中文摘要：

教学代理（Pedagogical Agent）是指为了促进学习者的学习而在计算机屏幕上呈现的人物形象,主要包括内部特征和外部特征。对相关理论进行梳理发现,教学代理能否促进学习存在争论：拟人效应、社会存在感理论和社会代理理论认为在多媒体环境中加入教学代理能促进学习,干扰理论则认为教学代理会阻碍学习。综述以往的实证研究发现,教学代理特征（代理类型、语词讲解、声音类型、手势有无）、学习环境特征（学科性质、呈现步调）以及学习者特征（先前知识经验、被试群体）可能会影响教学代理的效果。通过元分析发现,对于不同的测量指标,教学代理的有效性不同：教学代理能显著提高保持测验（g=0.19）、迁移测验（g=0.39）和其他测验（g=0.31）成绩,但对学习动机和学习兴趣的影响不显著。该结果支持了社会代理理论,社会存在理论有待验证。另外,元分析发现对于不同的测量指标,起调节作用的变量不同：用系统合成的声音讲解学习内容时,教学代理才能提高学习兴趣（g=0.81）;学习者为非成人时,教学代理才能提高保持测验成绩（g=0.51）;当教学代理伴随手势动作（g=0.67）,学习材料为科学知识（g=0.46）时,教学代理能提高迁移测验成绩。未来研究需要关注：在多媒体环境中加入教学代理是否会吸引学习者的注意,干扰对核心内容的加工;教学代理能够促进学习的内在机制;教学代理的手势动作在多媒体学习环境中的作用。

英文摘要：

Pedagogical agents are lifelike characters presented on a computer screen that facilitate learners＇ performances.Pedagogical agents in the multimedia learning are important in computer-assisted learning and instructional design. First, we reviewed the definition of pedagogical agent, and introduced theories related with pedagogical agent. It was found that the effects of pedagogical agent on multimedia learning were debatable. Second, based on the empirical studies, a meta-analysis was conducted to examine the positive and negative effects of pedagogical agent on learning. Results of meta-analysis revealed that pedagogical agent could effectively improve scores on the retention tests（g = 0.19）, transfer tests（g = 0.39） and other tests（g = 0.31）. However, it did not facilitate learners＇ motivations and interests. These results were consistent with the Social Agency Theory. However, the Social Presence Theory still needs to be further tested. Third, moderator analyses indicated that different variables were moderated by different factors, such as voice, gesture, disciplines and self-paced or computer paced learning. Pedagogical agents with computer-edited voices facilitated learners＇ interests（g = 0.81） than those without voice. Pedagogical agents were more beneficial to K-12 students（g = 0.51） than adults. Pedagogical agents improved participants＇ scores on transfer tests when the materials focused on science and technology（ g = 0.46） rather than the humanities. Moreover, compared with no gestures, pedagogical agents with gestures could improve participants＇ scores on transfer tests（g = 0.67）. Future research should explore the effect of pedagogical agents extensively on learners＇ attention, and the mechanisms underlie the beneficial effect of pedagogical agents, as well as pedagogical agent＇s gestures on learning.