The Case Against Social Media is Weaker Than You Think

The evidence that social media broke America is surprisingly weak. So why are so many people convinced it did?

Many people have the intuition that social media broke the USA. More precisely, they place a lot of the blame for America’s current political problems on the emergence of social media platforms like Facebook, Twitter (X), YouTube, and TikTok.

Some examples:

• Alexandria Ocasio-Cortez describes Meta as “a cancer to democracy metastasizing into a global surveillance and propaganda machine for boosting authoritarian regimes and destroying civil society.”

• Jonathan Haidt has argued that social media platforms “dissolved the mortar of trust, belief in institutions, and shared stories that had held a large and diverse secular democracy [in America] together.”

• Obama suggests that one of “the biggest reasons for democracies weakening is the profound change that’s taking place in how we communicate and consume information.”

These viewpoints express a broader conventional wisdom among many intellectuals, pundits, and politicians.

In some ways, this conventional wisdom is understandable.

First, there appears to be a lot more brazen lies, falsehoods, bullshit, conspiracy theories, and plain stupidity and dysfunction in American political culture than there was a few decades ago.

Second, this development appears to coincide with the emergence of social media platforms. Facebook launched in 2004, YouTube in 2005, and Twitter in 2006. As these platforms were adopted by hundreds of millions of users over the next decade, the health of American democracy seemed to deteriorate. By 2016, the year of Trump’s first election, many experts began talking of an "epistemological crisis”, a “disinformation age”, and a new authoritarian era.

Third, social media platforms feature a shocking amount of false, misleading, and incendiary content. And this is not surprising. The absence of traditional gatekeeping allows charlatans and know-nothing pundits to reach vast audiences, and the profit-maximising design of platforms is often hostile to truth and rational discourse. In a competitive attention economy mediated by engagement-grabbing algorithms, outrage-generating, identity-affirming, and bias-confirming content usually outcompetes thoughtful analysis and debate.

For these reasons and more, many people view social media as a kind of technological wrecking ball. As the famous “rationalist” writer Eliezer Yudkowsky recently put it,

“My current rough sense of history is that the last "moral panic" about social media turned out to be accurate warnings. The bad things actually happened, as measured by eyeball and by instrument. Now we all live in the wreckage. Anyone want to dispute this?”

I want to dispute this. More precisely, I don’t think there is strong evidence to support this assessment, and there are some good reasons to be sceptical of it. In a recent article for Asterisk magazine, I present a detailed, data-driven argument outlining the reasons for this scepticism.

Scapegoating the Algorithm

To simplify greatly, the core argument¹ I advance can be summarised like this:

• People exaggerate the novelty of America’s political and epistemic challenges, many of which can arise (and historically have arisen) in the absence of social media, often in much worse forms.

• Social media doesn’t provide a very plausible explanation of recent political developments like intense partisan polarization, low institutional trust, and the political and epistemic dysfunction associated with MAGA and the Republican Party. First, increasing partisan polarization and institutional mistrust long preceded the emergence of social media. Second, social media-based explanations struggle to explain why these trends are so different (and sometimes reversed) in countries with similar rates of social media use. Third, the very fact that America’s most acute challenges are concentrated on the political right sits uneasily with social media-based explanations. (Platforms do not run different algorithms for conservatives and liberals.)

• A better explanation of this “asymmetric polarization”—in less technical language, the fact that the Republican Party is much more dysfunctional than the Democratic Party—has to do with America’s “diploma divide”. As the Democratic Party has become the home of college-educated urban professionals and the Republican Party’s base has shifted to white Americans without college degrees, conservatives have become hostile to institutions (science, media, academia, etc.), which they perceive—mostly accurately—as dominated by liberals who don’t share their values and look down on them. The result has been a “crank realignment”: the Republican Party now provides the natural home for resentful, anti-establishment, low-trust, conspiracy-minded cranks, as well as a media environment that is increasingly unconstrained by basic scientific knowledge and rudimentary norms of professional journalism and fact-checking.

• Large-scale, randomised field experiments that alter people’s social media experiences consistently find either minimal or no effects on political attitudes.

• Although these findings are very limited in various ways, they align with decades of research demonstrating that political manipulation and persuasion are highly challenging, media effects are minimal, and people are sophisticated and active consumers of information. In other words, people’s attitudes, worldviews, and voting intentions tend to be deep-rooted and shaped by far more consequential factors than the content they happen to come across in media, including social media, not least because people exercise so much control over which media (including social media) they consume. (Media bias is largely demand-driven, not supply-driven).

In this post, I will briefly consider various objections to these arguments:

Objection 1: The essay targets a “straw man”. Nobody, or nobody serious, or nobody serious worth criticising, thinks of social media like a technological wrecking ball that has “broken” America. Social processes always involve complex and dynamic interactions among multiple factors, rather than simple, mono-causal stories.

Response: First, many people do think of social media this way. (Return to Yudkowsky’s claim that we live in the “wreckage” of social media or Haidt’s viral Atlantic article.) I often encounter the idea that social media has single-handedly destroyed democracy or has had a vast range of deleterious consequences, so it seemed valuable to write a detailed article explaining why the evidence for this view is not strong.

Second, the “wrecking ball” framing is simply a helpful device for making a broader point: that we lack strong evidence for alarmist views about the harmful impacts of social media. Even if one believes that these impacts arise from highly complex, multi-causal, and dynamic processes, this point would still apply.

Objection 2: Questions like, “What are the effects of social media platforms?” or “Did social media have negative consequences?” are either meaningless or pointless. “Social media” refers to a diverse range of platforms that are constantly evolving, used by different people in various ways, and have many consequences, some positive and some negative. It’s not helpful to generalise about them. It’s like asking, “What’s the effect of food on the body?”

Response: First, although social media platforms are diverse, many people have argued that they share harmful properties (engagement-maximising algorithms, the absence of elite gatekeeping, etc.). This is why so many people believe it makes sense to attribute social and political problems to social media in general. So, it seems reasonable to see what the balance of evidence implies about the plausibility of such explanations.

Second, I nevertheless agree with the broader point here. When considering social media, society, and politics, many discussions would benefit from being more focused and nuanced. TikTok is very different from Facebook. X is very different from Substack. And people and cultures are also very different, meaning one shouldn’t expect platforms to have uniform effects on users. For example, I suspect that many social media platforms produce net benefits for intelligent and conscientious individuals, but net costs for those who are less intelligent and less conscientious. (I suspect many aspects of technology and modern culture are like this).

Objection 3: My argument that America’s political and epistemic challenges are worse on the political right merely reflects my own liberal bias.

Response: Maybe this is true. Everyone is biased, and it’s reasonable to point out that much of the research allegedly documenting that epistemic problems are worse on the right than the left is probably untrustworthy in various respects. (Breaking news: “Liberal researchers discover that liberals are better than conservatives!”).

As I wrote in the article, it’s challenging to measure these things “scientifically” in ways that would be persuasive to an ideologically diverse audience. Nevertheless, I find it hard to see how intellectually honest people could seriously look at mainstream liberal and conservative politics and media in the USA and conclude that the latter is no worse than the former. Yes, there are deep problems on the mainstream left—problems of bias, propaganda, highbrow misinformation, ideology, groupthink, and so on—but such issues pale in comparison to the frequency, brazenness, and egregiousness of falsehoods spread by figures like Donald Trump, Elon Musk, RFK Junior, Tucker Carlson, Steve Bannon, and Candace Owens.

I talked about this issue at length with Yascha Mounk here:

Dan Williams on Misinformation

However, suppose that you’re a MAGA Republican and think America’s epistemic problems are much worse among liberals. If so, the argument of my essay would be even stronger. Because liberals are so much more closely connected to legacy media and mainstream institutions than conservatives, it’s even more implausible that their problems have their roots in social media specifically.

Objection 4: Yes, problems are much worse on the political right, but liberals have also become more “structurally stupid” in recent years, which cries out for explanation.

Response: I don’t share this intuition. I’m also unsure how to adjudicate this disagreement with data. Typically, those who argue for this viewpoint (e.g., Haidt or Jonathan Rauch) point to the emergence of an aggressive and unpleasant form of progressive activism associated with the “Great Awokening,” which has become increasingly influential within the Democratic Party in recent years. Because woke activists launched a frenzy of social media-based cancellation campaigns aimed at destroying heretics and punishing wrongthink, it seemed plausible to many people that social media was the driving force behind this development.

Maybe—but I’m sceptical. First, social media remains, but the Great Awokening has subsided. Today, the suffocating political monoculture it creates is primarily confined to fringe parts of the internet, such as Bluesky, which most people happily avoid. Second, as Musa al-Gharbi documents in We Have Never Been Woke, these “Great Awokenings” have happened many times throughout history before the emergence of social media. Finally, to the extent that the mainstream liberal coalition has become more monolithically liberal and groupthinky in recent years, this seems easily explained by a general trend towards growing partisan polarization and political sectarianism that began long before the emergence of social media.

Objection 5: Even if nations with similar rates of social media use are experiencing some different trends and outcomes, this observation misses the forest for the trees. What explains the striking similarities we observe worldwide, such as the concurrent rise of right-wing populism in numerous nations?

Response: First, although it is very much not my area of expertise, I think it’s easy to exaggerate this trend towards growing right-wing populism, in part because “right-wing populism” bundles together phenomena that are very different. For example, although the UK’s “right-wing populist” party, Reform UK, shares some similarities with MAGA, there are also significant differences. To take only one, I struggle to imagine Reform UK running a sophisticated disinformation campaign to overturn the results of a democratic election.

Second, the features often associated with right-wing populism (xenophobia, anti-“elite” and anti-“establishment” politics, the desire for a strong leader, demagogic rhetoric, etc.) have historically arisen in much worse forms before the emergence of social media, not least in the many popular fascist movements in the first half of the twentieth century.

Third, although there is significant scholarly controversy over the causes of recent trends towards right-wing populism, many factors seem to have little to do with social media, including globalisation, mass immigration, the long-term economic fallout from the 2008 financial crisis, educational polarization, and a powerful cultural backlash against progressive social change. I also think people vastly underestimate the significance of the two World Wars and the Cold War in creating unusually cohesive, interdependent national cultures in Western countries during the second half of the twentieth century.

I don’t want to be overly dismissive here. For example, social media platforms have two features that have probably favoured populist movements. First, the complete absence of traditional gatekeepers provides “counter-elites” a direct way of reaching large, sympathetic audiences without going through established institutions. Second, as Martin Gurri points out in The Revolt of the Public, the information tsunami unleashed by social media makes it more challenging for these institutions to maintain public trust. For every mistake, gaffe, hypocrisy, and scandal, there is an army of influencers ready to identify, exaggerate, and serve them up to hungry audiences.

For these reasons, it’s at least plausible that the social media age has been favourable to populist politics. However, that concession is very different from much of the alarmist discourse about how social media broke democracy.

Objection 6: The evidence isn’t strong either way. Even if we lack strong evidence that social media has had devastating effects, we also lack strong evidence to suggest it hasn’t.

Response: I agree and try to reflect this nuance in the essay. When it comes to the harms of social media, people sometimes conflate an absence of evidence with evidence of absence. That is, they treat an accurate observation (there isn’t strong evidence that social media has had the catastrophic effects often attributed to it) as equivalent to an inaccurate observation (we have strong evidence social media hasn’t had catastrophic effects).

Suppose you approach this topic with a high degree of confidence that social media has had devastating effects. (You have a strong “prior” that this is the case). How much should you reduce your confidence in this view when confronted with the evidence and arguments I present in my essay? Probably not much.

For every point I make, you could respond as follows:

• Yes, there have always been epistemic challenges, but they are manifestly worse now than in the recent past.

• Yes, trends towards growing polarization and institutional mistrust predate the emergence of social media, but social media has accelerated them and made them worse.

• Yes, countries and political coalitions with similar rates of social media use are experiencing different trends. Still, there are also some worrying similarities in recent trends across countries, and social media has interacted with country-specific and party-specific factors in ways that exacerbate them.

• Yes, some large-scale, randomised field experiments show minimal or no effects on political attitudes, but these experiments are extremely methodologically limited. And some other experiments seem to point in a different direction.

• Yes, people are not gullible rubes easily manipulated by influencers and algorithms, but spending vast amounts of time consuming low-quality, biased, and misleading content online will still impact people’s attitudes and behaviours in many ways.

These are all fair responses. Nevertheless, the central point of my essay remains: There isn’t strong evidence for the hypothesis that social media has had the kinds of effects confidently attributed to it, and there are some good reasons to be sceptical of this hypothesis. Given this, people should be honest that when they make bold claims about its effects, these claims are not supported by rigorous social-scientific analysis.

This situation is nicely captured by Scott Alexander, who seems to endorse the “wrecking ball” narrative:

“I don't want to affirmatively come out in favor of social media, but I think we're still at our prior, rather than having gotten extra evidence… Most studies seem to show that social media does not increase political polarization… I'm not saying we should follow studies off a cliff, or that I'm sure social media had no negative effects. I'm saying that we started with this very pessimistic prior that social media would destroy society, and we still sort of pessimistically feel like social media is destroying society, and maybe we're right, but we can't say we "confirmed" our prior and now our pessimism has been "vindicated".”

The “most accurate description”, he writes, “is that the evidence contradicts our prior, but the prior is so strong that we're choosing to stick with it anyway.”

This is a good summary of the situation. The conviction that social media has broken democracy is a popular and powerful narrative, not a well-supported theory. People feel it's true, so they stick with it, even when the evidence doesn't support it.

The question is why people feel the story is true. My sense is that its popularity has much more to do with vibes, anecdotes, and subjectively compelling stories than with data and rigorous analysis. Hopefully my essay will motivate some people to make a more persuasive case for it.

1. This arguments draw on the work of many social scientists, including Alberto Acerbi, Hunt Allcott, Sacha Altay, Levi Boxell, Ceren Budak, Matthew Gentzkow, Matt Grossmann, Andrew Guess, David Hopkins, Lilliana Mason, Hugo Mercier, Brendan Nyhan, Talia Stroud, and Joe Uscinski.