什么样的聊天机器人更能促进学习?——基于60项实验与准实验研究的元分析

What’s the Better Chatbot in Learning?--A Meta-analysis of 60 Experiments and Quasi-experiments Studies

随着人工智能技术的迅猛发展,特别是大语言模型的兴起,应用聊天机器人辅助学生学习成为研究热点。相关研究表明,聊天机器人能有效激发学生的学习动机,提高学习投入度、自我效能感和学业成就等。然而,究竟什么样的聊天机器人更能促进学习,如何才能有效地应用聊天机器人,仍是亟待解决的问题。本研究采用元分析方法,系统分析近十年国内外应用聊天机器人促进学生学习的60项实证研究。结果表明,聊天机器人对学生学习效果有中等程度的正向影响,反馈类型、交互模态等交互设计对学习效果有显著的调节作用。不同年级、不同学习方式对聊天机器人的应用效果存在显著差异,但不同学科和学习环境无显著差异。文章最后展望了聊天机器人的未来研究方向。

With the rapid development of artificial intelligence technology with large language models,the application of chatbots to assist student learning has become a research hotspot.Relevant studies have shown that chatbots can effectively stimulate students'learning motivation and improve their engagement,self-efficacy,academic achievement,etc.However,what kind of chatbots can promote learning more effectively and how to apply chatbots more effectively are still urgent problems to be solved.To explore these issues,this study uses the meta-analysis method to systematically analyze the 60 empirical studies on applying chatbots to promote student learning conducted internationally in the past decade.The analysis results show that chatbots have a moderate positive impact on students'learning outcomes.In addition,the interactive design of chatbots,including feedback types and interaction modalities,significantly moderates learning outcomes.At the same time,we also found that the grades of students and the application context of chatbots also significantly affect the application outcomes.However,the dominant mode of interaction,subject discipline,and learning mode have no significant effects on learning outcomes.Finally,the article looks forward to future research on chatbots,hoping to provide more guidance and inspiration for the education field.