Trying to prove students are not robots is pushing them toward more AI

Hacker News discussion: https://news.ycombinator.com/item?id=47290457
Primary source: Techdirt article

This Hacker News thread surfaced a sharp argument from Techdirt: once schools start treating “looks human” as the target, students optimize for that metric instead of for clear thinking or good writing. The article describes a loop in which teachers distrust AI detectors, students learn that polished prose can trigger suspicion, and both sides end up playing a game around detection scores rather than learning outcomes.

What the article is saying

• Students are changing their writing style to avoid false positives from AI detectors.
• Some are intentionally making prose flatter or rougher so it reads as “authentically human.”
• Others are using services such as GPTZero not to avoid AI entirely, but to test whether their work will survive automated scrutiny.

The core criticism is that detection-centric assessment creates perverse incentives. If the system rewards text that merely appears unassisted, students may be nudged toward more hidden AI use, more score-gaming, and less revision. That is especially damaging in writing-heavy courses, where iteration and cleanup should be part of the craft rather than a signal of misconduct.

For AI and education practitioners, the useful takeaway is not that schools should ignore misuse. It is that weak detectors and adversarial grading heuristics can distort behavior at scale. Better policy probably means process-based assessment, oral defense, staged drafts, and clearer rules around acceptable assistance instead of a binary “human or machine” test.