How Virtual Agents Can Shape Human-Human Collaboration: A Systematic Review
摘要
Virtual agents have demonstrated considerable success in fostering learning experiences. Existing literature reviews of the field have found that in the context of virtual agent support for individual learners, agents enhance the social and cognitive learning experience. However, these reviews have not examined how virtual agents can facilitate collaboration between two or more human learners. This systematic review addresses this gap. We identified and analyzed 55 studies reporting on virtually embodied or text-based agents that facilitate collaboration among two or more human learners. We found that researchers operationalize and measure collaboration through a variety of lenses, including academically productive talk, knowledge building, and collaborative decision-making. The analysis also revealed a set of behaviors that virtual agents perform to support collaborative learning activities, including scaffolding collaborative behaviors, consolidating students’ ideas, and encouraging quieter students to participate in the conversation. We identify gaps and opportunities for refining and unifying existing frameworks and establishing shared evaluation methods. These advancements hold the potential to enhance the effectiveness of virtual agents supporting human-human collaboration.