A systematic review of pedagogical agent research: Similarities, differences and unexplored aspects

Technological advancements have recently enabled researchers to build increasingly human-like agents with many of these agents being used for educational purposes. Despite the reported successes of these agents, research findings offer a rather inconclusive picture to what extent agent design features contribute to an effective learning experience. This can at least in part be explained by differences in research methods and instruments used to examine the effect of pedagogical agents. The current systematic review provides a comprehensive overview of pedagogical agent research conducted over the last decade (2010–2021), with a focus on experimental design and instrumentation. We systematically reviewed journal articles extracted from five electronic databases. Seventy-five studies were included for evaluation after a three-phase selection procedure. For the analysis, three main directions were investigated: (1) design features and implementations of agents; (2) moderating variables in agent research; (3) instruments utilized to evaluate the effectiveness of agents. The review reveals some shortcomings in the field, including areas to which pedagogical research has not paid much attention to, such as the human-likeness of the agent and the underrepresentation of the K12 population. Furthermore, agents have seldomly been used in virtual reality environments, despite the fact that such environments have long been used in education and have demonstrated their potential to promote learning. Even though pedagogical agents have shown their effectiveness in assisting student learning, further research in new rather unexplored directions is recommended to assess the full potential pedagogical agents in education.