Adverse effects of personalized automated feedback

Abstract

[EN] In large classes with hundreds of students, it is rarely feasible to provide students with individual feedback on their performance. Automatically generated personalized feedback on students’ performance might help to overcome this issue, but available empirical effect studies are inconclusive due to lack of methodological rigor. This study uses a repetitive randomized control experiment to explore whether automatically generated feedback is effective and for which students. Our results indicate that feedback does not have a positive effect on performance for all students. Some groups benefit from receiving personalized feedback, while others do not perform better than the control group. Students that perform average benefit most from receiving personalized feedback. However, lower-scoring students who received feedback tend to have lower attrition rates and if they participate at the final exam, their performance is not higher than the control group. Therefore, providing automated feedback is not something that should be undertaken mindlessly.