What is a false positive?
A false positive occurs when human-written work receives an elevated AI-writing score. This happens most often with formal academic tone, uniform formatting, translated or edited prose, and short submissions with little scored text.
When we suppress or mask scores
Below 20% overall AI detected, we often mask exact percentages because the signal is weak. At 0% with no qualifying evidence, we show a calm green card instead of alarming language. These choices reduce harm from noisy measurements.
Populations at higher risk
Multilingual writers, students trained in template-heavy exam writing, and STEM students new to humanities essays may score higher without AI use. Slide decks and reference-heavy PDFs can distort excerpts if extraction includes bibliography text — use submitted-content-only mode.
Practical de-escalation steps
Compare the flagged passage to the student's prior work. Look for specific detail only they would know. Request a short in-person explanation of one argument paragraph. Prefer revision assignments over hearings when evidence is thin.
Documenting fairness
Integrity committees should log detector version, confidence label, and whether paraphrase risk was confidently detected. Reports include disclaimers for this purpose. Students should receive flagged excerpts, not just a number.