Using teachable agent feedback to support effective learning by teaching

Teachable agents build upon research showing that students can learn by teaching their peers. In these systems, students learn by teaching simulated pupils who can be designed to support productive types of feedback and teacher-pupil interactions. We found that students learned better when they taught an agent designed to mimic a self-regulated learner (i.e. pushed the teacher to ask questions, and probe the agent for explanations to check her understanding). Analyses of log file data showed that interacting with the agent helped students engage in beneficial learning and teaching behaviors