Public Outrage Is A Signal, But What Is It Communicating?
How outrage online spreads and is perceived has implications for how we respond to it
We’re all familiar with this narrative: Public anger shows up as engagement on social platforms when someone (or a brand or institution) does something wrong or appears to have done something wrong. The crisis communicator’s job is to accurately interpret that reaction, then respond in a way that addresses the genuine grievance and helps move the situation back toward calm.
Public outrage is a signal.
Or is it?
This story is overly simplistic. The social media environment systematically distorts the expression of outrage, amplifies it through reinforcement and algorithms, causes observers to perceive far more of it than actually exists, and then feeds those distorted perceptions back into the system, generating even more distortion. Let’s look at what the research tells us about the practical implications of this for crisis communications and social media users.
The Confrontation Effect
The conventional model of online engagement assumes that people interact most with content they agree with (congeniality bias). Extensive research on this across multiple domains (politics, religion, health, etc.) shows that people gravitate toward information that supports their beliefs and avoid information that challenges them. But Mochon and Schwartz (2024) document the opposite, referred to as the confrontation effect.
The Confrontation Effect: In online settings, users are more likely to engage with content that clashes with their ideology than with content that aligns with it.
The mechanism is outrage.
When users encounter ideology-inconsistent content, they experience a compound emotional response of anger and disgust that motivates the desire to publicly counter the content. Users are more likely to click on ideology-inconsistent posts and leave comments (but not follow the poster; following is still governed by congeniality bias). It is the emotional response to ideological challenge that drives this heightened engagement, rather than curiosity, and the effect is stronger for topics perceived as more threatening to core values than it is for lower-stakes or less identity-relevant content.
So, when a brand or institution becomes the subject of controversy, although spikes in engagement metrics can be considered a proxy for widespread, genuine grievance, they may also (potentially primarily, depending on the specifics) reflect how effectively the content has activated the confrontation response in ideologically opposed audiences. A communicator who interprets high engagement as evidence of broadly held public anger may be misreading the situation.
Because this misreading can lead to apologizing for positions that don’t warrant an apology or abandoning defensible stances under the pressure of what appears to be overwhelming opposition, we need to distinguish hostile from supportive or neutral engagement before drawing conclusions about the distribution of public opinion.
The confrontation effect is likely influenced by how social media has changed the way moral outrage functions. Crockett (2017) argues that digital media inflates the triggers of moral outrage and reduces its costs. We don’t encounter moral norm violations very often in ‘real life’, but they’re everywhere on social. Algorithms promote whatever is most likely to be shared, which often means content that elicits outrage. Interestingly, Crockett also found that people get more outraged by immoral acts they see online than by those they encounter in person or through traditional media (print/TV/Radio). Social media lowers the barrier to expressing outrage because there’s less personal and social risk involved. People can do it pseudonymously and address audiences they’ve self-selected, and there’s less of a sense of discomfort around inflicting harm on another person when they’re seen as a two-dimensional icon on a screen rather than a human. For crisis communicators, we need to acknowledge that although outrage on social can reflect the public’s genuine emotional state, it doesn’t just do that. Treating it as a direct readout of public feeling misses the distortion and signal amplification that’s produced by algorithms designed to maximize outrage expression.
Platforms also amplify the rewards of expressing outrage because a single piece of content can reach millions, and outrage fuels virality. It can be habit-forming because of the positive engagement (likes, shares etc) that arrives at unpredictable times, which is the type of reinforcement that makes you more likely to repeat a behavior. As Brady et al. (2021) found in their study of social learning and moral outrage, social feedback specific to outrage expression predicts future outrage expression. This is not surprising as reinforcement learning is a natural human behavior, but on social media, the inputs for the learning process are shaped by corporate interests because the algorithms determine how many users are exposed to any given post and, thus, how much social feedback it receives.
As Brady et al. note: “newsfeed algorithms can influence users’ moral behaviors by exploiting their natural tendencies for reinforcement learning”
So even if platform designers don’t intend to amplify moral outrage, the design choices that drive engagement can indirectly do so because outrage-provoking content draws high engagement. Brady et al. (2021) also demonstrated a norm learning effect, where users learn from observing the social network; in settings where outrage expression is common, we learn that it is the expected mode of communication and then engage in the behavior ourselves, regardless of whether we personally receive reinforcement for it. When outrage is already normative, we’re less sensitive to individual social feedback and don’t need the reward to maintain the behavior. So, from a crisis management perspective, in highly polarized online communities, outrage expression may be self-sustaining and resistant to interventions that are intended to reduce the rewards for outrage because the norm itself has become the driver.
Brady et al. noted in their 2021 paper that outrage expression may be decoupled from outrage experience, where if outrage becomes habitual, this visible outrage on social may not reflect the actual emotional experience of the people producing it. Their 2023 follow-up study, which I’ve written about before, confirms this. Observers on social media systematically overperceive the moral outrage of the people they are watching. And this overperception is specific to outrage, as no comparable overperception of happiness was found, so it’s not as simple as ‘people misread emotions online’. (Unsurprisingly, the over-perception effect is stronger when we spend more time using social media to learn about politics). Because social and group identity motivations encourage people to express outrage more frequently or more intensely than they actually feel, as a signal of group affiliation and trustworthiness, posters may express more outrage than they feel, while observers perceive more outrage than the posters express.
Individual Bias Becomes Collective Misperception
Individual overperception then becomes a systematic misperception of the collective emotional climate of entire social networks. When participants were assigned to a “high-overperception” (posts whose outrage tended to be overestimated by observers) or “low-overperception” (more accurately perceived tweets) feed, those who viewed the high-overperception feed judged the collective outrage of their social network as greater than those who viewed the low-overperception feed. The collective outrage judgements in the high-overperception condition exceeded the mean of the individually perceived outrage values, meaning participants were weighting the most intense outrage expressions disproportionately when forming their collective judgement, i.e., the “crowd-emotion-amplification effect” Compared with participants who viewed the low-overperception feed, those who viewed the high-overperception feed then considered outrage expression to be more socially appropriate on that platform; they also perceived the platform to be more polarized and ideologically extreme.
Crisis communicators need to be mindful of this effect because when an organization is caught in a social media controversy, the people monitoring that controversy may be perceiving a level of public outrage that substantially exceeds what the public actually feels. The crisis looks worse than it is, not because of bad faith or incompetence, but because of systematic perceptual biases built into the social media environment itself.
Users Want Something Different From What Platforms Give Them
Rathje et al. (2023) add another factor to this picture in their study of what type of content people think will go viral on social versus what type of content they think should go viral. The participants believed that content evoking intense emotions, divisive or polarizing content, moral outrage, misinformation and conspiracy theories, negative emotional content, and content featuring people criticizing their enemies all go more viral than they should. Conversely, they believed that content evoking positive emotions, accurate information, thoughtful and nuanced content, and educational content does not go as viral as it should. And the results were nearly identical for Republicans and Democrats. Republicans expressed less concern about misinformation going viral, but the overall pattern of beliefs about what does and should go viral was consistent across both. So, the public’s dissatisfaction with the current social media content environment is not itself a partisan issue, even if specific content disputes are.
The paradox Rathje et al. identify where users engage with divisive content but state they want something different reflects how algorithms are optimized to let a small number of highly active users produce a disproportionate share of the divisive content, and also that social media use is partly driven by self-control failures and habit rather than intentional choice. The implication is that because the content environment people experience on social media doesn’t accurately represent what most people want, engagement data are a biased measure of genuine public preference.
Implications For Crisis Communications
The most fundamental implication of this these finding is that crisis communicators need to carefully consider what social media engagement data tell us. The confrontation effect means that high engagement is often driven by ideologically opposed audiences, not by a broad consensus of genuine grievance. The overperception research means that the apparent intensity of that engagement substantially overstates the actual emotional experience of the people producing it. The social learning research means that much of what looks like organic outrage is actually behavior shaped by reinforcement schedules and network norms. And the preference paradox research means that the content environment users experience may not reflect their genuine preferences.
The signal is noisy and systematically biased. It’s also moving in a predictable direction: toward making crises look worse than they are.
This does not mean that social media outrage is never genuine or doesn’t matter as we know that it can be and that it clearly does. But social media crisis signals should be approached like any other data source with known systematic biases. We need to compare social media sentiment with direct feedback, survey data, media coverage, and other indicators before drawing conclusions about the true distribution of public opinion.
Avoiding Reactive Escalation
All of the factors in these studies highlight the risk of a specific failure mode that we want to avoid in a crisis, and that is reactive escalation. Reactive escalation happens when we, perceiving a level of public outrage that substantially exceeds the actual emotional state of the public, respond in ways that are disproportionate to the actual situation. We over-apologize, make unnecessary concessions, abandon defensible positions… engage combatively with critics in ways that amplify the confrontation effect.
Misreading perception is a predictable natural human behavior that has nothing to do with the misreader’s intelligence or experience. It’s a reaction to the social media environment. The crowd-emotion-amplification effect means that the communications team monitoring a crisis may also perceive the network as more outraged than it actually is. The fact that the effect is particularly strong among heavy social media users is particularly important because these are precisely the people most likely to be working on these teams. We need to be aware of our own risk to be able to effectively assess others’.
Of course, we’re not going to ignore social media signals, and the specifics of the approach will depend on the entity involved and the situation. But having baseline measures of typical engagement levels and sentiment before a crisis occurs helps identify deviations and assess them in context. Conducting rapid direct surveys of affected stakeholders rather than relying solely on social media monitoring helps determine exactly how much of what is happening online is also reflected in the perception of the groups that are most important.
A rapid response is still important, and how you react in the initial stages of a crisis will always determine how well you recover. However, the rapid response doesn’t always have to involve a rapid public response. Building in deliberate delays before major response decisions to see how initial outrage spikes dissipate can be beneficial, as these may be artificially amplified at first by reinforcement dynamics that make the situation appear considerably worse than it actually is. I’ve had a few clients who wanted to issue an immediate public statement but, due to the specifics of their situations, saying nothing at that stage was the right move because it gave the ‘spike’ time to clear so we could look closely at the actual issue and potential damage and respond in a way that mediated that rather than amplifying (by responding to) a negative perception that appeared problematic but actually didn’t exist.
Framing And The Confrontation Effect
The confrontation effect is stronger when messages are perceived as more threatening to core values, so communications that are perceived as challenges to the values or identity of an opposing audience will generate more confrontational engagement than those framed in ways that reduce perceived threat.
In practical terms, we need to avoid language that could be read as dismissive of critics’ concerns, even when those concerns are misplaced. Dismissiveness can be perceived as threatening, so may activate the confrontation response.
Additionally, because the confrontation effect is driven by ideological inconsistency, emphasizing shared values rather than focusing on contested ones may reduce the ideological distance that triggers it.
We also want to be mindful about how we invite dialogue, as communications that invite comment will likely generate more confrontational engagement than those that don’t.
Platform Considerations
We should be cautious about allowing social media engagement metrics to drive crisis communications strategy in real time because the incentive structures of social media platforms are not aligned with the interests of organizations managing crises.
The research makes it clear that social platforms are active participants that shape the emotional content of public discourse in ways that systematically favor outrage. Understanding that the platform architecture is biased toward amplifying outrage should inform how organizations calibrate responses. A response that would be proportionate to the actual level of public concern may appear inadequate when measured against the amplified signal on social media; a response calibrated to the amplified signal may be disproportionate to the actual situation.
We need human judgment, direct stakeholder engagement, and off-platform data sources to check social media signals.
Social media signals are important, but as standalone data, they can be misleading. We need to build systematic corrections for known perceptual biases into our crisis assessment processes and framing communications in a way that reduces perceived threat rather than responding to the amplified appearance of outrage.
We need to maintain strategic clarity about the difference between the social media environment and the actual distribution of public opinion.
Social media crises ARE real. The outrage they surface is NOT always manufactured. Sometimes organizations do genuinely wrong things, and the public is genuinely angry, and social media is the channel through which that anger finds expression. I’m not saying we need to throw away existing approaches to social media crises. But we do need to ask these questions:
How much of this outrage is real? Has the company actually received direct contact from stakeholders they care about complaining about this? This separates signal from noise. Are customers calling? Are donors pulling support? Or is it Twitter and Reddit?
How much is amplified? Is it affecting relationships that matter to the company’s survival? Is it on multiple platforms or contained to one?
How much are we perceiving that isn’t even there? Are the commenters saying anything of substance that is true and damaging because it is true? Sometimes the outrage is real because the underlying claim has merit. Sometimes it’s pure amplification of something false. You need to know which before you move.
Answering these questions well and honestly gives us a good foundation for responding to the actual issue and for checking our own biases when analyzing a client’s situation.
Sources
Brady, WJ et al. How social learning amplifies moral outrage expression in online social networks. Sci. Adv 7, eabe5641(2021). https://doi.org/10.1126/sciadv.abe5641
Brady WJ et al. Overperception of moral outrage in online social networks inflates beliefs about intergroup hostility. Nat Hum Behav 7, 917–927 (2023). https://doi.org/10.1038/s41562-023-01582-0
Crockett, MJ. Moral outrage in the digital age. Nat Hum Behav 1, 769–771 (2017). https://doi.org/10.1038/s41562-017-0213-3
Mochon, D & Schwartz, J. The confrontation effect: When users engage more with ideology-inconsistent content online. Organizational Behavior and Human Decision Processes 185 (2024). https://doi.org/10.1016/j.obhdp.2024.104366
Rathje S et al. People think that social media platforms do (but should not) amplify divisive content. Perspect Psychol Sci 19(5):781-795 (2024) https://doi.org/10.1177/17456916231190392.