Computers and Education Open (Dec 2022)
Game learning analytics can unpack Escribo play effects in preschool early reading and writing
Abstract
Randomized controlled trials usually cannot uncover why some participants benefit more or less from an intervention when they do not collect process data during the implementation. Beyond cost concerns, collecting multiple data points during implementation can be unfeasible or potentially harmful when the participants are young children. This research employed data collected by digital games to assess how different students benefited from the Escribo Play word reading and writing intervention during a trial with 749 preschool students. The learning analytics identified that the total number of incorrect answers was the data type that most differentiated students. The clustering method identified four clusters that displayed significant differences in how they benefited from the intervention. Cluster 1 had most students and displayed an effect on reading (d = 0.58) stronger than the overall intervention (d = 0.40). Cluster 2 only had 13 students and displayed an effect three times higher than the overall (d = 1.23). Clusters 3 and 4 had very low pre-test scores and did not benefit from the intervention, probably because they lacked prerequisite skills. These findings allow policymakers to target better the delivery of the intervention to improve its cost-benefit and the developers to improve the intervention for Clusters 3 and 4. The study demonstrated that learning analytics can be a cost-effective and unharmful alternative to process data to assess the effectiveness of game-based educational interventions for early childhood education.
