Improving knowledge gain and emotional experience in online learning with knowledge and emotional scaffolding-based conversational agent

Conversational agents (CAs) primarily adopt knowledge scaffolding (KS) or emotional scaffolding (ES) to intervene in learners' knowledge gain and emotional experience in online learning. However, the ill-defined design for KS and ES, as well as insufficient understanding of their interactive effects on learning outcomes, have hindered the advancement of CAs in theory and practice. This study proposed systematic KS and ES design principles based on Zone of Proximal Development and growth mindset theories. We investigated their individual and combined impacts on knowledge gain and emotional experience. A quasi-experiment was conducted with 128 undergraduate students divided into four groups, corresponding to four distinct CAs: a non-scaffolding control group (CG), ES, KS, and Knowledge and Emotional Scaffolding (K＆ES) CA. The results showed that K＆ES-based CA had a significant impact on knowledge gain and emotional experience, with both being slightly improved compared to CG. Besides, KS-based CA had a positive effect on knowledge gain and emotional experience, while ES-based CA only slightly improved emotional experience compared to CG. The results validated the effectiveness of the proposed ES and KS design principles. The fine-grained analysis revealed a significant correlation between the achievement positive emotion and knowledge transfer, highlighting the importance of integrating KS and ES. In conclusion, this study offers valuable theoretical, methodological, and empirical insights for utilizing CAs to optimize online learning experiences.