How Twitter misleads us about how many people have left — and what to do about it

Social feeds are engines for distorting social understanding. Here’s how to push back against the algorithm.

J. Nathan Matias
5 min readJan 5, 2023
A Mountain Bluebird attacks its own reflection in a car mirror. Photo CC-By-2.0 Tom Koerner

Why are so many people staying on Twitter, even after the company gutted its child safety team, struggled with information security problems, violated user privacy for political ends, and brought nazis back to the platform?

To outside observers, it can seem like Twitter users are continuing as before, seemingly unaware of the millions of people who have left. “You left Twitter?” a friend recently remarked, “I hadn’t noticed.” Yet many of the accounts I follow haven’t tweeted in ages, and roughly 15% of them have already set up accounts on Mastodon. (I’m at

By filling feeds with a continuous stream of voices, platforms create the illusion that the room is full when many people have already left

How can so many people be unaware of a mass exodus from the platform? One reason is that social media feeds are designed to mislead us about the average opinions and behaviors of the people in our lives. The distorting effects of Twitter’s timeline and algorithm cause people to stay on the platform by making it seem like life continues as normal, despite the wasteland they occupy.

How can this happen? One answer lies in social norms. As a scientist, I study the relationship between algorithms and collective behavior— especially when social norms escalate online violence or prevent it. When people make decisions about how to behave, we think about the different groups that matter to us, guess how they will respond, and adjust our behaviors in response. Decades of research in social psychology have shown that while it’s hard to change someone’s mind, it’s much easier to influence behavior by changing people’s understanding of what’s common or expected.

As a business, social media is an engine for influencing behavior by shaping our beliefs about others. That’s one argument that Sinan Aral makes in his 2021 book The Hype Machine. When people see multiple friends voting in elections, taking on eating disorders, donating to charities, or joining a genocide, we start to believe that it’s common and expected. Whether or not the social feed has misled us about how common those behaviors are, we’re still more likely to follow suit.

Social media feeds are designed to mislead us about the opinions and behaviors of the people in our lives

Whatever the algorithm, the feed itself creates the simplest, most misleading distortion from social media. By giving us a continuous stream of voices, platforms create the illusion that the room is full when many people have already left. If you leave a lecture, the room will have a visible absence. Leave a social media platform, and the feed will fill the empty space with something else.

Overcoming Twitter’s Distorting Influence

How can we overcome the distorting influence of social media feeds to make it clear that more people are leaving Twitter? The answer can be found in a classic idea from social science: threshold models of collective behavior.

In “Threshold Models of Collective Behavior,” Mark Granovetter pointed out that people often need to see a certain number of other people take an action before they’re willing to do the same. That can be hard when people’s beliefs or views are unknown to each other. Consider this example I show to my undergraduate class at Cornell, with a network of blue smileys and orange smileys.

A graph showing blue and orange smiley faces. Orange smiley faces are interested in taking action, but they don’t know each other and they won’t take action without observing at least one or two others.
Illustrating the Threshold Model of Collective Action. Image by Drew Margolin

In this example, five orange smiley faces are all willing to take some kind of action. But they each have a threshold of others they need to see act first. When the network is controlled by an entity that prevents them from seeing each other, only one person does anything, and no one sees it. That’s the current situation with Twitter.

We typically think of social media platforms as systems that enable collective action by making like-minded people visible to each other, as pointed out by Jackson, Bailey and Welles in their book #HashtagActivism. The opposite can also happen — platforms also prevent collective action by making departures invisible.

When people try to leave, social media platforms are designed to instantly fill in the empty chairs with more voices from the algorithm. That’s also why Twitter has banned people who mention competing services, added more people you don’t follow to the feed, and show (unreliable) view metrics on every tweet in a desperate attempt to convince people to stay.

So how can we make our departures more visible to others and encourage the same? One powerful idea from Daniel Gillmor is to set up the equivalent of an “out of office” message on Twitter.

  • If you have left entirely, post a message about your departure once or twice a day.
  • Even if you are still posting to Twitter, it’s still worth pointing people to other places where you have set up accounts, so they can follow you elsewhere

If enough of us do this, our departures will be more visible, and more people will realize this and follow suit as their threshold for action is met.

How to Create an Out-of-Office Note on Twitter

In a discussion on Mastodon, Darius Kazemi made a great suggestion for how to quickly and easily set up an out-of-office note. Cheap Bots Done Quick is a service that will automatically tweet a custom message for you on a regular schedule.

  • Log into the site with your Twitter account
  • Customize the text
  • Tell it how often to post your out-of-office message

If you want to add variety to your message, Cheap Bots Done Quick allows you to create a template that will cycle through different combinations of phrases when tweeting the announcement. Feel free to copy and edit my version.

Achieving a Better Social Media Ecosystem

While I’ve decided to move to on Mastodon, I agree that all our digital environments are feeling polluted and exploited — just as our rivers must have seemed in the era before conservation. I understand why people have accepted living in a digital trashfire — and if they’re online at all — staying wherever the audience seems to be. That’s why it’s so important to make our departures visible, so people can see there are other options.

It’s going to take collective effort to imagine and achieve a better vision for social media. That’s our vision at the Citizens and Technology Lab, where we organize community/citizen science for a world where digital power is guided by evidence and accountable to the public.

Just like scientists and communities work together to understand our rivers, air, and biodiversity, CAT Lab works with communities to understand our digital environments. I hope this post helps people see one small way that science can help us achieve a better ecosystem, and start putting it into action.



J. Nathan Matias

Citizen social science to improve digital life & hold tech accountable. Assistant Prof, Cornell. Prev: Princeton, MIT. Guatemalan-American