The emergence of generative artificial intelligence (GenAI) models, including ChatGPT, Gemini and Copilot, has transformed educational pedagogy by enabling new forms of interaction, reflection and learning. Yet the use of GenAI in software development education often fails to support deep conceptual learning. In a new article, researchers introduce Teachable AI, a prototype learning system that applies a learning-by-teaching approach in which students explain newly learned concepts to an AI agent.
The article was written by researchers at Carnegie Mellon University, American Heritage Schools and the University of Rome. It was presented at the 14th International Conference on Information and Education Technology and published in IEEE Xplore.
"The widespread availability of GenAI has introduced new opportunities for learning in computing education, particularly in courses on software development, where students often struggle to grasp abstract and fundamental principles and concepts," said Shihong Huang, teaching professor of information systems at Carnegie Mellon's Heinz College, who coauthored the article. "Using our system, students act as teachers by explaining coding concepts to the Teachable AI bot and deepen their understanding of the subjects."
