Communicating on social media is weird — why?

My aim in this post is to present an explanation of some features of social media in terms of linguistics, game theory, and economics. I’m interested in two questions:

  • Why do we talk about social/political/economic stuff with strangers on social media?
  • Why do we speak to, and read the words of, many people? Why do we typically follow and be followed by many people, in light of the fact that we don’t interact with most of them?

The first thing to note is that these features of social media are weird. It’s easy to overlook this fact because it’s so ubiquitous, but it’s strange that many of us spend a lot of our time talking about current affairs with people we’re not close to. After all, it’s a familiar piece of advice that with people you’re close to — say, around the thanksgiving table — politics and so on shouldn’t be discussed. And there’s good reason for this — on these contentious topics, it’s so easy to offend or get embroiled in arguments, a risk that’s even greater when talking with people you don’t know well.

It’s also strange that we speak to and hear from many, while ignoring most. I follow and am followed by around 800 people on twitter, but most of my tweets only get a handful of interactions, and I only interact with a handful of tweets I see. Most tweets are ignored by most readers.

Social media communication is not continuous with every day communication — it’s not a translation onto the internet of face-to-face conversation. So it’s worth asking why it has the features that it does have.

Let me introduce two concepts, from game theory and linguistics respectively, that I think explain the strange fact that we talk politics with people we’re not close to.

The first is that of a Schelling point. A Schelling point is a type of solution to a group action problem when the group must coordinate on some one response but can’t communicate about it. For example, say that you and a stranger are posed this problem — meet each other tomorrow, somewhere in New York City, at noon. If you could communicate, this would be no problem — you’d just agree on somewhere, it doesn’t matter where, to meet. But what if you can’t communicate? What should you do then?

Schelling’s answer is that you should each pick the place that you expect the other person to pick. But what is that? You know nothing about the person — you don’t know, for example, that they like big statues are so are likely to opt for the statue of liberty. So what you do is pick somewhere that seems like an obvious answer to the question. In this particular case, Schelling’s answer was that you should pick the information booth at Grand Central Terminal, since that’s a very obvious and salient place to meet in New York.

The key point is that in the absence of communication, if you need to coordinate on an action, you should pick the action that is in some sense most obvious or salient.

The second idea I will use is that of presupposition. Philosophers and linguists think of conversation as taking place against a background of shared beliefs that are presupposed by all the members of a conversation. These are things you can take for granted in the conversation. For example, if you’re talking to a good friend of yours, there will be lots of presuppositions: that you have a lizard, say, or that you used to live in Oregon. You can exploit these presuppositions to communicate more succinctly. You can say, for example, if you’re talking about strange road trips: “When I left Oregon, I drove across the country with my lizard”. If you’re talking to a stranger about the same thing, you can’t just say that — they’d look at you funny, and maybe say “I didn’t know you lived in Oregon!” or “I didn’t know you had a lizard!” Instead, you have to be more longwinded, and say something like “I used to live in Oregon but then I left and I have a cat and I drove across the country with it”.

Let’s consider how these concepts might apply in social media. Assume that social media posts are, in some sense, conversations. Given that they are often with(more or less) strangers, then there is no set of presuppositions. You don’t know what the other person knows, and so you don’t know what presuppositions you can exploit for quick and smooth communication.

(Contrast this with face-to-face conversation: there will be lots of presuppositions based solely on the shared physical environment that can be exploited. This perhaps explains why weather is a perennial topic of small talk in such circumstances.)

On social media, given the constraints of characters (as on twitter) and attention (tl;dr), it’s very important to have presuppositions. So you have a question: what can I presuppose as common ground with all my many followers, most of whom I don’t know well?

My claim is that social/political/economic events are a Schelling point solution to that question. You can pretty much take for granted that a person will have some interest in and knowledge of, say, Trump or Brexit or the latest climate change report or the viral Buzzfeed longread. These things are to conversation with strangers what Grand Central Terminal is to people meeting in New York, and so we talk about such things with strangers, despite the generally received wisdom that they are controversial, because we can presuppose common ground about it.

(Of course, not all social media conversation is about these topics, and much is more special interest. Most of my twitter interactions are about academic philosophy. But they work precisely because I and other academic philosophers have a large number of presuppositions in common.)

So that’s my explanation of why we talk about current events. You might object: isn’t this a bit complicated? Couldn’t the reason we talk about these things simply be … because we want to?

Well, I’m not sure. Here’s a way to see that this simple answer is perhaps too simple. The internet contains all human knowledge. In theory, at least, all of that is potentially discussable on social media. Once you realize this — that we could speak about ever so many different things — the fact that we do tend mostly to speak only about a very small subset of all the things becomes more puzzling.

Or rather, it’s puzzling before you consider the explanation I offered. The internet may contain all human knowledge, but not all human knowledge is discussable with people you don’t know well online, because it can’t be presupposed that a stranger will have any familiarity with it, while it can be presupposed a stranger will have familiarity with, say, Trump.

On the perspective I’ve adopted here, it’s not the case that we’re all news junkies who take to social media to talk about it. Rather, we’re all talking junkies who talk about the news because it’s something we can rely on others to know about. Communicating is the end; politics is a means to that end.

That means that there’s nothing inherent in social/political/economic topics that make them the fitting object of social media attention, just as there’s nothing inherent in Grand Central Terminal that makes it the best place to meet in New York. And that means, in turn, that we could easily have social media that didn’t work that way, which might be something we should consider trying to implement if we think that talking about such things does more harm than good (thanks to echo chambers, doxxing, bots, Trump, twitter mobs, etc.)

Grand Central Station is a Schelling point solution for meeting in New York. (Image credit:

Let’s turn to the next question of why we speak to so many people, and hear so many people, but interact with so few. I think this, in turn, can be explained by looking at the logic of mass production.

To see this, note first that the content of twitter is user-produced tweets. That is where the value of twitter lies, that is what gets us logging in and, all importantly, looking at ads.

Realizing this, the following facts cause a prima facie problem:

(1) Twitter doesn’t want to pay money for content

(2) Twitter users don’t want to expend much effort tweeting

(3) Twitter users want as much attention as possible
(4) Twitter needs good quality content

(1) is a fact, and (2) plausibly follows from (1) (generally, we’re reticent to spend too much time working for free), yet (2) is in tension with both (3)(we could get more attention if we expended more effort) and (4) (twitter’s content would be better if users expended more effort tweeting). I claim that what dissolves the tension here is the fact that we speak to and are members of large audiences.

Think of two ways we can sell products. You can find a small target audience, tailor the product to them, and reach out to them personally to try to get them to buy it. Or, you can instead make a generic product, and bring it to the attention of many, many people, in the hope that it will be suitable for at least some of them. In the former case, arguably, more effort goes into making the product — it is better quality — while in the latter, more effort goes into presenting it to as many people as possible.

Lower quality products can be produced and consumed provided each gets presented to many people, and there are many such products. If they are presented to many people, then they’ll be suitable for at least some of them, and so the makers will make some sales. If there are many such products, then users will be able to find one that suits them, and so users will be happy.

Let’s transfer this to social media. Users don’t want to expend much effort tweeting, because they do it for free. That means that their tweets are, on average, low quality. But because they’re presented to many people, there’s a chance that at least someone will appreciate any given tweet, and so the tweeter, as writer, gets attention (3) despite having put in little effort (2), and so they are happy. When we tweet, we are like people making a non-bespoke pair of jeans, which, although it won’t fit most people, will be on a rack in a superstore that gets enough footfall that it will catch the eye and fit the waist of at least a few browsers. That is why we have many followers.

Similarly, a typical twitter feed counts as good quality for a tweeter as a reader (4) because, even though most individual tweets are not good quality, there is so many of them that we’ll probably find at least a couple interesting. And that is why we follow many people.

By thinking, then, of tweets as obeying the logic of mass-produced products we can explain the strange fact that we speak and hear from many but interact with so few. But we can also realize how contingent this strange fact is, and how different social media could be if, for example, the number of users we follow(ed) was throttled. That might encourage us to produce higher quality content and lead to more interactions.

Social media, although so familiar, should seem stranger to us than it does. I have attempted to explain some of its strange features in the hope that, once we recognize how contingent they are — how easily they could have been different — we might attempt to create better social media sites, and better ways of communicating on the internet.

Novella "Coming From Nothing" at @zer0books ( Academic philosophy at: