Effectiveness of collaborative learning in VR: a meta-analysis

英文题目：The Effects of Pedagogical Agents on Learner's Motivation and Learning Outcomes  :  A meta-analysis

1.Introduction

With the development of artificial intelligence and virtual reality technology, the original digital educational resources have formed a learning resouce environment with intelligent characteristics through intelligent agents that sense external demands and use intelligent tutors and intelligent learning companions. A series of complex, visual avatars (pedagogical agents) are gradually appearing in the Intelligent Guided Learning System.Pedagogical Agents are virtual,teacher-like roles uesd in e-learning environments to meet a varity of instructional objectives. They often act as teachers or learning motivators, using language or facial expressions to communicate with learners. They are also often integrated into the e-learning environment to provide cognitive support (Baylor,1999) and socioculturally inclusive learning experiences.Instructional agents can not only save labour costs and personalise teaching, but also increase learners' interest in learning and improve learning outcomes (Johnson, Ozogul, & Reisslein, 2015). However, some suggest that teaching agents can increase learners' cognitive load and distract learners' attention. Therefore, there is still no agreement on whether instructional agents can motivate learners and whether they can improve learning outcomes.

2.Theoretical framework

对相关理论开展综述，主要包括以下4个理论：

2.1 拟人效应 - Persona Effect

Lester et al. (1997) experimental research direction that the inclusion of instructional agents in interactive learning environments positively affects the perception of the learning process and further facilitates learning. Therefore, he proposed the persona effect on the basis of this experiment. The Persona Effect refers to the fact that if a persona characters exists in an interactive learning environment, it is able to positively influence students' learning even if the characters are not very expressive. Although the Persona Effect states that the inclusion of pedagogical agents in multimedia learning can improve learners' process-awareness experience and academic performance, the Persona Effect does not give a detailed theoretical explanation of how pedagogical agents can positively influence learners' process-awareness. The Social Presence Theory gives a more reasonable explanation for this problem.

2.2 社会存在感理论 - Social Presence Theory

Social Presence Theory was first proposed by Short, Williams and Christie (1976) as a theoretical hypothesis to explain the impact of electronic media on communication.Social presence refers to the extent to which a person is perceived as a "real person" and the extent to which people perceive a connection to them when communicating through the media (Gunawardena, 1995; Lowenthal, 2009). Social presence has been found to be an important potential factor in improving teaching and learning outcomes, and in addition to playing an important role in real classroom environments, it also plays an important role in computer-based learning environments, positively predicting learning satisfaction(Gunawardena & Zittle,1997;Richardson & Swan,2003).Based on this theory, the inclusion of pedagogical agents in learning improves the perceived experience of the learner's learning process, increases interest and motivation, and reduces the learner's perceived task difficulty and mental effort (Frechette & Moreno, 2010). Social presence theory explains the relationship between instructional agents and the perception and experience of learning, but does not explain how instructional agents affect learning outcomes.Social agency theory gives another perspective on how instructional agency affects learning outcomes.

2.3 社会代理理论 - Social Agence Theory

Learning is not only a cognitive activity, but also a social activity, even if the learner sits in front of the computer and learns the knowledge alone it is still a social behaviour, so multimedia learning is also a social activity that needs to collaborate with the learner and the explainer (Mayer, & Dapra, 2012). And the previous emphasis only on learners' cognitive processing in multimedia learning is incomplete. Therefore, Mayer and Dapra (2012) regarded the agent image as a kind of social information and added the social dimension to the cognitive theory of multimedia learning to explain the role of the agent image in multimedia learning. This is Social Agency Theory (SAT).

Social agency theory is a commonly used theory in agency image research (Atkinson et al., 2005; Mayer, 2005; Moreno, Mayer, & Spires, 2001).Mayer and Dapra (2012) proposed a social agency theory model of pedagogical agency: firstly, social presence perceptions of a computer in learning can be triggered by small social cues, such as the presentation of agentic images and non-verbal behaviours; and second, once these social communication schemas are triggered, the learner may feel that he or she is communicating with another real person. Next, the learner tries to switch on its deep cognitive processing strategies to understand what the computer is explaining.Mayer and Dapra (2012) explain deep cognitive processing in learner learning as learners putting more effort into selecting relevant information, organising fragmented information, and integrating it into holistic representations; and most importantly, integrating these holistic representations with their knowledge and experience to encode them in memory, resulting in meaningful learning outcomes, which are mainly in the form of being able to better apply or transfer acquired knowledge to new situations. or transfer to new situations. Thus, social agency theory suggests that agentic images will have higher social awareness and interest in the learning process, and will be more effective in transfer performance.

2.4 认知负荷理论 - Cognitive load Theory

Cognitive load theory attempts to explain how different tasks store different cognitive demands in a limited capacity working memory. Cognitive psychologists believe that humans process information using a "three-element system", which consists of sensory memory, short-term memory, and long-term memory (Baddeley, 1992). Cognitive load theory relates to the short-term memory (also known as working memory) and long-term memory components of the human cognitive structure.Cognitive load theory focuses on which types of information are processed in working memory. According to Baddeley (1992), working memory is divided into multiple channels. The load on working memory may be influenced by the nature of the learning task (intrinsic cognitive load) as well as the design of the instructional material. For example, the design of instructional materials may affect the cognitive processes associated with learning (or cognitive schema) (external cognitive load)  and it may affect related cognitive structures such as concrete processing, construction, and automated manipulation (related cognitive load).

The core principle of cognitive load theory is to increase relevant cognitive load and decrease external cognitive load (Kester et al., 2006). Cognitive load theory suggests that dedicated instructional agent information that is not relevant to the content/task (e.g., a large number of facial expressions that are less relevant to the instructional goal) increases the external cognitive load on the learner because it also requires the learner to invest a certain amount of cognitive effort, and that investing limited cognitive resources in information or media that are less relevant to the learning task will inevitably increase the cognitive load and thus have a learning negatively (Clark & Choi, 2005), as learners need to divide their attention between the large number of visual elements of the teaching agent, or between the teaching agent and other information on the screen.

3.Method

3.1 确定研究方法为元分析，选择研究工具；

3.2 确定文献筛选标准，保留符合条件的文献；

以Web of Science、ScienceDirect、Springer Link、EBSCOhost、Education为检索库，以"Pedagogical Agent、Conversational Pedagogical、Animated Agent、Teaching Agent"交叉组合"Learning Outcomes、Motivation"为检索词，时间跨度为2010-2023年

3.3 对收录的文献开展编码与分析；

4.Result

4.1 研究样本的发表偏倚与异质性检验；

4.2 教学代理对学习者学习动机的影响；

4.3 教学代理对学习者学习效果的影响；

5.Discussion

5.1 讨论元分析过程中与设想不符的情况；

6.Conclusion

6.1 对研究过程与结果进行总结；

6.2 提出相关意见。