Introduction

The internet is one of the fastest technological revolutions in human history [1]. Almost 5 billion people worldwide use social media, and the average social media user now spends about two and a half hours a day online [2]. Alas, the online environment is far from a true representation of the offline world. In this paper, we argue that social media is akin to a funhouse mirror, reflecting and warping our collective sense of what is normative [3]. When people stare into the mirror they do not see a true version of reality, but instead one that has been distorted by a small but vocal minority of extreme outliers whose opinions create illusory norms. In turn, these outliers are often amplified by design features and algorithms that prioritize engaging content.

Online discussions are dominated by a surprisingly small, extremely vocal, and non-representative minority. Research on social media has found that, while only 3 % of active accounts are toxic, they produce 33 % of all content [4]. Furthermore, 74 % of all online conflicts are started in just 1 % of communities [5], and 0.1 % of users shared 80 % of fake news [6,7]. Not only does this extreme minority stir discontent, spread misinformation, and spark outrage online, they also bias the meta-perceptions of most users who passively “lurk” online. This can lead to false polarization and pluralistic ignorance, which are linked to a number of problems including drug and alcohol use [8], intergroup hostility [9,10], and support for authoritarian regimes [11]. Furthermore, exposure to extreme content can normalize unhealthy and dangerous behavior. For example, teens exposed to extreme content related to alcohol consumption thought dangerous alcohol consumption was normative [12].

The current paper explains who social media creates a funhouse mirror version of reality., We draw from work in political science, psychology, and cognitive science to explain how online environments become saturated with false norms, who is misrepresented online, what happens when online norms deviate from offline norms, where online people are affected, and why expressions are more extreme online. We provide a framework for understanding and correcting the distortions in our social perceptions created by social media platforms. We argue the funhouse mirror nature of social media can be pernicious for individuals and society by causing pluralistic ignorance and false polarization.

Social norms are defined as the “predominant behaviors, attitudes, beliefs, and codes of conduct of a group” [13]. Norms are powerful because they are inherently tied to people’s social identity – the norms one follows signal what group they are in and strengthen their ties with their ingroup [14]. Once a norm has been established within a social group, new group members are especially quick to learn and follow new norms, and those norms are relatively impervious to reinforcement learning [15,16]. Conforming to norms demonstrates commitment to the group, whether that involves supporting group values, policing wrongdoers, or adopting specific worldviews [17, 18, 19]. When a norm is ambiguous, such as deciding whether an action is permissible, people base their decisions on the group consensus [20]. Thus, correctly and accurately detecting social norms is critical for social acceptance.

Detecting social norms can be a challenge, as it requires one to attend to the behaviors and opinions of many group members to form a model of how to behave. Thus, rather than encoding and memorizing each individual exemplar of normative behavior and opinion, people instead form an average representation of a series of exemplars in a group via the process of ensemble encoding [21,22]. Ensemble coding is cognitively efficient, allowing people to encode a single representation of a set of stimuli, rather than encoding and memorizing every item [21]. Socially, ensemble coding allows people to form a single estimation of group emotion or opinion, rather than individually encoding each person’s reaction [23,24]. Thus, one might gather information about what opinion is normative over repeated interactions with others to form an average representation of a groups’ opinion. In this way, people encode the social norms from posts and comments in online forums and social media platforms.

While ensemble coding is efficient, it can become distorted online due to the structure of the normative information. False norms emerge, in part, because social media is dominated by a small number of extreme people who post only their most extreme opinions, and do so at a very high volume–often posting dozens of times more than others [25∗∗, 26, 27], while more moderate or neutral opinions are practically invisible online. Encountering a disproportionate volume of extreme opinions can lead to false perceptions that the norms are far more extreme than they actually are.

This appears to be a general phenomenon online as it occurs across domains and platforms. For instance, online consumer reviews often reflect extremely positive or negative experiences, resulting in ratings that are either perfect or terrible, with little in between [28]. Similarly, on platforms like Instagram, there is a norm to present oneself as interesting, attractive, and successful, showing only the most extremely positive and flattering content to garner peer approval [29]. On Linkedin, people disproportionately report successes and accomplishments rather than failures [30]. This leads to a feed where rich models and highly successful people appear to be the norm and can implicitly pressure users to distort, filter, and curate their posts to fit in or gain social status.

In online political discussions, the people who post frequently on social media are often the most ideologically extreme [31,32]. Indeed, 97 % of political posts from Twitter/X come from just 10 % of the most active users on social media, meaning that about 90 % of the population’s political opinions are being represented by less than 3 % of tweets online [33]. This is a marked difference from offline polling data showing that most people are ideologically moderate, uninterested in politics, and avoid political discussions when they are able [34, 35, 36]. In discussions of the Covid-19 vaccine discussion on Twitter, only 0.35 % of people were in true echo chambers, and yet those users dominated the overall discourse [37]. Similarly, an analysis of social media 448,103 users found that a third of low-credibility posts were shared by just 10 accounts [26] (see also [38]). This renders moderate opinions effectively invisible on social media, leaving the most extreme perspectives most visible for users (see Figure 1).

Receiving such biased inputs from the online environment can lead to extremely biased outputs when people rely on ensemble coding to form representations of the opinions of the general public. This may be especially problematic for topics like politics, where opinions are invisible and people are generally hesitant to share their opinions with others in everyday life. This increases the difficulty of ascertaining true offline norms. People base their perception of norms on an unrepresentative sample of opinions and images leading to a distorted view of social norms. This may be exacerbated by the fact that people tend to weight extreme content more heavily when taking the average of a set of stimuli [23], and assume greater moral outrage from a post than the authors of posts themselves report feeling [39]. Thus, the conceptualized average of opinions for one’s ingroup and outgroup in the online world can be far more extreme than true offline norms.

Considering that online discourse is dominated by the most extreme people, it is not surprising that negativity, intergroup hostility, and polarization appear strikingly prevalent online–often exceeding exposure to similar content in the real world and all other forms of media (See Figure 2 [40, 41, 42, 43])The most widely shared content on Twitter/X and Facebook is moralized content, such as expressions of outrage and hostility towards political outgroups [40,42,44]. When positive content does go viral, it also tends to be high valence, inducing emotions like awe [45]. As such, the norms generated on some social media platforms might be more hostile than the offline world.

Social media feeds are also curated to normalize extreme norms—from unrealistic beauty standards to outrageous benchmarks for success—thereby fostering a false reality [46]. Even if people recognize that certain visible behaviors (e.g., beauty norms) do not reflect how people actually are (i.e., descriptive norm), they are still reinforced through “likes,” signaling what is socially desirable (i.e., prescriptive norm). These extreme norms can leave people feeling inadequate and constant exposure to extreme outliers in body shape on platforms like Instagram may contribute to lower body image and depression in teen girls [47].

Research directly linking distorted online norms with offline norm misperception is still in its infancy, but early evidence suggests that spillover does occur in some domains (e.g., online content that normalizes alcohol consumption can lead college students to overestimate how common drinking is offline, to endorse drinking behavior, and to engage in drinking behaviors themselves [48, 49, 50]). However, it is plausible that some users may recognize the distorted nature of online content and place less weight on this information when calibrating their understanding of norms. We therefore call for more research on the impact of online content on perceptions of real world norms (see also [51]).

Not only are these norms different from offline norms, but they also contradict social media users’ expressed preferences. While most users acknowledge that negative and extreme content is most prevalent online, the majority of users say they would prefer more positive and nuanced content (see Figure 3 [52]). This discrepancy suggests that these false norms make divisiveness seem more pervasive than it actually is–and certainly more than people want it to be. Even though incivility from politicians is increasing and is increasingly socially rewarded online [53], people report that they want to hear less from uncivil politicians [54]. So, if most people say they want positive content, what causes the norms of negativity and extremity to be so prevalent online?

Social media operates in an attention economy, where design features and algorithms are designed to elicit as much engagement as possible [3,55]. Platforms then sell ad space to companies based on indices of attention. As such, there is a strong incentive for users to create content that captures attention and maximizes engagement–rather than content that reflects reality. Given that users who are the most active on social media are also the most extreme [25,32]), this creates a perverse incentive structure to reward surprising, negative, extreme or divisive content. For instance, news stories that express outgroup animosity are 67 % more likely to be shared on social media [42]. Thus, people with more extreme or hostile beliefs tend to dominate discourse – drowning out or overshading more mild or nuanced content [31] and leading to false beliefs about the norms of a community.

These online dynamics are amplified by the design features and recommendation algorithms on various platforms [56,57]. For instance, a recent analysis of the algorithm on Twitter/X found that it prioritizes evocative content [44]. This incentivizes users to create this type of content which can help them build a large following while warping public perceptions of norms. Indeed, people who are focused on gaining social status are the most hostile online [25]. These strivers may further distort norms as they rise in status.

This is compounded due to the fact that there is often little motivation for someone to post a nuanced or moderate opinion on social media. Moreover, nuanced or moderate posts often risk hostility from more extreme ingroup and outgroup members, especially since such hostility has little cost for the aggressor due to the social distance the online environment affords [58]. Indeed, people who are politically moderate were more likely to report being harassed online, even though they were also less likely to post [31]. People who hold less extreme beliefs have less investment into arguing, and when attacked by people who have more strongly held beliefs, perceive it as more hostile. The fact that people who “troll” other people typically have higher dark triad characteristics also does not encourage nuanced debate [59].

© 2024 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.