Post #3615310
2026-07-05 20:00 UTC
How dangerous is an AI-writing detector that is mostly accurate? A profile of Pangram argues a mostly-right detector is worse than a broken one: at a one-in-10,000 false-positive rate across 30M+ students turning in dozens of assignments each, the wrongful accusations never stop. And the black-box verdict leaves the accused nothing to appeal.
https://benjaminhan.net/posts/20260705-pangram-ai-detection-accuracy/?utm_source=mastodon&utm_medium=social
#AI #Education #Ethics
Replies (2)
-
@ramsey@phpc.social 2026-07-05 20:05
@BenjaminHan@sigmoid.social @paninid@mastodon.world Similarly, my wife was finishing her degree a few years back (before LLMs were used for this stuff), and all her work was scanned by a plagiarism detector. It seemed more harmful than helpful.
-
@antonio106@ottawa.place 2026-07-05 23:01
@BenjaminHan@sigmoid.social in my experience having taught a few college courses, it was *obvious* when someone was using an LLM, when the finished product was so much better than the student in front of you. What my college did to circumvent this was get rid of essay questions altogether and make the full course multiple choice. Which, BY PURE COINCIDIECE, means a whole bunch of us sessional instructors didn't get a new contract the next year.