that were made. However, typical error categories are unknown. Here we study English vocabulary training by German Gymnasium students mostly aged 10–13, analyzing more than 3.2 million responses to identify recurring mistake patterns that could support explanatory feedback. We find twelve categories, including missing elements (for example, “to ” in infinitives), spacing, capitalization, punctuation, double-letter errors, and phonological similarity, among others. After this coverage, 22% of wrong responses and 15% of near-correct responses remain unexplained. These results provide a practical foundation for automated feedback that names the error and suggests a fix, which can reduce uncertainty for learners. More broadly, they inform the design of language-learning applications for early secondary students and support more effective vocabulary acquisition in second-language learning.
| Item Type: | Thesis (Master) |
|---|---|
| Supervisor name: | Rijn, D.H. van and Sprenger, S.A. |
| Degree programme: | Psychology |
| Differentiation route: | Other [Master Psychology] |
| Date Deposited: | 17 Sep 2025 10:02 |
| Last Modified: | 17 Sep 2025 10:02 |
| URI: | http://gmwpublic.studenttheses.ub.rug.nl/id/eprint/5928 |
Actions (login required)
 |
View Item |