In the fall of 2017, Collins Dictionary crowned “fake news” word of the year, and on the surface it made sense. Journalists were using the phrase to sound the alarm about a polluted online information environment. Academics were publishing papers, naming conferences after it, and desperately trying to understand what was happening. And, of course, then-President Donald Trump was using it from behind the podium to discredit nearly anything he disliked. But by the spring of 2017, I had already grown exasperated. The term wasn’t just imprecise; it had become a weapon. It was being used to attack the news media rather than to describe the real problem unfolding online. Worse, “fake news” never actually captured the phenomenon it claimed to. Most of the content that worried researchers wasn’t fake at all. It was genuine material stripped of context: a real photograph with a misleading caption, a scientific quote cherry-picked to prove a false point, an old video resurfaced as if it were new. And very little of it looked like what we traditionally think of as news. So I joined a chorus of scholars pushing for more precise language. We adopted “misinformation” for false or misleading claims shared without intent to harm, and “disinformation” for the deliberate, harmful variety, with “malinformation” covering the weaponization of authentic content. These terms caught on, and they are useful. But I have come to believe they are too clean, too simple, too focused on the individual artifact. We spend so much time labeling dots that we fail to see the pattern they form. We ask whether a single post is accurate and whether the person who shared it meant harm, and in doing so, we ignore a much larger and more uncomfortable set of questions: why people create and share this content, how it helps them build identity and community, what needs it fills, and what narratives it reinforces over time. We will not strengthen information ecosystems by perfecting our typologies. We will only do that by shifting our gaze from the individual post to the social context, from the atom to the network, and from the single claim to the story it belongs to.
To understand what these labels miss, it helps to imagine a woman I will call Lynda. Lynda is not a real person, but she is a composite of many people I have tracked online, and she is far more complicated than our vocabulary gives her credit for. Lynda fervently believes vaccines are dangerous. She spends enormous amounts of time researching this belief. She scours databases for newly published scientific studies. She watches regulatory hearings about vaccine approvals. She reads vaccine inserts to analyze ingredients, side effects, and warnings. Then she shares what she learns with her community online. Now, how do we classify Lynda? She is not a misinformer in the usual sense. She is not lazily forwarding content she never bothered to check. She verifies relentlessly. She is not a disinformation agent either, at least not by the standard definition. She is not trying to cause harm or make money. My own sense is that Lynda is driven by something more complicated and more human. She feels an overwhelming need to warn others about a health system she sincerely believes has harmed someone she loves. She is selecting information strategically, not to deceive, exactly, but to connect with people and to promote a worldview. Her standards for what she posts are not based on whether the information is rationally valid; they are based on whether it resonates with her social identity and moral values. If we dismiss Lynda for lacking research credentials or for ignoring nuance, we miss what she is actually accomplishing. She takes snippets of accurate information from authoritative institutions—a scientist conceding that more research is needed, a disclaimer about known side effects—and shares them without context. The individual fragments may be true, but they are used to support deeply inaccurate narratives: that governments are rolling out vaccines to control populations, that doctors are dupes, that pharmaceutical companies are hiding catastrophic harm. This is not simply a failure of fact-checking. It is a success of storytelling. Lynda is not an outlier. She is a warning. To understand our information ecosystem, we have to stop obsessing over accuracy and start looking at how online spaces thrive on connection, affirmation, and identity. As the communications scholar Alice Marwick has put it, in social environments people are not necessarily trying to inform others; they share stories, pictures, and videos to express themselves and broadcast their identity, affiliations, values, and norms. That motivation is not unique to conspiracy theorists. It applies to Beatles fans and cat lovers, to social justice activists and vaccine skeptics alike.
Lynda’s online world also reveals something else that our tidy labels cannot capture: connection. If you were to follow her digital trail, you would likely find her not only in anti-vaccine Facebook groups, but also in #stopthesteal groups, sharing climate denial memes, or speculating about the latest mass shooting on Instagram. Yet most researchers never follow that trail far enough. The academic world is divided into silos. I work in a school of public health, so people assume I should study health misinformation. Colleagues in political science are funded to investigate threats to democracy. The climate communication scholars study climate misinformation. The security experts study Russian disinformation in Ukraine. Each group holds its own conferences, publishes in its own journals, and rarely talks to the others. Funders and policymakers reinforce these silos by concentrating on nation states and specific regions, or on scheduled events like elections, vaccine rollouts, and climate summits. But the actors who are most effective at polluting the information environment do not respect those boundaries. They exploit moments of tension and outrage, not ceremonial moments with advance calendars. They show up around a royal documentary, a celebrity divorce trial, or the World Cup, and they adapt their narratives to whatever people are already angry or anxious about. No one funds sustained research into those moments, even though studying them could teach us a great deal. The same siloing happens on the response side. In November 2020, my team published a report based on twenty million posts we had gathered from Instagram, Twitter, and Facebook about COVID-19 vaccines. We deliberately collected conversations broadly, not just content that looked false, because we wanted to know how people were actually talking. The team identified several major narratives: safety, efficacy, necessity, and the political and economic motives behind the vaccines. But the most frequent conversation of all was one we labeled “liberty and freedom.” People were less worried about whether vaccines were safe than about whether they would be forced to get vaccinated or required to carry vaccine verification. And yet institutions like the Centers for Disease Control and Prevention were only structured to address the safety and efficacy narrative. They had no framework for responding to a story about coercion and liberty. The result was a vacuum, and vacuums are where bad actors thrive.
Most scholars and platforms still think in terms of what I call atoms of content. Social media companies have teams that decide whether an individual post should be fact-checked, labeled, down-ranked, or removed. They have become increasingly good at playing whack-a-mole with posts that may not even violate guidelines. But this approach misunderstands influence. People are not persuaded by a single post. They are influenced by narratives, and individual posts only matter to the extent that they fit into those narratives. I like to think of individual posts not as atoms but as drops of water. A single drop of water will not persuade anyone, and it will not do much harm. But over time, repetition gathers power. Month after month, people see posts that are “just asking questions” about government institutions, public health agencies, or the integrity of elections. Like water dripping on stone, each drop leaves no visible trace, but eventually grooves are cut deep. Trust erodes slowly and then suddenly. That is the real damage of a polluted information ecosystem: not any single lie, but the slow destruction of shared reality and institutional credibility. To respond effectively, we have to zoom out. We have to stop treating accuracy as the only metric and start mapping the larger narratives that travel across platforms, languages, and communities. We also have to be honest that trusted institutions have not kept up. It has become easy to blame Russian trolls on Facebook or anonymous teenagers on 4chan, but the truth is more uncomfortable. There have been many moments when those responsible for providing clear, actionable information to meet communities’ needs failed to do so. Bad actors took advantage of that vacuum. They still do. We need a more integrated view of how and why information circulates, and we need to ask not just “Is this post true?” but “What story is this post telling, who needs to hear it, and why does it make sense to them?”
So what can be done? The first priority is to organize and fund cross-cutting research. We need a permanent, global institution to monitor and study the information environment, centrally funded and therefore independent of both governments and technology companies. Right now, many organizations are doing overlapping work on disinformation, but they fail to share data, coordinate classifications, or respond quickly in a crisis. A shared infrastructure would allow us to see the connections that Lynda’s path reveals, across topics, platforms, languages, and regions. The second priority is to learn how to participate, not just broadcast. The polluted information ecosystem is participatory, and the people who are best at it are constantly experimenting, engaging, and adapting to their audiences’ concerns. News outlets and government agencies may be on social media, but most of them still treat it as a one-way megaphone. Traditional science communication is still top-down, based on what I think of as the paternalistic deficit model: experts assume that if they just supply the correct information, audiences will passively consume it and do the right thing. That model has failed. We have much to learn from people like Lynda about how to connect with audiences rather than simply present to them. That starts with training government communications staff, community organizations, librarians, and journalists to genuinely listen to the public’s questions and concerns, and to take those concerns seriously enough to respond to the underlying feelings, not just the surface claims. The third priority is to support community-led resilience. Funders are currently obsessed with platforms, filters, and regulation—trying to expunge the bad stuff rather than expand the good stuff. But whack-a-mole content removal is not a strategy. Instead, major funders should invest in place-based responses built around what specific communities need. Consider the Health Advocates In-Reach and Research campaign, created by the health researcher Stephen Thomas, which trains local barbershop and beauty salon owners to listen to their customers’ health concerns, offer advice, and connect people to follow-up care. Or consider El Tímpano, the participatory online news site founded in 2018 by journalist Madeleine Bair after she found that the Spanish-speaking community in Oakland, California was woefully underserved by local media. These are not flashy platform interventions. They are human infrastructure, and they work.
Finally, we have to commit to the long game. There is no quick fix, no algorithm switch that will solve all of this, no single law or platform policy that will restore trust. Improving the health of our information ecosystems will be a decades-long process, and much of that time will be spent playing catch-up. The technology will keep evolving, and just when we think we understand the current landscape, a new disruption like ChatGPT will emerge seemingly overnight. But the underlying lesson will not change: the answer is not to build bigger lists of bad posts or to label more people as misinformers. The answer is to look beyond the neat diagrams and tidy typologies and see what is actually going on. It is to recognize that people are not empty containers waiting to be filled with correct information. They are complex, emotional, social beings who seek connection, affirmation, and community. They share things not because they are stupid or evil, but because the stories they share make sense inside the story of their lives. People like Lynda may be spreading harmful narratives, but they are also modeling something important: how to build trust, how to speak to people’s fears, how to show up consistently and sincerely. If we want to build a healthier information ecosystem, we need to understand what makes those voices compelling and what needs they are answering. We need to offer better stories, better communities, and better ways for people to feel seen and safe. That is not a task for fact-checkers alone. It is a task for everyone who cares about democracy, public health, and human dignity. The real project is not to clean up the information system for the humans who live inside it. It is to build systems with and for those humans, so that trust has somewhere to grow.