The Connection Machine

Freedom of speech in the USA is a pretty well-misunderstood concept. The basic idea is that you should be able to say whatever you want and you won’t get in trouble. But of course, that’s not anywhere close to true. Mostly because it only applies to the government. Your employer can fire you for what you say. So you have freedom of speech, but only when it comes to what the government can do about your ranting. (But there are lots of limits on that freedom, even when it comes to the government. You can’t threaten people, for example. Calling in a bomb threat doesn’t get you a pass just because of the First Amendment.)

So you have to be a little careful what you say, because it can get you fired. And wrapping yourself in the First Amendment doesn’t help. But what if your employer is the government? Well, you get a little more leeway in that case. Particularly if you are doing the speaking on your own time. But still, there are limits. Except you will never be able to figure out what the limits are, because the courts have been extremely inconsistent over the years.

So suppose you are a pretty big deal tweeter who has built a following of thousands by being really funny. But funny in that “your comedy special can’t be on before 10PM” way. Lots of jokes about drinking and fucking and swearing and stuff. And suppose you are also a public school teacher. And suppose you don’t really do a very good job at staying anonymous. Say you put up selfies now and then that actually show your face. And you have an AVI that looks like you. But you go out of your way to avoid ever connecting to anyone you know in real life. What do you suppose is going to happen?

Well, it turns out that what’s going to happen is a parent is going to discover your account. And they are going to report you. And you will think you’re going to get fired. But then you get to the meeting and it turns out it’s an intervention. And they start reading your tweets out loud. And you start laughing because, let’s face it, you’re the funniest fucker you know. But they aren’t laughing. They think you are an unstable alcoholic who whores around and maybe even has a couple murders under your belt.

So that’s a fun way to spend a Thursday.

But let’s roll back a bit. How did that parent discover your account? You have location stuff turned off. You aren’t using your real name, or even your real city. There are hundreds of millions of users. Isn’t the probability of a parent stumbling onto you really, really, really small?

Nope. First, there are hundreds of millions of twitter users, but only 34 million users in the USA. And the number of parents in a big school district is, say, around 34,000. So given that about 1 in 10 US citizens is on twitter (really? that seems high), about 3,400 of those parents are on twitter. If each of them sees just one tweet, the chance of you being seen is one one-hundredth of one percent. But they see tweets from lots of people. If each of them saw tweets from 100 different people, and each of them saw a different 100, that brings the probability of you getting seen up to just 1%. One in a hundred. But remember, you’re a big deal funny fucker and people retweet you all the time. So it’s now starting to look like the chances of being detected are actually pretty good.

But, in fact, that analysis greatly understates the reality. Because behind the scenes, there is actually a twitter machine that is trying to make that parent find you. Let’s call it the connection machine. And what that machine does is try to figure out who knows whom.

In general population twitter, people actually think this is a good thing. They import their list of contacts to see if they can find and follow people they know in real life. And from there, the connection machine makes recommendations and tries to help them find even more people they know. Because to the general population, Twitter is just another Facebook.

But if you have an anonymous account, location features turned off, no hashtags about local teams or events, or anything obvious, how can twitter make the connections to people around you? Well, there are lots of ways, actually. I don’t know which of these twitter actually uses, but I’d guess all of them.

Location, location, location! When you turn off location services, that prevents those stalker-enabling tags on all your tweets, and it keeps your phone from telling twitter exactly which bathroom you are in, but it doesn’t keep twitter from knowing your location. The messages between your device and twitter have an internet address, which is just a big number that tells the network how to get messages to you. That network is a collection of physical things in physical places, so there are big tables of addresses that let the equipment get the bits where they need to go. So although your internet address doesn’t locate you like a GPS does, it certainly reveals your city, and if you are tweeting at work on company Wi-Fi, probably your employer’s address.

So twitter knows where you are. And that’s certainly something it can use to match people up.

You probably also connect from a variety of places. So it can even make stronger connections between people who, for example, go to the same gym or wait in line at the same Starbucks.

It’s not what you know. It’s who you know. The next obvious place to look for connections is in the follower graph. Twitter clearly leans on this database heavily for recommendations. It often tells me I should be following people who follow me. It also tells me to follow people that a lot of the people I follow follow. (Yeah, I just said follow follow. I’m leaving it.)

Spooky stuff. How about other, more subtle connections? For example, I bet most of general population twitter follow a lot more blue-check-verified accounts than people in funny, sexy, weird, or poetic twitter. So you could cleave the population into groups who like verified accounts, and those who don’t.

You can cleave the population into those who use lots of hashtags, and those who think hashtags are so last year.

People who @ celebrities, and those who are not idiots.

People who RT news stories, and those who are not annoying.

People who live-tweet sporting events, and those who mute them.

Build up enough of these discriminators and you will create a fingerprint. And that fingerprint is probably what drives Twitter’s “similar to” recommendations.

The ways we are all connected are sometimes so subtle and obscure that you can’t even explain them. But a computer can see them. When you read the hype about “big data,” particularly in marketing, this is what they are talking about. Algorithms that let the connection machine match people who are otherwise not obviously connected. For example, not too long ago, Facebook started suggesting twitter people to me. My Facebook life and my twitter life could not be more separate. There are not any connections between these accounts, and only a couple of people from twitter have become close enough friends to justify subjecting them to the banality of my Facebook account. Yet Facebook was able to focus in on these people find connections between them, and boom, all sorts of twitter people started popping up in my “People you may Know.” Big data = spoooooky data.

So my friend, the public school teacher, has shut down her old account and started over. And she is being more careful. Keeping it anonymous. But I told her—there is no way you are going to avoid having yet another parent find the account. It’s virtually guaranteed to happen eventually. What you can do is make it impossible to prove that is your account. And that will have to be good enough. Because freedom of speech only goes so far. And for teachers, that isn’t very far at all.