How Bad Is Social Media?

[This is a transcript with references.]

What should you do if you’re spending too much time on social media? Asking for friend. Well, you try to convince yourself that social media is actually good for, something. It’s got to be good for something, right? But they say that social media increases polarization, and gets you stuck in echo chambers full of fake news and so on. How bad is social media? That’s what we’ll talk about today.

Social media has changed society profoundly. About 60 percent of the world’s population now uses social media. It has made it vastly easier to find people all over the globe, to connect with them, and to get insulted by them.

What does that do to society? It’s complicated. American social psychologist Jonathan Haidt and sociologist Chris Bail have compiled a public Google Doc in which they collect references on questions such as “Does social media make people angrier?” and “Does social media create political echo chambers?”. The most relevant thing you learn from this document is that whatever your opinion, you can find a paper that supports it.

Honestly, I began working on this video thinking it’ll end up being one big shrug because that’s how sociology generally looks like to a physicist. But turns out it isn’t quite as bad. You just have to be really careful with phrasing the question.

For example, you may remember the headlines claiming that fake news spreads faster than the truth. Then again there were headline saying that those headlines about fake news were themselves fake news. What is going on?

Well, the original headlines were based on a 2018 paper published in Science by researchers at MIT. The authors compared how true and false news stories spread on twitter.

They had a sample of about 126 thousand news items from 2006 to 2017, tweeted by about three million people over four point five million times. So, not a small study.

These new items were classified as true or false according to certain fact-checking organizations. The conclusion of the study was, in the authors’ own words, that “Falsehood diffused significantly farther, faster, deeper, and more broadly than the truth in all categories. The effects were most pronounced for false political news.” And it wasn’t a small difference. They found that it took true stories approximately six times as long as false stories to reach one thousand five hundred people.

But in 2021 other researchers pointed out that the 2018 paper looked at news that had been fact-checked by certain institutions, but that those institutions pay more attention to news that have already spread quite successfully. An article in Science then claimed that this means the original study had been debunked. This is why you’ve seen the headlines saying news about fake news are fake news.

That wasn’t the end of the story. Because the authors of the original study then said they’d never claimed their study applies to all fake news, it had just been misreported. And the authors of the new study said they had never claimed the earlier study was wrong because they knew it’d been misreported. Then the author of the science news article who had claimed that the misreported fake news study was fake news apologized that his article had misreported the story. Hope that clarifies it!

But wait what does all that mean now? Do fake news spread better or do they not? The answer is: they do, but. It turns out that the major difference between true news and fake news is that fake news spread to a larger audience. And since they appeal to a larger audience, they also spread faster. But if you compare true and false stories that have reached an audience of the same size, then the sharing patterns look the same. This was the point of the 2021 paper. It’s not like fake news networks have a different connectivity, the size of audience that they attract is the major difference.

And yes, that was strictly speaking only demonstrated for fake news stories that were fact checked by certain institutions. One of the authors made this diagram to show the difference between what they *said they did and what the headlines said they did. But the authors’ also say they’re reasonably confident their finding will carry over to false news more generally, but that remains to be seen. I am guessing there are people working on doing exactly this as we speak.

But the 2018 Science paper has made an interesting that didn’t spread widely. They found that bots accelerated the spread of true and false news equally. This means that if false news spreads better than the truth, then it’s because *humans are more likely to spread false news. We can’t blame it on the bots. The authors conjecture that the reason may be that people like novelty, and it’s easier to be original with something that’s made up.

Anecdotal evidence. I had my first encounter with fake news on facebook in 2016 when Trump ran for president. It was a quote attributed to Trump from some anti-republican facebook page, shared by an American friend. Several people pointed out that there was no evidence Trump actually said that. The guy who’d shared it reacted by saying it’s funny even if it isn’t true.

And that’s why false news spreads: We share it for reasons other than accuracy. Because it’s funny or upsetting or because it allow us to express our own opinion. Whether it’s true doesn’t really matter for that. The problem is that the next person who comes across shared fake news believes that the person who shared it believed it to be true and is therefroe more likely to also believe it’s true.

What can be done about it? It’s easier than you think! Because most people agree that fake news is bad and they’re actually quite good at spotting it, you just have to occasionally remind them to think before sharing. At least this was the conclusion of a paper published in Nature last year.

The authors recruited about a thousand Americans and presented them with 36 actual news stories taken from social media. Half of the headlines were false, and half were correct, half favourable to Democrats and the other half favourable to Republicans. The participants were then asked to evaluate the accuracy of the news items.

They quite reliably rated correct headlines as correct and false ones as false. And while they did rate headlines in favour of their own political orientation as correct more often than those in favour of the other camp, the partisan influence was much smaller than that of the actual accuracy of the headline. So, the problem is not that we’re just bad at spotting bad news.

But the authors also found that whether the headlines were right or wrong had little effect on whether people intended to share a news item. They then encouraged people to consider the accuracy of the news item and afterwards asked again how likely they’d share it. This simple tactic led to a big reduction of the intention to share *false news but didn’t affect the intention to share real news. According to the paper, accuracy “often has little effect on sharing, because the social media context focuses [users’] attention on other factors such as the desire to attract and please followers/friends or to signal one’s group membership”.

According to another paper that just appeared two months ago, the misinformation problem is particularly pronounced in the United States. The authors of the paper found that while people from the UK, Canada, Australia, and New Zealand are exposed to misinformation on social media at about the same rate, Americans are three times more likely to share it.

Earlier this year, a review paper in the journal Nature Medicine looked at the spread of misinformation about public health in particular by reviewing 123 papers. The two major conclusions that the author draws from his literature survey is that (a) people are sometimes duped by misinformation just because they are distracted or not paying attention and (b) some people believe in and share misinformation because it reinforces their beliefs. So that’s consistent with what the other papers had found.

As to what to do about it, one thing he suggests is also just reminding people to think about accuracy before sharing. Another interesting suggestion he has is to give people information about the tactics of misinformation spreaders with browser games. There are two of those games, one is called “Go Viral!” and the other one “Get Bad News”. Studies have found that people who played these games were much better at spotting health-related misinformation. You can try them out yourself, links are in the info below. I think this is a really good idea and I’d like to have a game like this about physics please.

So, yes, social media spreads a lot of misinformation. The good side of social media is that it also seems to generally benefit information literacy. In 2018, a team of American researchers recruited almost 3000 Americans. They offered half of them 20 dollars to deactivate their facebook accounts for 4 weeks, just prior to the 2018 midterm elections. Four weeks later, those who disconnected from facebook were less able to correctly answer factual questions about recent news events. But they also reported increased well-being and less political polarization. Let’s therefore look at what we know about polarization and echo chambers.

If you watch a few of my videos on quantum mechanics, soon all your recommended videos will be about quantum mechanics. Suddenly the whole world is quantum mechanics! Such an “echo chamber” seems to be an inevitable side-effect of algorithms that want to help you find what you’re interested in, and as a result, you get more of the same.

This leads to the dreaded conspiracy rabbit holes that you fall into on YouTube, as it happened to me when I was working on my video on flat earthers, though YouTube seems to have tweaked their algorithm since to prevent that from happening quite as easily.

These “more of the same” algorithms help you make contact with people who think like yourself. So, the idea that we live in echo chambers sounds plausible, but plausible ideas are the ones you should be most sceptical about. What do the data say?

According to a 2021 paper by a group from the University of Oxford, echo chamber issue are real but the problem’s been hugely overstated. They looked at surveys from 7 different countries in which people reported what news they typically consumed. Turns out only about 5 percent of social media user are properly stuck in a political echo chamber in which they almost exclusively consume news from one political side. Though the numbers differ somewhat by country. The overall largest fraction of people in echo chambers is that of the American left.

The previously mentioned Chris Bail is lead author of a 2018 paper about an experiment in which they tried to get people out of their echo chambers. They surveyed about one thousand five hundred Americans, about half Democrats and the other half Republicans who visited Twitter at least three times each week. After one week of tracking, a randomly selected group of those people was offered 11 dollars to follow a Twitter bot for 1 month, but they were not informed about the purpose of this study. The bot initially just tweeted landscape pictures but then began tweeting that promoted the participants’ opposing political ideology.

At the end of the month, the participants were surveyed again. Turned out that republicans who followed a liberal Twitter bot became even more conservative. The more they were exposed the larger the effect. For democrats the change was not statistically significant.

By the way, Chris Bail is the director of an institute called the Polarization Lab that lets you check how deeply stuck in an echo chamber you are on twitter. Turns out rather unsurprisingly I’m deeply stuck in a liberal camp, that’s what you get when you mostly follow people with PhDs.

The question whether social media increases polarization in society has been extensively studied, especially in the United States. The risk, sociologists say, is that social media makes it easier to find people whose opinions we like, and we get encouraged by like-minded people to distance ourselves from the perceived enemy.  

Indeed, in 2018, a leaked internal presentation at Facebook warned senior executives that facebook “algorithms exploit the human brain’s attraction to divisiveness,” and that “If left unchecked,” the algorithm would feed users “more and more divisive content in an effort to gain user attention & increase time on the platform.”

Now, it’s quite well-established among sociologists already that increased levels of polarization in society are associated with erosion of constructive political debate, social trust, and inter-party cooperation. The question we’re interested in here is whether social media *increases this polarization.

In 2021, researchers from the UK and the US set out to answer the question whether “Out-group animosity drives engagement on social media”. They analysed almost three million Twitter posts by news media accounts and US congressional members. They found that posts about the political out-group were shared or retweeted about twice as often as posts about the in-group. And almost all of the post about the other political camp were negative, leading to more negative engagement.

Again though, you have to be really careful to keep in mind what question a study was asking in the first place. Because this finding doesn’t necessarily mean that the negative engagement caused polarization to *increase, it might just means that social media is a good platform to live out your feelings. Just which way the causation goes is at the moment rather unclear.

For one thing, a study from 2017 , found that self-reported polarization is higher among elderly Americans who are less likely to be online to begin with.

Another thing to keep in mind is that the USA isn’t the *only country in the world. A team of American researchers pointed out last yearthat while social media usage has increased world-wide, polarization has not. It *has increased in the US, but in many other countries polarization has in fact decreased.

Indeed, a 2021 study among more than 3000 Dutch citizens found that self-reported polarization *correlates with social media use, but the causation goes from polarization to social media usage, and it depends on the platform you’re using. At least in the Dutch sample, people who became politically more polarized spent more time on facebook, but less time on twitter.

It’s hard to interpret what this means because god knows what’s up with Dutch twitter. But I think what we can take away from this is that the idea that more social media use increases polarization is almost certainly too simple to be correct. What happens depends both on cultural context and on platform design.

So what do we learn from all this?

Most obviously we learn that this is a very active research area. I certainly hope that the results will eventually lead to better algorithms for social media. While we wait for that to happen, I think the best we can do is focus on the part that’s in our hands, which is to decide what we pay attention to, and what we share.

What I have taken away from all those papers is that we have to be especially careful with headlines that upset us. Yes, they might get a lot of comments. But they might also spread misinformation and hate. So be careful out there.