AI detectors flag neurodivergent writing as AI-generated, causing real harm
sanjanasinghx · x · 2026-10-19
A viral thread highlights that AI detection tools, including Turnitin, wrongly flag neurodivergent writing (autism, ADHD, dyslexia) as AI-generated — a distressing false-positive problem.
Why it happens: detectors rely on perplexity (word predictability) and burstiness (sentence-structure variation). Neurodivergent prose tends to be literal, structured, and uniform, scoring low on both — statistically mirroring AI text. The BBC has documented cases of autistic students whose entirely original, handwritten work was flagged 100% AI-generated; formal academic prose is similarly vulnerable, leading to false academic-dishonesty accusations.
More from AGI Musings
- Mathematician argues we may be entering a golden age of mathematics — thebasepoint · 2026-10-19
- Google Research grounds AI agents in lab robotics, finds nanomolar cancer protein binders — DeryaTR_ · 2026-10-19
- A decade after Theranos, skeptics ask if AI diagnostics will be worse — krishnan · 2026-10-19
- Pyramid Replacement: how AGI could hollow out companies from entry level up — luke_drago_ · 2026-10-19
- Hinton's 'Stop Training Radiologists' Prediction Aged Badly, and Everyone Noticed — chrisalbon · 2026-10-19
- Many open math problems turn out to be 'obvious' relative to AI — thebasepoint · 2026-10-19