Hacker News Reader

In contrast to all the dubiously useful AI applications springing up lately, one that would be great is an AI that can reliably and accurately detect bots, shills, astroturf, disguised advertisements and other sponsored content masquerading as organic. I’d install a browser plug-in that could put a warning under comments and tweets that are detected as fake or a bad faith actor.

View on HN

There's a lot of research papers on this topic and you can detects some kinds of bad behavior consistently, but mostly involving spamming. Sophisticated information operations distribute the work across multiple participants, and the technological side of it is a constantly evolving arms race.

I would totally pay for this

Even just a slight deboost to "probability of shill". I'm good enough at noticing it myself, but not until I wasted 5 seconds reading their post.

I would as well... but how would you actually target an actual shill? You'd need a dataset of confirmed shills.

If you have a dataset of people who you speculate to be shills, doesn't that trend close to a dataset of people who's opinion you don't like?

To be clear, I'm not saying that wouldn't be personally valuable... but uh, if I was creating a system like this for myself, I'd curate that dataset VERY carefully (assuming I could even get the system to work :\\\)

Reply on news.ycombinator.com