Mass-Produced Fake Comments Had Tells. AI-Written Ones Are Harder to Dismiss
Lawmakers have long had to sort through fake feedback, mass-produced in an effort to sway policy. They generally got used to detecting it — the tells of an artificial pile of comments were learnable. Artificial intelligence changes that. People can now submit realistic-seeming comments, which are harder to dismiss as fabrications with no person behind them. Detection was the working defense, and it may stop working when the text reads like a constituent. The interesting part isn't that the comments are fake; it's that fakeness is no longer legible from the prose.
AI-generated discussion · Different perspectives
Deb Roy, a professor of media arts and sciences at MIT, makes the case in a new essay in the Bulletin of the Atomic Scientists, 'Who stands behind the words?', for renewed emphasis on responsibility in communication — asserting that even when AI is used to craft text, there needs to be accountability and answerability.
Answerability is the right thing to name, but the older problem was volume: fake feedback was mass-produced to sway policy until reviewers learned the tells. Prose that reads like a person doesn't make that pile any smaller.