Eventually You Have to Let Gary Be Wrong

At some point, nearly everyone who has spent enough time online has encountered a version of Gary.

Gary may be a real person, a distant relative, a former classmate, a stranger with a flag in his profile photo, or simply the temporary name we give to whoever has just posted something so confidently wrong that leaving it alone feels like negligence. You open the comments intending to kill thirty seconds. Ten minutes later you are searching for a primary source. Twenty minutes later you are explaining what the source actually says. Forty-five minutes later Gary has accused you of being brainwashed, three people you have never met are arguing underneath you, and you have somehow become Assistant Regional Manager of Being Right on Facebook.

We joke about this because the alternative is admitting how peculiar the arrangement is. Social platforms contain industrial quantities of questionable information, but one of the systems we rely on to push back against it is astonishingly informal: ordinary users volunteering to correct one another. No shift. No paycheck. No editor. No hazard pay. Just a person deciding that something wrong cannot be allowed to sit there unchallenged.

New research from the University of Vienna suggests there is a predictable cost to that arrangement. Simone Moser, Stephanie Bührer, Diana Kostikova and Jörg Matthes studied what happens after people correct misinformation and receive feedback from other users. Their finding is not merely that online hostility feels unpleasant. It is that hostile feedback can deplete the emotional resources of the people doing the correcting, making them less willing to do it again.

The researchers ran a cross-national experiment with 2,001 participants in Austria, Bulgaria, Denmark and Poland. Participants corrected misinformation and then received different kinds of feedback. The tone and emotional intensity of that feedback mattered. Angry, high-arousal responses produced more emotional exhaustion than civil disagreement or positive feedback.

That distinction is important because it removes an easy escape hatch. The problem was not simply that people dislike being contradicted. A healthy conversation can contain disagreement. Someone can tell you that you are wrong without turning the exchange into an emotional knife fight. What depleted participants was the hostility.

Emotional exhaustion, in turn, predicted greater avoidance of future corrective engagement. Even confidence in the correction did not provide the protection we might expect. Knowing you are right is apparently not an infinite energy source when the social cost of saying so is high.

The paper describes this as a barrier to citizen-led misinformation correction. That sounds clinical. In ordinary internet language, it means the people willing to do the homework can eventually get tired of being screamed at for bringing the homework.

There is a strange moral pressure embedded in the modern comment section. If you know something is false and you have the evidence, are you obligated to correct it? If nobody responds, does silence make the claim look more credible? If you walk away because the conversation is miserable, have you abandoned the field to whoever is most comfortable being loud and wrong?

Those questions turn ordinary participation into a tiny civic duty. The conscientious user starts to feel responsible not only for what they personally say, but for the accuracy of whatever happens to pass through their feed. A false statistic appears. Somebody supplies the correct number. A fabricated quotation circulates. Somebody tracks down the transcript. A misleading screenshot appears. Somebody finds the original context.

None of this is necessarily wasted effort. Corrections can matter. Other people are often reading even when the person being corrected refuses to move an inch. A citation left in a thread may be useful to the quiet majority rather than the loud participant who triggered it.

But the labor is still labor. It requires attention, emotional regulation, research, judgment and time. The person doing it must decide whether the claim deserves a response, find reliable evidence, explain it clearly and then absorb whatever social reaction comes back. We have built a system where a user can arrive for vacation photographs and leave having performed unpaid information maintenance for a platform valued in the billions.

The nastiest implication is not that fact-checkers become tired. It is what happens to the population that remains after they do.

Imagine two people in a comment section. One cares deeply about accuracy, checks sources and feels responsible for correcting bad information. The other does not mind posting dubious claims, repeating rumors or responding aggressively when challenged. If repeated hostility gradually convinces the first person that participation is not worth the emotional cost, nothing requires the second person to leave too.

That creates a selection problem. The people most sensitive to the quality of the conversation may also be the ones most likely to withdraw when the conversation becomes relentlessly hostile. The people least troubled by the degradation of the space can remain perfectly comfortable inside it.

Over time, that can change what the room feels like. It may become less informed not because everyone was persuaded by misinformation, but because some of the people who reliably challenged it decided they had better things to do with their lives.

If the people most willing to correct bad information eventually withdraw, the people least bothered by spreading it do not necessarily leave with them.

There is a personality trap here too. Correcting misinformation can become part of how a person understands themselves online. You are the one who checks things. You are the one who posts the source. You are the one who refuses to let nonsense slide. That identity can feel useful and even principled.

But there is no natural endpoint. The internet can manufacture one more incorrect claim faster than you can finish disproving the previous one. No amount of individual vigilance can close the queue. If your sense of responsibility expands to match the supply of things that are wrong, the job becomes infinite.

Eventually the boundary has to become personal rather than informational. Not: Is Gary wrong? He may be spectacularly wrong. The better question is: Am I the person who needs to fix this particular instance of wrongness right now?

Sometimes the answer will be yes. The claim is harmful, the audience is important, the correction is easy, or the conversation still contains enough good faith to make participation worthwhile. Sometimes the answer will be no. The thread is hostile, the same people have already rejected the evidence, the correction exists elsewhere, or you simply do not have the emotional budget for another unpaid shift.

That is not surrender. It is resource allocation.

The University of Vienna researchers point toward moderation of hostile comments as one way platforms can sustain user-led correction. Their experiment suggests why that matters. If the emotional tone surrounding correction determines whether people are willing to keep participating, then civility is not merely a cosmetic preference. It affects who remains willing to do information work.

Platforms frequently talk about community notes, user reports, crowdsourced context and other systems that invite users to help maintain the information environment. Those tools can be useful. But every system that depends on volunteer correction also depends on the volunteers having enough energy to return tomorrow.

A platform cannot reasonably celebrate citizen fact-checking while treating the hostility directed at those citizens as somebody else's interpersonal problem. If correcting falsehood predictably comes bundled with abuse, then the design has made accuracy more expensive.

And when accuracy becomes expensive, fewer people will keep paying.

There is a broader lesson here about the internet's strange economy of attention. We are constantly told to participate responsibly, verify information, correct false claims and improve the quality of public conversation. Those are admirable goals. But responsibility without boundaries turns quickly into an assignment nobody can finish.

You are allowed to care about whether the internet is accurate without personally becoming its maintenance department.

Correct the thing when the correction has a reasonable chance of helping. Leave the source for the people quietly reading. Support systems that make good information easier to find. Report abuse when the platform gives you a useful way to do it. And when the thread has become a furnace fueled entirely by strangers enjoying the heat, close the tab.

Gary may still be wrong tomorrow. The internet has more wrong answers than you have nervous system.

That is not a failure of civic responsibility. It may be the boundary that allows you to have enough energy left to participate when it actually matters.

SOURCE NOTES

• Primary research: Moser, Bührer, Kostikova & Matthes, Computers in Human Behavior, Article 109179. Available online Sept. 14, 2026; announced by the University of Vienna research group Sept. 18, 2026. DOI: 10.1016/j.chb.2026.109179.
• University record: University of Vienna research portal. The peer-reviewed paper is accepted/in press for Computers in Human Behavior, volume 187, and will appear in print in 2027.
• Method note: The study was a cross-national experiment with 2,001 participants in Austria, Bulgaria, Denmark and Poland. Its findings concern responses within the experimental social-media setting and should not be read as a claim that every online correction produces the same emotional effect.