Hacker News Reader

Comment by ohmaigad | Hacker News Reader

The issue is that there is no way their algorithm understands context. Saying "Fuck nazis" will probably flag you for swearing and being a nazi. An actual person looking your tweets/posts can understand the context.

View on HN

An actual person looking at a list of tweets/posts hundreds-long is not going to spend more than a few seconds at most on average per tweet. That's not enough time to understand context. As a matter of fact, I've given up trying to engage on social media debates precisely because most of the time, the other party(ies) in the conversation still do not understand context despite being part of the conversation! Often because they are not spending more than a few seconds themselves. I remember conversations being a lot more civil when done by e-mail.

You don't need a lot more than a few seconds to understand that "Fuck, i am tired" or "Fuck nazis" is not something inherently bad. For an algorithm - i am not so sure (even a company as huge as youtube cannot figure it out).

That's a poor example. How about a message that gets flagged, where the message is a strong frustration at an issue that is very nuanced? Do you jump to conclusions on the person's point of view? Do you assume that the person is prone to being angry, just because you've seen a selection of angry posts from them?

There's many things that can go wrong there, and they will, given a system like this operating in the wild.

Reply on news.ycombinator.com