Gautam Biswas

Teachable Agents represent a technology that emphasizes the social aspects of learn-ing as much as the cognitive processes. Over several years, our research team has developed Bet-ty’s Brain, a multi-agent environment that utilizes the learning-by-teaching paradigm to help middle school students learn science. In Betty’s Brain, students teach a virtual Teachable Agent (TA) called Betty using a visual causal map representation. Once taught, Betty, can answer ques-tions, explain her answers, and when requested by the student take quizzes, which are a set of questions created and graded by a mentor agent named Mr. Davis. The TA’s quiz performance helps students indirectly assess their own knowledge, and it also motivates them to learn more and improve their TA’s quiz scores. Overall, the learning and teaching task is complex, open-ended, and choice-rich. Thus, learners must employ a number of cognitive and metacognitive strategies to succeed in their tasks. Experimental studies run in middle school classrooms show that students learn science content and do develop some metacognitive strategies through interactions with Betty and Mr. Davis.

Other researchers have also developed Teachable Agent systems, and I will discuss some of them. I will briefly discuss how Teachable Agent technologies have evolved over the years, and look into the future, where advanced AI and Machine Learning technologies can support Teach-able Agents that van interact in more natural ways with their human instructors.