Hacker News Reader

Comment by petargyurov | Hacker News Reader

The biggest sign mentioned in the article is the insane number of tweets since the inception of the account - surely this must set off some alerts on Twitter's backend?

I know that this is an extremely difficult problem to solve and that they probably ban accounts in waves but with something as obvious as this you'd think that the ban could/should come more swiftly...

View on HN

IIRC someone did some analysis on political accounts that did nothing but high volume retweets supportive of a particular cause. Problem is, quite a few of the identified probable bot accounts turned out to be real people that had just become infatuated with the cause...

[Related phenomenon: the UK PM gets lots of obviously bot -like almost identical 'good job Boris' comments on his Facebook account. Trouble is, it appears that it's actually mostly real people with otherwise perfectly normal accounts and verifiable identies: the whole thing probably started off as a forum campaign to rally the appearance of support in an election buildup, but ended up being mostly people who think it's a funny meme. https://www.bbc.co.uk/news/blogs-trending-50218615]

> Trouble is, it appears that it's actually mostly real people with otherwise perfectly normal accounts and verifiable identies

Just hypothetical: how do they come up with ground truth when studying this kind of thing? Seems difficult unless you witness the person interacting in the real world.

The BBC actually exchanged messages with some of them. People on Facebook in particular use real names and have profiles on other websites, including places where they and their family work and can be contacted using real contact details. And there's some extent of life history faking that bot creators aren't going to go to just to be one of dozens of people posting 'Good job Boris' a decade after the account's first formed.

Obviously it's more difficult on Twitter when accounts tend to be anonymous, connections are one-way, a lot of activity is retweeting and very little of it is personal photos, and many real accounts have been around for a lot less time.

From what I’ve understood, the wave ban is partly because they don’t want the spammers to be able to figure out exactly what triggered the ban.

Given the pressure that Elliott is putting on Dorsey to aggressively grow Twitter, I don't think it is likely that things will improve.

https://www.vox.com/recode/2020/3/9/21171482/twitter-jack-do...

They got 3 seats on the board and agreed to create a committee literally to search for Dorsey's successor. It's far too late for Dorsey. Realistically they're just letting him go out on his own terms at this point. In a few months Dorsey will announce that he's stepping down because he wants to focus his attention on his other job and that'll be that.

Do you have a source for the claim that this committee is searching for Dorsey's successor? I found this from the WSJ about three days ago, but the gist of it seems to be that two new board members were added, they're searching for a 'new independent director' and a deal was struck that allowed Dorsey to maintain his position.

https://www.wsj.com/articles/twitter-elliott-strike-truce-th...

https://uk.reuters.com/article/us-twitter-elliott/twitter-el...

>The board will also create a temporary committee to evaluate Twitter’s leadership structure and CEO succession plan that will share the results publicly before the end of the year.

I expect so, but I don't expect Elliott to scale back its expectations significantly after that happens.

I think it could be a relatively easy problem to solve: give any user the option to verify their identity.

In this way, as verified accounts increase, the unverified accounts become more obvious and I imagine we will start to view them with more scrutiny.

It would allow anonymity to remain, just make it so those accounts are more obviously anonymous or pseudonymous.

What do you think the marginal cost of verification is, and how many users will Twitter have to verify in order to make this work?

The marginal cost of verification could be zero. All Twitter has to do is to get together with Facebook, Google et al and agree to support a number of approved independent identity systems such as Yoti [1]. That uses a combination of government documents (eg passports) and biometrics.

Basically, users would be able to sign up for a sort of "digital passport" ID that would, in the long term, be accepted by governments and most websites. Enough people might find it convenient enough to sign up.

Twitter already allows you to block all users who have not registered a phone number and not uploaded a photo or whatever. I used these settings. They actually do remove a lot of bots and bad players. Again, in the long term, Twitter would give people the option to see only verified accounts.

I really wouldn't care if it meant I only saw tweets from a relatively small number of verified accounts. It might be more like Twitter in its first half dozen years, when it was much more engaging and fun than it is today.

[1] https://www.yoti.com/

Yoti also works offline. For example, you can use it to prove that you are over 18 without giving away your name or other irrelevant information.

Verify how

Not exactly sure which method they'd use but seems like if they wanted to do it, they could put a lot more effort into figuring it out. I mean, Bumble, a dating app, does a simple (maybe too-easy-to-game) verification, but with so many banks having to abide by know your customer (KYC) requirements, then I think social networks could figure it out.

You should try signing up on Nextdoor. This isn't a hard problem, nor would anyone be forced to verify themselves if they don't want to.

I don't know if Nextdoor's method of mailing a card to someone's mailbox would work for Twitter, but I like your example of a social network that has taken verification much more seriously.

Nextdoor doesn’t think I’m using my real name and told me to contact support instead. Yeah F that...

Reply on news.ycombinator.com