One of my bigger concerns with AI is its potential to spread misinformation, but does anyone know of folks using the models for detecting it? I'm not into the AI hype but I feel that's one area that has some potential benefit.
I find trusting AI to detect misinformation to present essentially the same problem as using AI to spread it - both cases require implicit trust in an untrustworthy system, and both remove the impetus to educate people and have them practice critical thinking and skepticism (as opposed to cynicism and contrarianism) themselves.
Although there was a paper posted here recently that suggested AI was effective at deprogramming conspiracy theory[0,1]. No idea about its credibility but claims like being able to reduce belief in conspiracy theory by "20%" seem iffy to me.
The most effective system I've seen so far, ironically (given who now runs the site) is the community notes on Twitter. But even that gets gamed by activists and bad faith actors. Still I think removing human beings from the loop isn't always helpful.
Comments
One of my bigger concerns with AI is its potential to spread misinformation, but does anyone know of folks using the models for detecting it? I'm not into the AI hype but I feel that's one area that has some potential benefit.
I find trusting AI to detect misinformation to present essentially the same problem as using AI to spread it - both cases require implicit trust in an untrustworthy system, and both remove the impetus to educate people and have them practice critical thinking and skepticism (as opposed to cynicism and contrarianism) themselves.
Although there was a paper posted here recently that suggested AI was effective at deprogramming conspiracy theory[0,1]. No idea about its credibility but claims like being able to reduce belief in conspiracy theory by "20%" seem iffy to me.
The most effective system I've seen so far, ironically (given who now runs the site) is the community notes on Twitter. But even that gets gamed by activists and bad faith actors. Still I think removing human beings from the loop isn't always helpful.
[0]https://news.ycombinator.com/item?id=39927578
[1]https://osf.io/preprints/psyarxiv/xcwdn