The area of Teachable Agents is one that seeks to study how learning and cognition is improved when a student teaches an agent concepts within a given domain. The agent adaptively interacts with the student to expose misconceptions, prompt elaboration, and encourage the students to persist in the face of failure. While these systems have been generally successful and have potential to address achievement gaps, how to design and implement teachable agents to maximize student learning is still not fully understood.
It has been demonstrated that students involved in peer tutoring experience learning gains themselves, and this is called the Tutor Learning Effect. Instead of peers, however, we utilize a software based agent such as Betty, or a hardware agent such as Quinn. This facilitates the capturing of metacognition data, which in turn allows us to study how students learn, what effective learning habits and successful learning patterns are, and how to develop and refine better open ended learning environments.
Key Research Questions:
- Which aspects of teachable agent interactions directly correlate with the tutor effect?
- How do cognitive and social interactions between a student and teachable agent improve learning outcomes?
- What is the best method to implement a teachable agent?