Comment by ilamont | Hacker News Reader

Comment by ilamont | Hacker News Reader

Suspicions were aroused after researchers found a third of the accounts under investigation to have been created within minutes of each other. They exclusively promoted UK-based PRC diplomats, ignoring those in other countries.

So human researchers were able to spot a pretty obvious pattern. How come Twitter's own staff and algorithms never seem to catch on, despite having even more data to look at?

It amazes me that Twitter fails at something so basic -- identifying and shutting down bot networks -- and has been failing at it for so long. Here's a report from 2017 (https://www.bbc.com/news/technology-38724082). Another one from 2014 (https://medium.com/i-data/fake-friends-with-real-benefits-ee...). If you look through the archives you will find similar reports going back 15 years.

Someone on HN (can't find the ref) once speculated that Twitter has an incentive not to shut down the fake accounts because the fakes boosts user numbers and "engagement" which correlate with profit. Twitter is certainly making a lot: I once looked into running a Twitter ad campaign for my business until I saw projected $2+ clicks. Are the massive margins suggested by those CPCs built on the backs of fake accounts?

Well yeah, what we see as obvious bots, twitter sees as MAUs which it can include in figures given to investors and shareholders.

Because they haven't developed any systems to identify the bots, they can pretend that they do not know they are bots - and their active user figures are not fraudulent/illegal. If they identified the bots, then they would not be able to lie about these accounts.

What is "MAUS"?

Monthly active users

If a man's salary depends on him not understanding something, you can bet that it won't be understood.

Also - I have seen lots of vigorous defense of the Chinese government here on HN lately. I wonder if there is a similar, but better run, effort here?

If they blocked chinas bots they would also block the follow, like, spam and click fraud bots. Removing 50% of twitters traffic would be hard to justify to higher ups, you could argue that the robots just produce noise and lower the quality of the product but higher ups just want to see the numbers get bigger.

I know I’ve had that thought here but I’m sure I’m not alone. The incentives are perverse, and moreover it explains why the Apple App Store and Amazon are populated with fake reviews and general chaff.

The fakers are parasitic to the buyer but symbiotic to the seller.

Twitter will not actively remove bots and fake accounts because without the new bots their user base has been stagnant for years and probably declining really fast.

that would expose twitter as a cesspit where the most mentally ill extroverts in society argue with bots and no actual conversation happens

This would remove %90 of botnet traffic: if ( from_amazon ) exit; You can get the list in json or csv below.

https://docs.aws.amazon.com/general/latest/gr/aws-ip-ranges....

I remember some years ago I would get messages from Twitter bots that all had an overly sexualized picture, and were named a womans name and then some random digits. It was so obvious that I don't understand why Twitter didn't stop it.

It's similar to the problem of security holes. Defenders have to find all bugs. Attackers have to find a single bug that the defenders missed. Twitter has to find all bots. These researchers have to find a single (cluster of) bot that Twitter missed. It's a lot easier.

From big co's understanding point of view, I think a more sensible option would be that Twitter doesn't look at it, or even cares, as long as it doesn't impact the bottom line.

I don't think Twitter allows it maliciously, as it tends to hurt their PR every time, and paints a target on their back for governments to attack them.

Wake me when they're meaningfully hurt and/or attacked. Until then, they will continue to do these things intentionally (which they do), whether the intent is malice or a political agenda.

Reply on news.ycombinator.com