The effects of robot-assisted language learning: A meta-analysis

Given the importance of learner–interlocutor interactions, social robots have received considerable attention in the field of language learning, as this technology can physically accompany learners, is equipped with common social behaviors, and can effectively communicate with learners for language acquisition. In line with the continuous efforts to examine the effects of using social robots in language learning, this study is a meta-analysis aiming to integrate the findings of previous experimental studies on robot-assisted language learning (RALL), focusing on language learning achievement. To this end, we selected 12 independent samples (N = 522) from the initial pool of 16 empirical studies—including 23 independent samples (N = 1147)—and computed effect sizes along with their standard errors. The results indicated that RALL had a medium-sized average effect (d = 0.59, SE = 0.09, 95% CI: 0.41–0.76) compared to non-RALL conditions, such as traditional learning conditions without technology (e.g., teacher-only) or other technology-mediated conditions (e.g., virtual agents and tablet PCs). We further identified several moderator variables (i.e., age group, target language, language domain, robots’ role, interaction type, type of non-RALL condition), and provided an in-depth discussion on how these variables were related to the effectiveness of using robots for language learning.