The COVID-19 pandemic acted as a digital pressure cooker, creating an environment where uncertainty and fear turned social media platforms like Twitter—now X—into breeding grounds for both dangerous health myths and vital scientific corrections. Amidst the chaos, we saw everything from bizarre claims about garlic water cures to dangerous misinformation regarding parasite medication and absurd conspiracy theories about vaccine-induced connectivity. While this flood of disinformation threatened public safety, it was mirrored by a parallel wave of dedicated users fighting to clear the air. A recent study from the University of Southern California sought to look past the noise, using sophisticated AI to categorize the human behavior behind these competing streams of information to understand who was spreading falsehoods and who was working to debunk them.
Led by Professor Emilio Ferrara and PhD student Eun Cheol Choi at the USC HUMANS Lab, the research team developed a custom AI tool to scan vast datasets from 2020 and 2021. By training large language models to dissect the tone, structure, and intent of these interactions, the study aimed to pull back the curtain on the “counter-misinformation ecosystem.” Their final paper, titled “Angry but Accurate,” is a deep dive into the online psychology of those who police the truth versus those who propagate deception. These findings, slated for presentation at the ACM Hypertext 2026 conference, highlight that understanding the people behind the posts is just as essential as tracking the misinformation itself if we ever hope to manage the integrity of our digital public squares.
One of the most striking findings of the study was the clear behavioral divide between the two groups. Surprisingly, those who actively combat misinformation tend to be more established digital citizens, often wielding older accounts and larger, more loyal follower bases than their counterparts. This suggests that credibility and longevity on a platform are significant tools for those trying to steer public discourse. Perhaps most counter-intuitively, the study revealed that the “truth-tellers” were actually the ones expressing more intense negative emotions. While we often assume that conspiracy theorists are the loudest and most aggressive, the researchers found that users debunking myths frequently displayed higher levels of anger, sadness, and disgust in their attempts to set the record straight.
The linguistic differences were equally telling. Users spreading misinformation relied on longer, more descriptive, and “surprising” narratives to ensnare readers, leaning into the clickbait appeal of shock value. In contrast, those fighting the tide of fake news were remarkably concise; their corrections were direct, sharp, and impatient. The intensity of their language reflects a specific kind of digital frustration—a defensive reaction to the life-threatening consequences of medical myths. This emotional profile challenges the prevailing wisdom that misinformation is always the most emotionally charged content; in reality, the anger behind the truth-seekers serves as a form of social alarm, a defensive mechanism intended to cut through the confusion of a global crisis.
This research carries heavy implications for the tech giants of Silicon Valley and their automated moderation systems. Most content moderation algorithms are designed to flag or downrank posts that exhibit high levels of “negative emotion” to keep the user experience civil. Here, the USC study highlights a systemic blind spot: if AI tools interpret anger as “toxicity,” they are inadvertently silencing the very people who are providing the most effective fact-checks. When legitimate, urgent corrections about public health are suppressed alongside aggressive misinformation simply because they sound angry or confrontational, the platform loses its ability to self-correct. Effectively, the infrastructure intended to foster positivity may be actively sabotaging the most vocal defenders of the truth.
Ultimately, the team’s work serves as a sobering reminder of why the pandemic was such a “wild time” for information science. The sheer speed at which health myths could travel,, and the tangible risks they posed to human lives, turned social media into an urgent, high-stakes battlefield. By uncovering the motives and behaviors of these two groups, researchers like Ferrara and Choi at USC are not just analyzing past tweets; they are mapping the behavioral patterns of our new digital reality. As we move forward, the challenge for platforms will be to adopt more nuanced moderation techniques that can distinguish between a malicious troll and a frustrated citizen trying to save lives, ensuring that in the future, the loudest voices of truth aren’t accidentally muted by the very systems meant to protect us.