@article{klose2026_112131,
  title = {Embracing the challenge : Predicting self-testing in non-formal online courses using machine learning},
  author = {Klose, Maria and Handschuh, Philipp and Steger, Diana and Artelt, Cordula},
  year = {2026},
  journal = {Computers \& education : an international journal},
  volume = {242},
  number = {105507},
  pages = {1-13},
  publisher = {Elsevier Science},
  address = {Amsterdam [u.a.]},
  issn = {1873-782X},
  doi = {10.1016/j.compedu.2025.105507},
  url = {https://fis.uni-bamberg.de/handle/uniba/112131},
  language = {eng},
  abstract = {Technology-driven advancements have made adaptive and interactive learning techniques more accessible. Online courses increasingly integrate self-tests that offer automated and immediate feedback. Self-tests can help learners to identify knowledge gaps and to reinforce their retention and comprehension. However, not all learners readily use self-tests, raising the question of which factors may impact learners' engagement with self-tests. The present study focused on non-formal education, covering 45 online courses offered by Bavarian universities. Analyses were based on a sample of N = 1261 participants aged 16–84 years. We used a machine learning approach to predict learners' engagement with self-tests and to identify important influencing factors. Therefore, we included 50 predictor variables in an elastic net regression to explore the role of learner- and course-related characteristics. The predictor variables were drawn from self-report, process, and meta data. Overall, learners differed substantially in their self-testing behavior. The prediction model explained 11 \% of the variance in learners' engagement with self-tests. Despite the model's modest explanatory power, the analysis identified potentially relevant predictors. The two most important predictor variables were learner commitment and the intention to obtain a confirmation of participation. Accordingly, course designers might implement extrinsic incentives—such as confirmations of participation—as a potentially useful strategy to encourage learners' engagement with self-tests. From a methodological perspective, the study highlights the importance of using appropriate statistical methods—such as machine learning algorithms—to understand complex learning behaviors.},
  keywords = {Online learning, Technology-enhanced learning, Lifelong learning, Self-testing, Machine learning}
}
