How An AI Detector Made Me Trust People Less

Summary

A personal trial raises questions about labels and trust.

After a week with a social-media detector extension, Watkins reports attending to labels rather than arguments and observing inconsistent classifications across contexts. He favors deliberate disclosure and process explanations.

Read against Mills’s limited use of detection: distinguish classifier accuracy, changes in the reader’s behavior, and opportunities to contest a label.

The author’s personal tests were not reproduced and vendor accuracy was not verified.