Christopher Petrie "Learning-by-Teaching AI Students in Computing Education"

This talk can be viewed via zoom. (Note: this talk will be recorded)

Abstract: 

Students can now easily generate working code without being able to explain why it works or how to fix it when it goes wrong. This creates a major challenge for computing education: How can AI support students while preserving the learning needed to develop independent understanding? Learning-by-teaching offers one scalable approach by reversing the role of AI, placing students instead in the role of explaining concepts to an AI learner.

This public talk first presents research on Explique at the Swiss Federal Institute of Technology of Lausanne in Switzerland (EPFL), which is an AI teachable agent deployed over 11 weeks in an undergraduate algorithms course with 546 enrolled students. Analysis revealed that there was evidence of sustained participation at scale and changes to the depth of students' explanations. Additionally, results indicated that students needed slightly fewer quiz attempts after teaching their AI than reading explanations. However, differences in the time spent and the absence of delayed assessments limited conclusions about its learning impact. 

Building on these findings and recent research on simulating novice programmers by Assistant Professor Juho Leinonen at Aalto University, I propose extending our work to investigate whether learning-by-teaching interactions encourage students to actively build their knowledge and improve delayed unaided learning. This project intends to contribute to computer science education and human-centric computing by rigorously testing how an AI learner with persistent and responsive difficulties shapes a student's reasoning and subsequent understanding. The wider ambition is to make explanatory practice more available and to understand its learning value alongside human teaching.

Bio:

Christopher Petrie is a computing education researcher with experience in classroom teaching and developing learning technologies. He taught computer science in New Zealand before completing his PhD at the University of Helsinki (awarded with distinction) where he studied the design and user experiences programming platforms that aim to develop computational thinking skills through music. Recently, he was a postdoctoral researcher at EPFL — Swiss Federal Institute of Technology of Lausanne in Switzerland at the CHILI Lab with Professor Pierre Dillenbourg. There Christopher taught web accessibility and usability research in their undergraduate Human-Computer-Interaction course. He also collaborated with Professor Ola Svensson and colleagues on the design and research of a learning platform called Explique. Additionally, Christopher has developed and founded a learning technology company called Kwizie.ai in Finland which has had 200000+ users globally.