基于生成式人工智能的人机协同学习更能提升学习成效?——基于20项实验和准实验的元分析

Does Human-computer Collaborative Learning Based on Generative Artificial Intelligence Enhance Learning Outcomes?-A Meta-analysis of 20 Experimental and Quasi-experimental Studies

生成式人工智能的崛起为人机协同学习注入了动力。然而,学界对基于生成式人工智能的人机协同学习成效仍有争议。本研究采用元分析法对国内外20项研究进行整合分析,并从学科领域、知识类型、干预时长等维度探讨调节变量对基于生成式人工智能的人机协同学习成效的影响。结果显示,相比于传统学习方式,基于生成式人工智能的人机协同学习能有效提升学习成效。调节分析表明,基于生成式人工智能的人机协同学习在社会科学领域与程序性知识学习中的表现更好;干预周期越长影响越弱;群体学习、角色设定与翻转课堂对人机协同学习成效的影响最突出;学科领域、知识类型、学习方式等的组间差异不显著。研究者需加强实践适切性设计,开展围炉群证式学习,注重角色精细化指引和探索模式深度融合,以助推基于生成式人工智能的人机协同学习理论与实践的发展。

The rise of Generative Artificial Intelligence(AI)has energized human-computer collaborative learning.However,there is still debate within the academic community regarding the effectiveness of collaborative learning facilitated by generative AI.This study utilized a meta-analytical approach to synthesize 20 domestic and international studies and examined the impact of moderating variables such as academic discipline,type of knowledge,and duration of intervention on the effectiveness of human-computer collaborative learning based on generative AI.The results indicate that,compared to traditional learning methods,human-computer collaborative learning enabled by generative AI enhances learning outcomes more effectively.The moderation analysis shows that the collaborative learning performs better in the field of social sciences and in the learning of procedural knowledge.Longer intervention periods correlate with weaker effects.Group learning,role-setting,and flipped classrooms have the most pronounced impact on the effectiveness of collaborative learning.Differences in academic disciplines,types of knowledge,and modes of learning are not significant among groups.Based on the results,the study suggests that researchers need to strengthen the design of practical applicability,conduct hearth-group evidence-based learning,enhance role-specific guidance,and explore deeply integrated learning models,to further advance the theoretical and practical development of human-computer collaborative learning integrated with generative AI.