Teach2Learn: Assessing the Learning among CS1 Students on How Well They Can Teach LLM-Simulated Students
The surge of generative AI in education calls for new learning and assessment approaches that are resistant to cognitive offloading. We explore a solution drawn from the principle of learning by teaching. This poster introduces Teach2Learn, a platform to assess learning among CS1 students by asking them to demonstrate their knowledge by teaching an LLM-simulated peer. We conducted two in-class pilot studies (N = 87) with unique student cohorts: one with non-engineering majors and another with engineering majors. Our preliminary results indicate an increase in student confidence in programming. However, the impact on conceptual understanding was mixed: one cohort showed improvement while the other did not. Teach2Learn presents a framework for teaching in harmony with AI in CS1 courses. Our findings highlight the potential for learning-by-teaching tools while underscoring the need for further research into their pedagogical impact.