Comment by AlphaWeaver | Hacker News Reader

Comment by AlphaWeaver | Hacker News Reader

Parent Full thread

Something particularly concerning about this incident is buried further down in the article.

Under the heading "Why Facebook Doesn't Ban Everyone", Lincoln explains how social network theory can identify individuals in networks that are influential. This isn't an exaggeration - basic graph heuristics (such as comparing in/out degree of an individual relative to other parts of the social graph) can easily identify such individuals. Censoring those individuals can have an amplified effect on the graph structure as a whole.

For an example of how just a few individuals can make a big difference on graph structure, check out Watts & Strogatz's 1998 paper, Collective dynamics of 'small-world' networks [0][1]. I'd highly recommend reading it - it's very short and easy to understand.

Essentially, a small number of connectors connecting with random individuals can seriously reduce the average number of steps to connect any two individuals. If Facebook wanted to stop the spread of a narrative that they didn't like, censoring just these few individuals would make it much harder for information to flow through the graph as a whole, and hardly anyone would notice (since Facebook's only censoring a few people.)

Scary.

[0]: https://en.wikipedia.org/wiki/Watts–Strogatz_model

[1]: https://pubmed.ncbi.nlm.nih.gov/9623998/

I’m banned from Twitter without explanation. I made an account, followed 3-4 tech celebrities like Paul graham, and posted a single tweet similar to “I finally got pressured into making a dumb Twitter account...” The account got banned for community violations within an hour and my support email said that my requests for support would not be responded to. Years later, I’m working on a video game and I’m worried this will make it hard to do marketing coming up in case my ip got flagged along the way.

Other good suggestions here, but here’s another: you tripped the New Bot Account detector.

As your account ages there will be more information about whether you are real or not. Stopping the influx of bots before they do damage is important. Therefore it’s quite likely that Twitter ban accounts early if they look like bots based on not much info (and potentially if they look low value).

You followed a few people who are popular account (not friends, fairly anonymous) but not mainstream accounts (so you’re a little more atypical) and then you post once with, sorry to say it, not particularly inspired/original/specific/human content. Perhaps there were other factors. I’d guess you didn’t go through and fill out your profile much if you weren’t that invested in the idea? No birthday or avatar maybe? Not following the easy setup steps that Twitter encourage you to do?

Combine this looking a little like a bot with the likelihood that you’re going to be low value because you’re not that engaged (debatable I know), and it kinda makes sense.

All speculation, but a possibility. Should Twitter be doing this? I don’t think it’s the worst thing, but it’s a trade off. If this is it then it would imply that your marketing efforts for your game would probably not be affected at all.

That is one explanation, but the explanation I tend to believe is another: Twitter quickly bans new accounts to force you to "prove your identity" by adding a phone number.

It is a very good way to confirm that you're in fact a person though.

Obviously, but it doesn’t help me when support has a “we will never speak with you” policy

You’ll probably find that these kind of systems don’t have just a single ‘New Bot Account’ detector, but a tangle of multiple subsystems that automatically temp-lock accounts based on various heuristics written by different people at different times.

A new account might incorrectly trigger an account lock due to signals for proxy detection, another lock for a specific 2017 bot wave with Romanian IP addresses and a third generic fake account lock. That, in turn, triggers a recidivism threshold that sets off a permanent disable.

Working ‘at scale’ means that developers who unravel old code to fix a use-blocking problem for 5000 users can be considered astronomically less productive than someone who fixes a minor problem for 100,000+.

> Stopping the influx of bots before they do damage is important.

Why?

Likely because if bad actors add bots and gain followers (across all bot accounts) faster than Twitter kills off the bots, the ratio of bots to real people will eventually (on a long enough timeline) become infinite.

When people notice they're only seeing, hearing from, and seeing retweets from bots, they're likely to leave Twitter. That would be bad for Twitter.

Twitter requires a phone number these days or you can't post. Twitter also finds ways to leak your phone number to others so I will never give them that.

Your account was probably hacked and used for something that broke the rules. I got an email warning of someone trying to change my password within a week of making an account.

Hogwash.

Twitter would be a lot more credible (although anything times zero is still zero) if they didn't have the checkbox "Let people who have your phone number find and connect with you on Twitter" on their "security" lockout verification page.

And, somehow, Twitter keeps spamming my email address in spite of my account having been "locked".

Sorry ... not sorry. No phone number for you.

The account wasn’t hacked. It was a day old

Changing your IP address should resolve your concerns?

Of course you can get around their "security" system, but you should ask yourself whether it is a game worth playing especially since the rules of the game can change overnight. Getting tied to a platform is a dangerous bet.

It's absolutely terrifying to me that they can do this sort of thing, or selectively censor DMs containing certain links, or from certain people or geofenced places, and it would never make the news.

I wrote a few months ago about this sort of danger: https://sneak.berlin/20200421/normalcy-bias/

These sorts of communication systems work, until the moment they don't, and we won't receive advance notice.

It seems that influence of a link is comprised of their awareness of the cliques discourse and reputation among the clique members. The algorithm on the other hand seems to latch on to traffic. There is a room for a wedge here.

Perhaps the algorithm can be defeated by employing a few extra "link" accounts (either sock puppets or simply ardent followers) with a shared reputation (a shared brand).

On other words a link between two cliques could be held by a special kind of clique.

Your concern seems to ignore that humans are not static agents.

They’ll route around black holes.

Conversation will not simply stop. It will move to a different platform or, should the network of platforms using HTTP look black holes, where useful into goes in but little useful info comes out, will pivot to a new protocol.

The really important infrastructure is the hardware and basic protocols.

There’s no reason see the network “inside Twitter” as important. Highly visible, but it’s just one groups choice of what to store in a database.

Alternatives can be copy-pasted together in hours by competent groups.

Tech is so up it’s own ass these days because it’s measuring inside a few Petri dishes and ignoring the Internet as a concept in and of itself.

The internet is not FAANG and Silicon Valley. It’s the hardware everywhere and fundamental protocols.

So, yes, the math is right. But the math applies at various scales. For as visible as Twitter and the rest are, they’re not in and of themselves a distributed global network of hardware and routing protocols. They’re black holes in a bigger universe.

That's interesting. The disgorzation produced by all the bans will probably impact votes for conservatives for a long time despite not censoring that many people indeed.

Maybe. The outcome of banning "connectors" typically reduces connectedness between cliques, rather than within cliques themselves. (In graph theory, when a group of people are well connected that share similar interests, we typically call that a clique.)

When people complain that people are increasingly finding themselves in echo chambers, and that "people don't talk to people from the other side," connectedness across the graph is dropping, while connectedness within cliques remains strong or increases.

Reply on news.ycombinator.com