In the early, frightening months of the COVID-19 pandemic, when information was shifting by the hour and uncertainty filled every newsfeed, a strange thing happened: posts and videos promoting natural remedies for the virus—steam inhalation, ginger, garlic, vitamin C megadoses—began to spread like wildfire. Some were harmless; some were dangerous. And yet they were shared by people we love, by people who read the news, by people who should know better. Why? Why does misinformation, so obviously false to some, travel so quickly through our social networks? Many scholars and commentators have pointed fingers at cognitive bias, at lazy thinking, at poor media literacy. But a growing body of research suggests something more subtle and more structural is at play. It’s not that people are foolish or don’t care about the truth. It’s that the reward systems built into social media platforms—the likes, the comments, the shares, the endless notifications—are inadvertently training us to share without thinking. A new study from Gizem Ceylan of Yale SOM, with Ian Anderson and Wendy Wood of the University of Southern California, makes this case powerfully. Their research suggests that social media platforms have created an environment where the very act of sharing has become habitual, almost automatic, divorced from the content being shared. And it is those habits—not politics, not ignorance, not malice—that may be driving the spread of misinformation online.
The researchers began by designing a set of experiments to understand how different kinds of social media users engage with true and false headlines. In the first experiment, they asked participants to review eight true and eight false headlines and decide whether they would share them to a simulated Facebook feed. The researchers also asked questions designed to measure how habitually each participant used Facebook—how much time they spent on the platform, how automatically they scrolled, how often they shared content without really thinking about it. The results were striking. On average, participants shared far more true headlines than false ones, which sounds reassuring. But the pattern changed dramatically when the researchers looked at the most habitual users. These heavy users shared a much larger volume of content overall, and among them, the split between true and false headlines was nearly even: they shared 43% of the true headlines and 38% of the false ones. By contrast, lighter users shared only 15% of true headlines and just 6% of false headlines. That means the heaviest users were not only more active; they were also far less discriminating. In fact, Ceylan and her colleagues calculated that the 15% most habitual Facebook users were responsible for 37% of all the false headlines shared in the study. This is a striking finding: a relatively small group of people, acting almost unconsciously, can have an outsized and disproportionate impact on the entire information ecosystem. They are not necessarily the loudest or most ideological voices in the room. They are simply the ones who have been trained by the platform to share, share, share.
But why exactly do these heavy users share so much misinformation? Is it because the false headlines align with their political beliefs? That’s the natural assumption, and it’s what prior research has often concluded. Yet Ceylan’s second experiment tells a different story. In this round, the researchers asked participants to consider headlines with clear partisan bias—some liberal, some conservative. They then looked at “concordant” sharing, meaning liberals sharing liberal headlines and conservatives sharing conservative headlines, as well as “discordant” sharing, meaning liberals sharing conservative headlines or conservatives sharing liberal headlines. If political ideology were the driving force, habitual users should show a strong preference for headlines that confirm their worldview. Instead, the researchers found something surprising. Less frequent users showed exactly the expected pattern: they strongly preferred to share concordant headlines, reinforcing their own political identities. But among the heaviest users, this preference largely disappeared. They shared more of everything, and their bias for concordant information was significantly weaker. They would share headlines from the other side just as readily as ones that matched their own beliefs. For Ceylan, this is powerful evidence that habit—not ideology—was driving their behavior. When people have been conditioned to share automatically, the content itself becomes almost irrelevant. The platform has taught them that sharing brings rewards in the form of engagement, and so they share whatever crosses their path, regardless of whether it’s true, false, liberal, or conservative. The authors call this “a shocker” for the misinformation research community, because it challenges the comfortable idea that we can solve this problem simply by correcting people’s beliefs or improving their media literacy.
This is not to say that heavy social media users are lazy or that they don’t care about the truth at all. In fact, Ceylan is careful to emphasize that they are not. The problem is deeper and more systemic. Human beings are creatures of habit. When we perform an action and receive a reward, our brains encode a connection between the cue, the behavior, and the reward. Over time, that connection becomes automatic, so the next time we see a cue—a compelling post, a notification, a buzzing phone—we perform the behavior without deliberate thought. Social media platforms are engineered to exploit this process. Every like, comment, and re-share is a small, sporadic reward, and intermittent rewards are the most powerful way to build habits. The platforms’ goal is engagement, not accuracy, so they reward any engagement at all. As the researchers explain, the content that gets attention becomes part of habitual users’ mental representations—their internal model of what is worth sharing. They learn not because they are trying to learn, but because the platform repeatedly tells them, in a thousand tiny ways, that this is what gets rewarded. Eventually, they just share content that fits that mental representation. It feels natural, almost reflexive. But it is not natural. It is a learned response, built by the architecture of the platform itself. This reframes the misinformation problem entirely. It is not an individual problem of bad actors or foolish people. It is a system problem, one that requires us to look at the environments we have created and ask whether they are helping people make good decisions or quietly training them to make bad ones.
The good news is that this research also points a way out of the trap. In their final experiment, Ceylan and her colleagues simulated a different kind of social media platform—one that rewarded people for accuracy. Instead of earning likes and comments for engagement, participants were given points for sharing accurate information, and those points could be redeemed for an Amazon gift card. The impact was dramatic. When accuracy was rewarded, everyone shared lots of true headlines and very few false ones. The participants’ previous social media habits no longer mattered. Heavy users and light users alike adapted quickly, learning that accuracy was the key to reward. Even more striking, the change persisted. After the researchers removed the accuracy reward, the participants continued to share accurate headlines at a higher rate than before. They had begun to form a new habit—a positive one. This suggests that human beings are not doomed to share misinformation. We are highly responsive to the structures around us. If those structures reward truth, we learn to share truth. If they reward attention-grabbing sensationalism, we learn to share that instead. The same psychological mechanism that makes us vulnerable to misinformation can be used to make us more resistant to it. The problem is not that our brains are broken. The problem is that the reward structures guiding us are broken, and they can be fixed.
This is the heart of Ceylan’s conclusion: it’s time to stop blaming individual users for the spread of misinformation and start looking at the systems we have built. Social media platforms have made a choice, consciously or not, to reward engagement above all else. They profit from time spent, from clicks, from outrage, from the endless sharing of content whether it is true or false. In doing so, they have created millions of habitual sharers who click and post without thinking, because that is what they have been trained to do. If we are serious about addressing misinformation, we cannot simply tell people to think harder or take a media literacy course. We have to create better environments. We have to change the incentive structures. Platforms could reward users for accuracy, as the experiment did. They could label reliable sources more prominently, reduce the amplification of sensational content, or design algorithms that prioritize quality over engagement. They could even create new systems where sharing accurate information earns visibility, while sharing false information carries a cost. These are not impossible changes; they are choices. As Ceylan says, “We cannot keep on blaming users for showing political biases or being lazy for the misinformation problem. We have to change the reward structure on these platforms.” It’s a humbling message, but also a hopeful one. It means that misinformation isn’t an inevitable feature of human nature. It is a consequence of the environments we create. And if we have the courage to change those environments, we can also change the habits they build—one share, one like, one thoughtful moment at a time.