How do pedagogical approaches affect the impact of chatbots on learning performance? A meta-analysis and research synthesis

Document Type

Review

Publication Date

5-1-2026

Abstract

Given the growing attention to educational chatbots and their potential to enhance students' learning performance, prior meta-analyses have examined overall effects and explored moderators such as educational level, subject area, and intervention duration. However, the role of pedagogical approaches remains underexplored, even though the effectiveness of chatbot-based interventions depends not only on technological affordances but also on how chatbots are pedagogically integrated. This meta-analysis synthesized 96 effect sizes from 56 studies using robust variance estimation (RVE) to examine how pedagogical approaches shape the effectiveness of chatbots in education. Cross-analyses further explored interactions between pedagogical approaches and contextual variables, including subject area, educational level, chatbot type, chatbot's role, and intervention duration. Results indicated that inquiry-based learning (IBL) yielded the largest effect (g = 0.87), followed by situated learning (g = 0.81), problem-based learning (PBL) (g = 0.72), project-based learning (g = 0.59), and collaborative learning (g = 0.57), while game-based learning showed a smaller yet still significant effect (g = 0.45). Within PBL, upper secondary students showed larger learning gains than tertiary learners. Within IBL, post-generative AI chatbots and those adopting dual conversational and resource roles tended to yield more favorable outcomes. Overall, these findings highlight the importance of aligning pedagogical approaches with chatbot integration and offer evidence-based guidance for designing effective, theory-informed chatbot-assisted learning interventions.

Keywords

Chatbots, Learning theory, Meta-analysis, Pedagogical approach

Publication Title

Educational Research Review

ISSN

1747-938X

DOI

10.1016/j.edurev.2026.100783

Volume

51

Publisher

Elsevier

Link to Full Text

Share

COinS