Reviewing the Evidence on Social Media and Social Cohesion

Scholars at University of Pennsylvania's Annenberg School for Communication collect the findings of the relevant research to date.

In a speech last year at Stanford, former President Barack Obama noted that “the internet and the accompanying information revolution has been transformative,” and that the age of social media has indeed been accompanied by positive developments. But, he said, “like all advances in technology, this progress has had unintended consequences that sometimes come at a price. And in this case, we see that our new information ecosystem is turbocharging some of humanity’s worst impulses.” Obama’s speech, on one level, represented a kind of liberal consensus on the role of technology in society: a sense of tarnished promise, and of mounting peril.

Questions about the role of social media and the internet in the health of society and democracy are a core concern at Tech Policy Press. For instance, we’ve explored how platforms such as Facebook exacerbate political polarization (as well as circumstances where they may not), the degree to which social media drives partisan sorting, and why there is cause for concern over the role of digital media more generally in the decline of democracy. And we’ve looked at potential interventions that might strengthen democracy, and the role that social media could play in healing divisions. Fundamentally, the internet and social media have rapidly changed the way that the species forms relationships and shares information, producing a set of questions about the effects on collective behavior that some scholars believe should be regarded as a crisis discipline.

Now, two scholars at the University of Pennsylvania’s Annenberg School for Communication have published a new paper in the journal Social Issues and Policy Review, “Do social media undermine social cohesion? A critical review.” Referencing some of the same research studies cited above, as well as many more, Sandra González-Bailón and Yphtach Lelkes ask:

But how much do we know about social media’s role in weakening democracies and undermining social cohesion? Most popular treatments of social media and social cohesion are based on anecdotes and, at best, correlational data. Do social media merely reflect larger societal trends, or is there something specific to these platforms that creates or aggravates conflict? Do the effects of social media vary by platform or the affordances within platforms?

González-Bailón and Lelkes set out to gather and evaluate the existing empirical evidence on these questions, and “to connect this evidence with ongoing policy discussions on how to regulate social media companies.” A complex set of “sociological and psychological factors are in constant interaction in the information environment social media create,” the authors say, and so they set out to consider those sociological and psychological factors in turn.

From a sociological perspective, González-Bailón and Lelkes note that social media can affect social cohesion through changing the structure and composition of networks, influencing individual perceptions and beliefs, and affecting the access to resources and “opportunities for engagement and cooperation” between individuals and groups– in other words, by influencing “networks, information and norms.”

Networks

• Social media “can enlarge or rewire networks and create new opportunities for membership and affiliations.”

• Online platforms “allow monitoring social signals through ties that are not rooted in physical proximity,” allowing users to span geographical and social distance.

• “The formation of more bridging ties (i.e., ties spanning social distance) has the potential to increase overall levels of information diversity, which in turn can trigger several positive consequences for individuals and groups, including increasing trust, resilience, and economic opportunity.”

• Online networks “can also encourage extreme self-selection and polarization, fragmenting the fabric of public conversations and creating different information realities for different people.”

Information

• Social media networks “allow a constant pulse of information, enabling recurring exposure and the fast diffusion of content that would be invisible otherwise.”

• While these are “the type of information cascades that always preceded collective action and successful mobilizations,” in fact, social media “allows information cascades to grow faster and reach more people more quickly.”

• “The cascading effects of social media have positive implications for political participation and mobilization.”

• But it can “also have negative consequences” when it is “misinformation or incivility that are spreading and increasing their reach.”

Norms

• When it comes to norms, “platform governance and content moderation policies are intended to regulate some of the dynamics of social media through the implementation or reinforcement of norms.”

• “Researchers in this space are trying to determine which norms are successful at reducing hate, incivility, hostility, and conflict– clearly still prevalent in most online platforms– and how to ensure that those norms are maintained.”

• But, research suggests, efforts to establish new norms on social platforms “often backfire” and have unintended consequences.

• The authors then “discuss in more detail the impact of social media on these dimensions (networks, information, and norms) and the mechanisms that reinforce or undermine prosocial behavior.

González-Bailón and Lelkes find that when it comes to the sociological research on social media’s impact on social cohesion, the “stronger evidence” produced by researchers so far “is suggestive of positive outcomes that can benefit social cohesion,” but that “there is also powerful observational evidence of destructive dynamics, including the fast diffusion of misinformation, manipulation campaigns, ideological segregation and extremism.” Here, González-Bailón and Lelkes suggest more research is necessary, as well as more access to data from social media platforms, in order to arrive at better conclusions. Even better, of course, would be the ability to study these phenomena across platforms and the broader information ecosystem, including looking at “spill-over dynamics” and understanding the potential for unintended consequences of changes on one platform or one part of the ecosystem.

When it comes to the psychology of social cohesion, González-Bailón and Lelkes say this “stream of research considers the relationship between social media use and individual level indicators of ideological or affective polarization, social trust, and bias.”

In general, it appears that within the broader information environment, social media drive “dynamics that shape the activation of psychological mechanisms at the individual level, which can lead to behavior that feeds back into the aggregated network dynamics.” The authors state that “the structure and composition of networks on social media platforms” can determine” how much cross-cutting exposure users get while using those platforms,” and that network dynamic “can shape attitudes, perceptions, and beliefs through social influence and information exposure," and can change behaviors and “prompt actions that can either reinforce or undermine social cohesion.” Further results in this area will require getting past methodological hurdles, since it is very difficult to look at aggregate dynamics and individual behaviors in the same study.

Information diets and polarization

• A number of studies find that “social media is marked by homophily,” with users grouping in like-minded clusters, but “researchers have largely concluded that users do not exist in complete echo chambers.”

• The authors reference work by researchers such as Chris Bail and Petter Törnberg (both recent guests on the Tech Policy Press podcast) to get at the dynamics driving the relationship between social media and polarization, such as its effect on partisan sorting.

• Social media also contribute to other dynamics that may drive greater exposure to political content, including inadvertent exposure.

Identity formation and signaling

• Social media allows users to “cultivate and signal their social identities,” and to “form communities around shared identities.” Identity signals play a number of roles on social media, from helping people decide what information to trust to driving engagement and commercial outcomes.

• Identities can be gleaned “from ostensibly nonpolitical behavior,” such as a post about “the start of hunting season or an iInstagram photo of the poster eating Chick-Fil-A” which might signal an individual has Republican views, for instance.

• Identity signals, say the researchers, “may harm social cohesion,” since increasing “the salience of identities can increase intergroup bias”.

Hate,

incivility and outrage

• Toxic language, “seemingly ubiquitous on social media,” can “change perceptions of what is normatively acceptable and not” and can “directly influence attitudes and perceptions.”

• Social media is increasingly the barometer people use to understand social norms.

• “These norms tell us what beliefs are acceptable to hold and which we should sanction. However, this barometer does not give an unbiased account as the most vocal users are often the most extreme.” Multiple studies show that exposure to racist posts, for instance, increase further racist expression.

• Toxic speech “increases support for anti-democratic behavior, such as electoral violence.”

González-Bailón and Lelkes point out that since “well-functioning communities” are a “pre-condition for building healthier democracies,” there are a number of policy implications for both governments and platforms from this emerging body of research results.

Social media platforms have been placed at the heart of high-profile policy discussions on how to regulate user-generated content or how to curtail information campaigns designed to steer conflict and group animosity. The societal desire for more cohesion, and policy attempts to deliver on it, often run into conflict with the goals of social media companies: many of the features that make these platforms popular, and profitable, also stoke discontent. For instance, the type of content that creates misperceptions of social norms, like outrage and incivility, is often the content that is amplified by news feed algorithms. Identifying the empirical conditions and social mechanisms that work toward social cohesion and healthy forms of political disagreement is necessary to design feasible and effective policy interventions. These can take the form of regulations (including efforts to guarantee researchers can access the data necessary to find answers of societal relevance) but also of platform design in the form of specific affordances (e.g., adding friction to the ability to share content in order to prevent the fast diffusion of misinformation).

The authors identify “three main features intrinsic to platform design that affect social cohesion: social norms, algorithmic recommendations, and social feedback mechanisms (e.g., likes, reshares),” and suggest these are the areas where policymakers should focus. More broadly, they conclude that social cohesion is, in fact, an externality of the function of social media platforms and the broader attention economy, “akin to car manufacturers and pollution.” The toolkit for addressing externalities in other contexts has included taxes to discourage bad behavior, and subsidies to encourage good behavior or to provide other offsets to a given externality. The authors suggest it is necessary to look at the social costs of social media, but point out that such research may entail “high costs and requires data access and platform transparency– all of which can benefit from regulation and federal funding.” That should make data access the first priority for governments.

Certainly, such access is necessary to make progress on other research priorities that González-Bailón and Lelkes identify, such as understanding “how on-platform dynamics” relate to “off-platform behaviors,” and the role of algorithms in a complex set of feedback mechanisms. These questions seem urgent, particularly after another attack on democracy, this time in Brazil, that appears to have been driven in part by information and dynamics on social media.

And yet, one might wonder, given the perpetually changing conditions and affordances on these platforms, whether scientists will constantly have to rediscover the answers to these questions. How much of the dynamics of Twitter will remain intact, for instance, after a few more months with Elon Musk at the helm?

"It’s absolutely true that platforms will constantly change," Dr. Lelkes told me by email, "and new features will emerge that will impact social cohesion. The goal with the review was to transcend any temporary feature of any one platform and identify key mechanisms that are common across social media outlets and which may impact social cohesion. For instance, all social media outlets allow, to different extents, the formation of different types of ties and exchange of information. All social media outlets allow, each to a different degree, users to cultivate and signal identities. I am sure dynamics on Twitter and other platforms will change, as these platforms enable different policies that affect these features and mechanisms. The recent Science piece by Sinan Aral et al is a good example of the ways in which platforms can affect these mechanisms, which then has, in that case, salubrious effects."

Future experiments will rest on the science the authors have so helpfully organized, and the consensus on some of these questions to date. But while Europe is already putting in place the mechanisms to move ahead on the next wave of research and requirements for data access and transparency, similar initiatives are languishing in the U.S. Perhaps lawmakers should consider the conclusions and recommendations offered by González-Bailón and Lelkes, before the next major challenge to democracy.