TY - DATA T1 - Supplementary data for the paper 'Social robots in education: A meta-analysis of learning outcomes' PY - 2024/09/16 AU - Joost de Winter AU - Dimitra Dodou AU - Fleur Moorlag AU - Joost Broekens UR - DO - 10.4121/a78b6b99-3fdf-4ae0-97b0-3b618b00805e.v1 KW - social robot KW - cognitive skills KW - sentiment analysis KW - meta-analysis N2 -

Social robots are an active research area, particularly for their use in education. Previous meta-analyses have shown that social robots have a positive impact on learning, with a Cohen’s d compared to control conditions of around 0.6. However, these analyses were often limited in scope or had problematic inclusion criteria, such as different control conditions (e.g., different learning methods or pre-post comparisons). In this meta-analysis, we examined learning outcomes with more studies and a focus on the type of control conditions. Results from 79 studies shown that social robots produced larger learning gains when compared to no intervention. Robots were also more effective than human teachers, though there was large variability in the effect sizes. This variability was partly explained by co-teaching, where robots paired with humans were more effective than robots alone. We also show, based on 370 effects, that pre-post effects are mostly greater than 0, which can be explained because learning inevitably occurs with prolonged practice over time. A sentiment analysis using a large language model revealed that papers from outside Europe used more positive language when describing the robots. The conclusion drawn from the current meta-analysis is that the effect size does not stand on its own but is influenced by the way the robot is used and the control condition chosen. For future research, we recommend fair comparisons where only one parameter (such as embodiment) is varied at a time.

ER -