Science, Social Media, and the Loss of Context

I know it’s not an airport, and I’m not announcing my departure. But I’ve been asked why I deleted various social media accounts, so I’ll explain my thinking.

We lose something easy to overlook when we take scientific discourse to algorithmic, for-profit social media platforms.

I used to advocate for using social media for scientific communication and exchange. At its peak, #AcademicTwitter was an amazing space. I learned from it. Discovered papers and conferences. I know colleagues who found jobs on Twitter. It was great. But now I’ve chosen to mostly opt out. That decision goes beyond my academic life, but I’ve also concluded that today’s social media technologies, especially for-profit algorithmic platforms, are probably not very good for science. They might be more harmful than we realize.

This isn’t a manifesto or a call to action. Nor a condemnation. I am not a moral authority. I’m just a guy with a few thoughts, none of which are original. Most of what I share here repackages arguments from Lanier, Newport, Ward, and others; I’ve collected many of the sources in my antisocialmedia bibliography. It’s also the product of a years-long discussion with my closest friend, Brett Wertz. I’m sharing it for those who ask or happen across it on their own (impressions be damned). But first, if you read only one essay about science and social media, make it Simon DeDeo’s The 11th Reason to Delete your Social Media Account: the Algorithm will Find You. In fact, you should probably just go read that instead.

But if you’re here for my rant, here it is:

Many academics feel obligated to be active on social media. To stay informed about current trends or take part in scholarly debate. To learn about jobs and grants for themselves, their students, or their colleagues. And, for some, to communicate their expertise to the broader public. These are all reasonable, even noble, motivations. For scientists, talking about science is fun. It’s our passion. And for those of us who don’t have easy access to intellectual colleagues, social media seems like an easy solution.

Social media is also a marketing tool, and academics are professionals whose livelihoods often depend on promoting their work. It is useful for many people, even now, in 2026. Many of those same people, academics included, know how unpleasant and distorting the current social media landscape can be. Most of us would agree that it is something different from what it was in 2016. These platforms often bring out something other than our best selves. We all know that colleague: so kind and endearing and always in good faith in real life, yet somehow their social media avatar regularly spews vitriol 280 characters (or more) at a time. Still, many of us—and many colleagues I deeply respect—remain active users. Maybe because of habit, compulsion, or sunk costs. Or maybe they’ve weighed the pros and cons and decided that their own use is worthwhile. It can certainly feel as if there is no other option, that this is just the way things are now: accept it or get left behind. The benefits are often easy to see. They’re just variable, depending on the person and what they want from a platform.

But what are the costs?

And are the assumptions we use to justify handing the format and content of our scientific conversations to black-box algorithms and a handful of tech CEOs actually warranted?

Academics often treat widely used social media platforms, like X, as a digital public square where everyone has an equal chance to speak and ideas compete on merit. An idea marketplace. It is indeed a market. But the invisible hand is not what neoclassical economics has in mind. The market isn’t organized around user welfare, and the user is the product, not the consumer. Just to be clear: you are the product (even if you’re paying). As we’re all well aware, proprietary, opaque algorithms control distribution. In a sense, the algorithm is the real user. The rest of us supply the material. It rewards outrage, novelty, and emotional display, not because it cares about emotions but because they drive engagement. Evidence and nuance are not necessarily valued. What looks to us like earnest scientific dialogue may instead be an exchange arranged by the algorithm to maximize engagement across the network. In other words, to make money.

Science is a human enterprise, conducted by humans, and we are a deeply social species. Our communication depends on local and cultural cues, etiquette, and layers of norms that make cooperative exchange possible. Many of those cues disappear when we move the conversation onto social media. The mechanisms we’ve developed over millennia to reduce confrontation and help complex ideas travel between minds are either absent or not working very well. That’s a vulnerable position to be in. Algorithms exploit it. They selectively amplify and suppress parts of what could be a meaningful exchange in order to capture a particular kind of attention. We end up discussing complex subjects through a medium stripped of much of what makes complex communication work. That mismatch is part of the product.

Offline, we are constantly reading contextual cues. Often quite subtle ones. Tone, posture, dress, and status signals help us judge other people’s intentions and regulate our own behavior. Online, many of those cues disappear. When we don’t really know who we’re speaking to, we may assume similarity or hostility. Academics who would normally tread carefully through epistemological minefields can easily forget to do so online. The cues that elicit empathy are gone, while the prestige economy of likes and reposts encourages forms of engagement that are often less than prosocial. Perhaps there are—or could be—norms for online social behavior. Maybe we have already developed some, or will develop more, just as we have offline. But online interactions still happen under algorithmic control, without face-to-face contact and often without any expectation of a continuing relationship. You can always just block people. So I’m skeptical that these platforms provide the conditions in which prosocial communication norms can really take hold.

Academics are people (for the time being), and people are status-seekers. Academics are often hyper-status-seekers. That’s not a pejorative. It’s a fact of human psychology and of how the profession works. Social media exploits these evolved incentives. Its prestige economy can make scholars feel that their tightly packaged insights amount to real scientific communication. But the feedback may be pushing them toward performance: optimizing for the appearance of science rather than its content, even if they never consciously choose to. It’s just operant conditioning.

We also might think these platforms act as a leveling mechanism. Anyone with something worth saying gets heard. But the algorithm that curates, amplifies, and buries content without context is itself a form of ideological and social stratification, and maybe a more controlling force (or at least the devil we don’t know) than what it replaced.

Perhaps the highest cost of sharing thoughts and conversations on algorithmic social media is the loss of control over context. These platforms lack many of the structures that support serious discussion: a shared vocabulary, peer review (formal or informal), shared standards, and disciplinary norms. In their place, we get decontextualized snippets and reactive discourse.

Imagine being invited to present your work, but when you arrive, you discover it’s actually a competition—a TED Talk Battle Royale. You don’t know the rules. You don’t know who the other contestants are or what they’re saying on the stages next to yours. You don’t know who’s in the audience, what they watched before they got there, or what they’ll watch afterward. The host, who has an agenda of their own, decides on the spot whose mic gets turned up, whose gets turned off, and what gets shown to whom. That’s roughly what it means to have a scientific conversation on social media. In a journal, a seminar, or a conference, you know the “room.” On social media, you mostly don’t know what’s going on, and someone else—the algorithm, the platform CEO, the engineers—is running the show.

You have little control over how what you share gets broadcast. Science communication is already delicate, even under the best conditions. On social media, scientific discussion becomes fodder for the algorithm, which presents it to other users with one goal: capture their attention and monetize it, usually through advertising.

There’s a version of this debate that gets stuck on which platform. Heterodox types will tell you Bluesky is a left-wing echo chamber, a curated safe space for progressives who’d rather not be challenged. Progressives will tell you X is a cesspit shaped by one man’s politics, with content moderation gutted to serve his ideology. Both are probably right, at least to some extent, about the other. What neither camp often admits is that the echo-chamber problem doesn’t come only from which platform you choose. It also comes from algorithmic curation itself. Every for-profit algorithm is, by design, a bubble. It learns what keeps you engaged and feeds you more of it. The diversity of perspectives you feel you’re getting—or the lack of it—is part of the product. If you think your feed has escaped this, if you feel as though you’re genuinely encountering the full range of opinion on a given platform, then the algorithm is working. It’s got you (cf. DeDeo).

Using social media and investing in other forms of scientific communication are not mutually exclusive. Someone can be active on X or Bluesky and still write long-form essays, maintain rich email correspondences, and show up to regional conferences. I’ll admit I’m not fully off these platforms myself. I autoposted this blog to Bluesky so people could comment through the AT Protocol. That kind of deliberate, narrow use can be productive while avoiding some of the costs I describe. But I’m thinking about another trade-off too. Social media platforms increasingly seem to replace other forms of interaction rather than add to them. That concerns me most. If that’s what is happening, we need to be honest about what we’re accepting. If algorithmic platforms become the main medium of scientific communication, then their owners, CEOs, engineering teams, and proprietary goals will exert real and largely unaccountable influence over how science is communicated, encountered, and ultimately shaped. I don’t think this is an overly doomer take or a hypothetical risk. It has already happened. It is a structural feature of an arrangement we’ve quietly, and often uncritically, accepted.

Algorithmic social media offers real benefits, even for scientists. You might discover an idea you wouldn’t have found otherwise. A new colleague. A funding opportunity. A job. A way to sell more books. All of that can advance careers and, sometimes, science itself. What we don’t discuss enough is the filter. On for-profit algorithmic platforms, only certain ideas, colleagues, and opportunities surface. The algorithm decides what crosses your feed, and it is not optimizing for scientific progress. We haven’t seriously considered what that might mean in the long run. If the main way scientists encounter each other and each other’s work is controlled by an objective function that has nothing to do with the health of science, we should at least ask what that does to the enterprise over time. I’m skeptical it’s a good thing.

One cost is harder to quantify. The behavioral logic of algorithmic platforms doesn’t just influence what we see on our screens. Eventually it starts shaping how we think. The feedback loop of likes and engagement is a form of conditioning, and one of its subtler effects is that you stop encountering ideas as ideas. You start seeing them as potential posts. Your intellectual life begins to pass through the question of how it would “perform.” That isn’t just a matter of some people using the platforms badly. It is a predictable result of sustained exposure to their incentives.

I’m not a tech doomer. I’m not arguing against the internet, or even against social media as a whole. But I don’t think we’re honest enough about what algorithmic platforms cost us. We mostly see one side: what we gain. We rarely see what’s through the looking glass.

I’m aware there is no obvious replacement. Even platforms less driven by algorithmic feeds, like Bluesky or Substack, carry versions of the same problems. Likes and reposts strip context regardless of who owns the server. Substack has its own engagement and algorithmic dynamics worth being skeptical of.

But I also think we’ve quietly accepted a premise worth questioning: that science requires this kind of scale and speed. Does it? I don’t think so. Maybe getting a job does these days. That’s another problem. And it isn’t only PhD students and postdocs trying to make a name for themselves online. Science did reasonably well without any of this. People built networks through email, listservs, phone calls, societies, conferences, and the slow accumulation of trust among people working on shared problems. That still works. It just takes effort. Maybe that friction is worth keeping. It gives us a science that moves deliberately and rests on substantive exchange with people we actually know, rather than one optimized to capture strangers’ attention and make algorithms happy.

I have plenty of friends and colleagues who are fully bought in on one platform or another. Most of them aren’t naive about it. They recognize the costs, the distortions, and sometimes even the dangers of the broader ecosystem. I also recognize that there are many ways to be a user (product) of these technologies. Casual users (products) do not have the same experience as frequent ones. Some colleagues get genuine value from the particular way they use these platforms, and they’ll tell you so. I’m not thinking of any one person here, or any one field. But I’ve noticed a pattern in the justifications I hear. So I want to work through the counterpoints that come up most often and offer my gut response to each.

“I just block trolls and only interact with thoughtful colleagues.”

Algorithms still shape what you see and how your posts get rewarded. Even a carefully curated feed sits inside a system that promotes controversy, emotional reactivity, and performance that drives engagement. Your careful curation can help, but it can’t remove the platform’s incentives.

“It’s the easiest way to meet collaborators and stay visible.”

Visibility is not the same as credibility or impact. Online networking often rewards people who are accessible, witty, or simply good at posting. Those aren’t the same things as producing rigorous scholarship. The result can be another shallow reputational hierarchy rather than meaningful scientific exchange.

“It lets me communicate science directly to the public.”

Algorithmic distribution can distort that communication by amplifying controversy and suppressing nuance. Public outreach on social media can easily slide into spectacle, with engagement metrics masquerading as understanding or impact.

“Social media democratizes visibility for those outside elite institutions.”

In practice, many of the same hierarchies carry over online. Algorithms tend to favor people who are already prominent and culturally attuned—consciously or not—to what performs well. Those who make the algorithms happy get more visibility. So various structural inequities are still there.

“I get encouragement and useful feedback from peers and readers.”

Some of that feedback is genuinely useful. But the reward loop itself can be deceptive. Likes provide dopamine and instant gratification, which is likelyb to steer us away from careful self-reflection. They can also push scholars toward performative self-presentation without realizing it.

“Academic publishing is slow; this is a real-time conversation.”

Yes, it’s fast. That’s both the appeal and part of the problem. Speed can come at the expense of reflection and accuracy, replacing slow, cumulative discussion with reactive commentary.

“Posting about injustice or science helps raise awareness.”

Posting can simulate activism while substituting for action. Platforms reward performative outrage and symbolic gestures more reliably than difficult, coordinated work, channeling energy back into engagement metrics. Taking the conversation or the activity offline is what the algorithm most fears.

“There’s nowhere else to reach such a large audience.”

A large audience isn’t necessarily an attentive one. Academic influence still depends on credibility, context, and sustained attention, all of which can be difficult to maintain on social media. So, scale does not always equal impact.

There are, of course, valid counterpoints to all of these takes.

But I think the deeper problem isn’t just that algorithmic social media is a poor venue for scientific discourse. It’s a convincing simulacrum of one. It gives you the feeling of building an audience, developing ideas, and engaging a community while crowding out the slower, more effortful forms of communication where those things happen better.

Some alternatives worth investing in instead:

Cultivate small-scale communication ecologies. It’s easy to overlook that algorithmic platforms didn’t just add something to our interactions. They also displaced, and continue to replace, other forms. Our attention is finite, and social media is designed to win that battle. Sometimes now, when I send a group email or text a photo or article to a handful of friends, it almost feels strange, like a relic of another era.

Recently, a senior scholar in my field—a true landmark figure—emailed two colleagues and me about a paper we’d published. It was so refreshing. We replied. He replied. The thread had a beginning, a few points of engagement, and a resolution. Later I saw him at a conference, and we continued the conversation in person. And it was ours. It wasn’t a spectacle for the algorithm. That is a scientific relationship. A conversation driven by the people having it.

There are more formal alternatives worth investing in too: small regional gatherings in my field, such as FOSSILS, NEWEPS, the Northwest Evolutionary Science meeting, and the French Network for the Evolutionary Study of Humans. I’m also still in a WhatsApp group from my postdoc institution. It’s mostly people asking who’s getting lunch or who got locked out of the building, but fairly often something scientifically interesting comes up. The researchers currently there use it for longer scientific conversations, which they can then take to lunch or the pub. Those communities don’t maintain themselves. They take a bit of leadership and investment. But what they offer is a conversation that belongs to the people in it, not to an algorithm.

Focus on depth over reach. Reinvest time spent chasing engagement in writing for long-form venues: Substack, Aeon, SAPIENS, The Conversation, Works in Progress, or university and personal blogs. Long-form writing makes room for argument, reflection, and context—the things that make scholarship meaningful. Even a thousand words can be enough to put an idea out there. Anyone can start a personal blog on their own website for free, and posts can be archived through services that issue a DOI. Public comments are probably a good thing.

Engage public audiences directly. Work with schools, museums, or libraries. Give short talks or Q&As for students and community members. These forms of engagement are slower and reach fewer people, but they may be far more meaningful. They build real relationships.

Use private or semi-private forums for discourse. Create online communities that aren’t optimized for “user engagement”: moderated Slack, Zulip, or Discord groups; mailing lists; or Google Groups organized around specific research interests. These can host dialogue without the distraction of metrics.

Practice slow outreach. Instead of reacting to trending topics, work on essays, podcasts, or short video explainers that communicate research clearly and on your own schedule. Communication can be scholarship, and need not be content. It’s no coincidence that blogs and vlogs no longer occupy the place they once did in our information landscape. They didn’t stop being useful. Ad-based, algorithmically driven platforms developed a near-monopoly over our attention. Like moths to a flame.

Collaborate with journalists and editors. If visibility is a goal, partner with professionals who understand and value storytelling and the ethics of representation. Work to get your research to circulate through curated, editorially accountable platforms rather than algorithmic feeds.

Build micro-networks of mentorship and exchange. Use professional associations, graduate seminars, or small workshops as places to exchange ideas. That has long been one purpose of scholarly community, before social media came to mediate so much of it.

In a recent episode of the Ezra Klein Show, guest Derek Thompson paraphrased Robert Putnam, the author of Bowling Alone. The observation has stayed with me. Putnam’s point was that we too often adopt a technology and then adopt that technology’s values, without stopping to think about how the technology might fit within our own values. Scientists are generally pretty sensitive to costs and benefits. But I don’t think we’re always as honest with ourselves about the values embedded in the platforms we use, or about the subtle ways those values begin to shape not only how we communicate science, but how we think about science, and perhaps how we think.