Replicating the advantage of classifying over generating examples in an authentic relational category learning context

Educators use psychological knowledge about categories to adapt their teaching to learners. Learning these categories usually requires understanding the underlying relational structures. To learn relational categories, learning tasks should effectively foster generalization and discrimination processes. Previous research indicates that tasks requiring learners to generate their own examples are less effective for learning relational categories and more likely to elicit overestimation of learning than tasks requiring learners to classify provided examples. We conducted an experiment with N = 104 students in degree programs that encompass educational science to investigate whether the disadvantage of generating examples in contrast to classifying examples can be mitigated by deliberately promoting generalization and discrimination processes during learning. The experiment was designed to reflect an authentic context for learning relational categories and to assess how students apply these categories in real-world settings. The results showed that despite the deliberate promotion of generalization and discrimination processes, classifying provided examples was still more effective for learning relational categories. Classifying examples presumably fosters a high level of cognitive engagement while simultaneously providing instructional guidance. In contrast, generating examples may offer insufficient guidance and thereby restrict learning. Moreover, students who generated examples during learning not only overestimated their learning but the overestimation also persisted throughout the knowledge test and after its completion. Based on the findings, educators should consider using classification tasks rather than generation tasks when teaching relational categories while also taking into account both learning and metacognitive processes.