How social media drives a “Regression to the extremes.”

Our digitally intermediated conversations keep sorting us into extremes — and driving our polarization.

Last week, we commemorated the somber day one year ago when thousands of Americans tried to ransack our democracy. That came on the backdrop of a global pandemic’s fifth wave, despite now having a safe, effective vaccine that nearly 40% of Americans still refuse to take.  

Meanwhile, America is just, well, angrier. 

Videos of airline passengers assaulting flight staff, shoppers harassing other shoppers over masks, or road rage incidents populate our newsfeeds. It doesn’t just offend our sensibilities. It’s deadly — traffic accident fatalities have spiked over the past two years.

Though each of these offers substrate for a complex sociological analysis, there is a clear thread that binds them. I’ve written quite a bit in these pages about the way that social media has corrupted our public discourse. But here I want to hone in on a particular mechanism, which I’ll call “regression to the extremes.”

In statistics, “regression to the mean” is the phenomenon by which if one sample of a certain variable tends toward the extreme, the next sample is more likely to to tend to the mean average of that variable. The regression to the extreme that I’m describing here is the phenomenon by which engaging in social media discussions will tend to drive opinions to one of either extremes on an issue. 

Let me use an example to demonstrate how this happens. As you all know, I deeply believe in Medicare for All — to the point where I literally co-wrote the book on it. When I post on social media about Medicare for All, I will inevitably get a response to the effect of, “While I agree that Medicare for All is a worthy policy goal, it is politically untenable and therefore we should support alternative approaches to reform.” I obviously disagree with this position. Though achieving Medicare for All will certainly be politically challenging, it is the most important healthcare fight we ought to be invested in (here’s a reminder of why). To get there, we will have to uproot the absurd power of healthcare corporations who sit on the wrong side of achieving it. But on this post, someone who clearly supports Medicare for All will respond to the quote above with something far more extreme, something like this: “You obviously don’t care about people and are willing to let them die in a ditch. And because Black people are far less likely to have healthcare in America, you’re clearly racist and support genocide against Black people.” 

What should be clear at this point is that the person upon whom this invective has been heaped probably supports the cause of more healthcare, and probably cares about healthcare inequities. On the grand continuum of healthcare policy, they’re closer to Bernie Sanders than Donald Trump. A rational, reasonable approach could probably even sway them to see the argument for Medicare for All as a politically tenable policy goal. But now, what little space for common ground we had has been obliterated. Their main incentive is to defend themselves against charges that they are a racist, genocidal maniac hellbent on taking everyone’s healthcare. So instead of reiterating their disagreement, they now argue something like this: “Medicare for All is going to lose us the next election — so even though you pretend to be a progressive, you actually support Trump and Trumpism!” And this is how regression to the extremes works: it takes a potentially thoughtful conversation about Medicare for All and turns it into mutual accusations about racist genocide and Trumpism. 

I have read too many unproductive threads on too many platforms that look something like this. You probably have, too. It’s not because people really believe these things about each other, it’s because the incentives of the algorithm are designed to bring them out. How? Social media algorithms are all about trying to keep your eyes fixed — that’s how they sell you ads, after all. So they promote the material with the most engagement. That tends to be the most inflammatory material. This was one of the revelations of the Facebook Papers. This, in turn, drives even more engagement. It raises the stakes on the conversation by increasing the audience. 

But there’s more. Neither of these people know each other in real life. They have no stake in maintaining a relationship. Rather, the entire interaction is intermediated by the social media platform, where the incentives drive virtue signaling to a base rather than achieving any sort of mutual understanding.

What does this have to do with airline fights, road rage, an insurrection, all of the absurdity we’re seeing offline? 

Well, after two years of pandemic life, when so many of our interactions have been digitally intermediated, it has reprogrammed our social incentives. We are porting the incentives of social media into the real world, with disastrous consequences. We assume the worst possible things about each other and act accordingly. The stress of living in a seemingly hostile world leaves us cynical, pessimistic, and closed off. That drives us back into our online echo chambers. Rinse and repeat.

It also has serious implications for the real challenges we are facing as a society. Our democracy, action against the pandemic, climate change, or the rise of global autocracy — all of these require us to find and occupy the common ground we share, to act collectively in the face of challenges grander than any one individual’s capacity for problem-solving. 

Make no mistake, the internet and the social media that operationalizes it aren’t going away anytime soon. While we need meaningful policy action to regulate the deleterious effects it's having, that doesn’t look to be coming in the near future. So in the interim, it’s worth trying to minimize our role in the regression to the extremes and the impact it has on our own outlook. 

Here’s a few simple guidelines to which I’m working to hold myself (and often failing). 

1. If you wouldn’t say it, don’t type it. Part of the mechanism of the regression to the extremes is that the digitally intermediated interface allows us some anonymity. It encourages us to type out things we probably wouldn’t say in person. 

2. Build meaningful online spaces.  Not everything on the internet is the cesspool that has become Facebook or Twitter. But usually, the spaces that are more earnest are cultivated to be that way … places like The Incision and others where people are invested in the search for truth together.

3. Cultivate offline thought spaces. One of the more insidious consequences of social media is that it has displaced our more traditional means of conversation. So be purposive about building irl spaces for just that.

4. Be proactive, not reactive. While sharing your perspectives online can be fulfilling and lead to meaningful conversations, responding to comments usually isn’t. If you’re going to post, post first instead of last. 

5. Assume the best in others. We have a lot more ownership over our interactions in shared public spaces than we often assume. Try to assume the best in others, and excuse them when they fail to live up to it. It’ll open to you to the good that’s all around you.