Hacker News Reader

Comment by Sir_Cmpwn | Hacker News Reader

There's no way they audited 300,000 accounts and got conclusive results about all of them. I wonder how many innocent accounts were swept up in this?

View on HN

The article EXPLICITLY states: > Of those, roughly 95 percent were identified by the company’s spam-fighting automation tools

So yes, your hunch is right, but it's also mentioned in the article.

Missed that, thanks. Great to know that robots are running our censorship tools now!

What is the alternative? Allow terrorists to spread propaganda? You aren't going to keep up with them without automated tools.

>What is the alternative? Allow terrorists to spread propaganda?

Yes

Well than I would say automated detection is the better of the two options.

"Robots" have run our "censorship" tools for two decades of spam-fighting. Even on the decentralised USENET platform, back when one third of traffic was spam and one third was spam cancels.

To be fair, it also EXPLICITLY states:

> Twitter said about 75 percent of the blocked accounts this year were spotted before a single tweet was sent

So the impact to average users is probably not very substantial.

One of those was probably mine. I was banned within five minutes of signing up, I'm guessing because I was using semi-public wifi and didn't provide a phone number (although the signup form says it's optional.) The stated reason was for "bot-like activity".

How do they know the account was from an actual terrorist or from a Viagra salesbot if it never made a post?

I get spam follows from bots with no posts all the time. It's generally pretty easy to tell who will turn out to be a real person or not — they'll have random, super botty names (Svetlana Nakamoto with the handle jonesdevin), or their profile pics will be stock art or porn stars, or something like that.

Reply on news.ycombinator.com