Peter Fowles, Erik Falor, Sulove Bhattarai, John Edwards, Seth Poulsen (2026) — Utah State University. arXiv:2605.21374 [cs.HC]
Oral code reviews paired with a flipped classroom represent a pragmatic harm-reduction approach to generative AI in CS education. Rather than banning LLMs, Fowles et al. (2026) designed weekly formative assessments where students must explain their submitted code regardless of its origin. Over three semesters at Utah State University, keystroke logs confirmed significantly higher AI usage (increased paste-to-type ratios), yet exam scores showed no statistically significant decline. Students reported positive attitudes toward the code review format, with negative feedback centered on logistical issues (scheduling, TA training) rather than pedagogical objections.
This study addresses a core tension in the ai-k12-evidence-base: how to maintain academic-integrity while acknowledging that AI tools are now part of professional practice. The approach aligns with ai-literacy frameworks that emphasize critical engagement over prohibition. Unlike AI detection tools, which create adversarial dynamics, oral assessments reframe the conversation around understanding and metacognition. The findings also complement research on over-reliance and cognitive-offloading, suggesting that structured accountability mechanisms can preserve learning outcomes even as AI usage rises.
The flipped classroom component is particularly notable — it creates dedicated time for one-on-one assessments without sacrificing content coverage, addressing a scalability concern that has limited adoption of oral assessment in large CS1 courses. This connects to broader discussions about formative-assessment design in AI-rich environments and the evolving role of teacher-role as learning facilitator rather than knowledge dispenser.
Related Pages
- ai-k12-evidence-base — related
- ai-literacy — related
- over-reliance — related
- academic-integrity — related
- cognitive-offloading — related
- formative-assessment — related
- teacher-role — related
- metacognition — related
- student-experience — related
- cs-education — related
- buggy-genai-code-student-responses — When AI Is Wrong on Purpose: How Students Respond to Buggy GenAI Code
Citation
APA: Peter Fowles, Erik Falor, Sulove Bhattarai, John Edwards, Seth Poulsen (2026). Combating Harms of Generative AI in CS1 with Code Review Interviews and a Flipped Classroom. arXiv:2605.21374. arXiv:2605.21374 [cs.HC].