An Explainable AI Assistant for Introductory Programming Education: Improving Feedback Reliability with Instructor-AI Collaboration

AI systems work better when instructors remain in control: The assistant improved reliability by having instructors define the pedagogical knowledge upfront—mapping student errors to known misconceptions and authoring feedback themselves—rather than relying solely on the AI to generate responses independently.

Explainability matters for educational trust: The system prioritized explainable AI methods that could show why it identified a particular error in student code, which helped ensure instructors could verify the AI's reasoning before it reached students, addressing concerns about "black box" AI in learning environments.

Timely, personalized feedback at scale is a real problem: The paper identifies that students in introductory programming courses often lack access to sufficient instructor feedback due to class size and workload constraints—a gap the AI assistant was designed to help fill while maintaining quality.

Human-AI collaboration outperformed either alone: Expert evaluation confirmed the system provided accurate, instructor-verified feedback, and classroom deployment showed students had positive experiences with it, suggesting the hybrid approach of AI analysis + instructor oversight was more effective than either approach independently.

Responsible AI in education requires validation beyond just performance metrics: The researchers conducted both expert evaluation (instructor verification) and real classroom deployment (student perception assessment), demonstrating that responsible AI implementation needs multiple forms of evidence, not just technical accuracy scores.