At the 25th Koli Calling International Conference on Computing Education Research in Koli, Finland, Patrick Bassner presented our paper Towards Understanding the Impact of Context-Aware AI Tutors and General-Purpose AI Chatbots on Student Learning, coauthored with Anna Lottner and Stephan Krusche.
In a randomized, mixed-methods study with 33 students, we explored how learners tackled the Burrows-Wheeler Transform with the context-aware Iris tutor, a general-purpose chatbot (ChatGPT), or no AI support. Performance converged across conditions, giving us a strong baseline, while Iris users showed an encouraging uptick in learning scores alongside slightly longer focus time. Interviews brought energetic feedback: context-aware guidance removed friction, guardrails kept learning on track, and students openly discussed how time pressure, trust, and verification shaped their choices.
Read the full paper in the ACM Digital Library.
Key findings
- Virtual tutors need adaptive guardrails: constraints should flex to keep students engaged without pushing them to external systems like ChatGPT.
- Performance differences were not statistically significant, but descriptive trends matched reports of stronger over-reliance in the ChatGPT group—students notice the risk, yet acting on it depends heavily on situational pressure.
- Context awareness in AI tutors was universally appreciated and should remain a design priority for future tutoring systems.
- Patrick Bassner presenting the study at Koli Calling 2025
Resources
Citation
Towards Understanding the Impact of Context-Aware AI Tutors and General-Purpose AI Chatbots on Student Learning
Patrick Bassner,
Anna Lottner, and
Stephan Krusche.
25th Koli Calling International Conference on Computing Education Research
Koli, Finland,
November
2025.