4 ms·

I think what they meant was that they detected bot activity associated with the account.

by q1w2 5y ago

logicalmonster 5y ago
There's multiple huge problems if Twitter takes this approach to rules enforcement even if they're genuinely trying to be honest and unbiased. 1) Every account on Twitter has "bot activity associated with this account": if for no other reason that random advertising bots retweet, follow, or like random people and posts to attempt to blend in as a genuine account and hide their true purpose. This reasoning gives Twitter essentially carte blanche to delete everybody, even assuming they were willing to be fully transparent about the evidence surrounding their bans (which they're not). 2) Bad Actors can easily sabotage accounts they dislike. Here's just 1 method how this can be done: there's 10,001 shady fly by night firms in India, Pakistan, Nigeria, Bangladesh, Vietnam, and other countries that offer services to "buy followers" or "buy retweets". It's trivial for somebody without programming skill and even a few hundred dollars to just buy mass spam and shill the hell out of an account (the lower quality the better) until somebody notices and that account gets banned. You don't even have to assign bad motives to Twitter to see the obvious problems here.
- afavour 5y ago
  I suspect a key difference here is “all of the users other twitter accounts”. If a bunch of accounts registered to the same email address all start boosting each other that’s probably pretty easy to detect, and distinct from someone else trying to do it to you.
- joshuamorton 5y ago
  You appear to be assuming that the entire team of people at twitter who work on these problems either haven't considered or aren't able to address these issues. While there often are challenges with moderation and rules enforcement, this kind of stuff is pretty easy to address.
  - dmpk2k 5y ago
    Given that I have wrestled with 2) in a different context, I'm quite interested in hearing the various (hopefully realistic) methods to address it. Actually serious.
    - logicalmonster 5y ago
      I would like to know too. But if you look at that profile, he says he works at Google. Think about that for a second. He claims this is an easy problem. But consider Google's approach to spam detection algorithms and see how there are so many false positives and innocent people getting screwed out of their ad accounts and livelihood. If this kind of stuff was easy, why hasn't anybody informed Google about it? I'm not trying to pick on anybody here, but I believe there's a level of arrogance in thinking that this form of sabotage is somehow a solvable problem unless you somehow have access to all communications on the planet.
      - joshuamorton 5y ago
        I very specifically said that there were hard problems, just that this wasn't one ;) In this case, a fairly straightforward account age rule would get you like 95% of the way there. My twitter account, which is 7 years old and has a few hundred followers and tweets once a week isn't going to benefit from this. You don't ban it. Similarly, you don't ban a 7 year old account with 10s of thousands of followers that's been doing the same kind of tweeting for 2 years and recently got a bunch of bot follows. On the other hand, a 3 week old account with a ton of bot follows is suspicious. This'll miss older accounts that were unused and changed (e.g. if my account suddenly started tweeting 17x a day about divisive political content and suddenly got a bunch of bot follows, so you need slightly fancier approaches). You probably further improve this with some kind of multifaceted account realness score based on a pagerank like analysis of followers (following bots lowers your score, as does being followed by them). And ofc you still ban the bots, but you don't punish any accounts above a certain threshold of "real". The other part of it is that randomly sabotaging twitter accounts isn't going to happen very often, it costs money, and unless there's an ROI (extorting people), no one is going to do it except a handful of people for teh lulz or whatever. But importantly, most of those people will be bad at doing this kind of attack in a convincing way. There are lots of difficult problems, but OP chose one that doesn't happen and is straightforward to address.

I think what they meant was that they detected bot activity associated with the account.

by q1w2 5y ago

I think what they meant was that they detected bot activity associated with the account.

There's multiple huge problems if Twitter takes this approach to rules enforcement even if they're genuinely trying to be honest and unbiased. 1) Every account on Twitter has "bot activity associated with this account": if for no other reason that random advertising bots retweet, follow, or like random people and posts to attempt to blend in as a genuine account and hide their true purpose. This reasoning gives Twitter essentially carte blanche to delete everybody, even assuming they were willing to be fully transparent about the evidence surrounding their bans (which they're not). 2) Bad Actors can easily sabotage accounts they dislike. Here's just 1 method how this can be done: there's 10,001 shady fly by night firms in India, Pakistan, Nigeria, Bangladesh, Vietnam, and other countries that offer services to "buy followers" or "buy retweets". It's trivial for somebody without programming skill and even a few hundred dollars to just buy mass spam and shill the hell out of an account (the lower quality the better) until somebody notices and that account gets banned. You don't even have to assign bad motives to Twitter to see the obvious problems here.

afavour 5y ago
I suspect a key difference here is “all of the users other twitter accounts”. If a bunch of accounts registered to the same email address all start boosting each other that’s probably pretty easy to detect, and distinct from someone else trying to do it to you.
joshuamorton 5y ago
You appear to be assuming that the entire team of people at twitter who work on these problems either haven't considered or aren't able to address these issues. While there often are challenges with moderation and rules enforcement, this kind of stuff is pretty easy to address.
- dmpk2k 5y ago
  Given that I have wrestled with 2) in a different context, I'm quite interested in hearing the various (hopefully realistic) methods to address it. Actually serious.
  - logicalmonster 5y ago
    I would like to know too. But if you look at that profile, he says he works at Google. Think about that for a second. He claims this is an easy problem. But consider Google's approach to spam detection algorithms and see how there are so many false positives and innocent people getting screwed out of their ad accounts and livelihood. If this kind of stuff was easy, why hasn't anybody informed Google about it? I'm not trying to pick on anybody here, but I believe there's a level of arrogance in thinking that this form of sabotage is somehow a solvable problem unless you somehow have access to all communications on the planet.
    - joshuamorton 5y ago
      I very specifically said that there were hard problems, just that this wasn't one ;) In this case, a fairly straightforward account age rule would get you like 95% of the way there. My twitter account, which is 7 years old and has a few hundred followers and tweets once a week isn't going to benefit from this. You don't ban it. Similarly, you don't ban a 7 year old account with 10s of thousands of followers that's been doing the same kind of tweeting for 2 years and recently got a bunch of bot follows. On the other hand, a 3 week old account with a ton of bot follows is suspicious. This'll miss older accounts that were unused and changed (e.g. if my account suddenly started tweeting 17x a day about divisive political content and suddenly got a bunch of bot follows, so you need slightly fancier approaches). You probably further improve this with some kind of multifaceted account realness score based on a pagerank like analysis of followers (following bots lowers your score, as does being followed by them). And ofc you still ban the bots, but you don't punish any accounts above a certain threshold of "real". The other part of it is that randomly sabotaging twitter accounts isn't going to happen very often, it costs money, and unless there's an ROI (extorting people), no one is going to do it except a handful of people for teh lulz or whatever. But importantly, most of those people will be bad at doing this kind of attack in a convincing way. There are lots of difficult problems, but OP chose one that doesn't happen and is straightforward to address.