Process Over Detection – What Does It Actually Look Like in Practice?
I hear “process over detection” as the answer to AI academic integrity but I rarely see it explained in practical terms. Here’s what we’ve implemented at the board level and what the evidence says after one academic year.
What process documentation means in practice:
Students submit a learning portfolio alongside any major assignment. Minimum requirements: an initial brainstorm (rough, unedited), a first draft with their own track-changes visible, a brief process note explaining major changes between draft and final.
Results after one year:
– Formal AI dishonesty complaints: down 40% vs previous year
– Student appeals of findings: down 60%
– Teacher satisfaction with the integrity process: up significantly
Why it works: AI cannot fake an authentic messy first draft that shows real thinking. The process portfolio provides positive evidence of learning, not negative evidence of cheating. It’s legally defensible in a way detection scores aren’t.
The upfront cost: redesigning assignment structures. Higher per-student prep time initially. But no detection workflow costs, no false accusation management, no parent complaints about scores.
5 Replies
Join the discussion.
Log In to Reply
Principal perspective: the reduction in formal complaints matters as much as the educational outcomes. Every formal AI dishonesty complaint takes significant admin time. If process documentation reduces that by 40% it's cost-effective even before considering the improved educational experience.
the legally defensible point is the one that should get admin's attention. detection-based findings can and will be challenged. process portfolios produce positive evidence of learning. different standard entirely.
implemented something similar in my grade 11 class. three mandatory check-ins during any major writing assignment - outline, first draft, final. AI can help at each stage but the student has to be present at each check-in. haven't had a single integrity concern this semester.
the process-over-detection approach scales best when it's embedded in assignment design from the beginning, not layered on as a monitoring mechanism after the fact. schools that are retrofitting process requirements onto existing assignments are finding it awkward because the assignments weren't built around process evidence to begin with. new course design is where this has to start.
the 40% drop in formal complaints is the number I needed to make this argument to my department. I've been saying process documentation works better for two years. now i have an actual outcome to cite. sharing this immediately.