Moving from Policing to Process Coaching

Tony Wong, School of Mathematics and Statistics
Tony Wong, Associate Professor, Applied Mathematics

When I noticed a clear performance gap between perfect homework grades and poor closed-book exam scores, I realized students were copy-pasting answers from AI. Instead of trying to police their tool usage, I restructured my course to prioritize effort and deep learning. I built a custom NotebookLM tutor bot seeded with my specific course materials to coach students through problems step-by-step, rather than simply dumping out the final answers. I also reduced the grading weight of standard homework, evaluating it purely on honest effort and showing steps, and introduced open-ended, "choose-your-own-adventure" style projects. Finally, I use AI to help me generate qualitative rubrics and am shifting toward activity-based classes so I can directly observe my students' problem-solving process in real-time.

To implement Tony's streamlined approach in quantitative or technical courses, follow these five steps:

Build a course-specific assistant using Google Notebook by uploading your lecture notes, handouts, and homework descriptions. This ensures the AI acts as a 24/7 coach that guides students through the problem-solving process instead of providing a copy-pasteable answer. (Remember that source materials in Google Notebook will be visible to students.)

Reduce the grade weight of regular homework and evaluate it purely on visible steps and completion. This removes the incentive to use AI as a shortcut to get the "correct" final answer.

Replace some repetitive, formulaic assignments with open-ended, "choose-your-own-adventure" style projects that require custom decision-making and are more difficult for an algorithm to automate.

Feed your assignments and solutions into an LLM like Gemini or ChatGPT to generate rubrics with qualitative "buckets" (such as excellent, satisfactory, or needs improvement) to make grading larger, multi-step problems much more efficient.

Use class sessions for some hands-on, active learning exercises. This allows you to watch students solve problems in real-time and address their actual conceptual gaps immediately.

  Click here for another example of how RIT faculty have adapted their teaching methods in response to AI.