Nieman Foundation at Harvard
This article for Nieman Lab was adapted from a Sept. 21, 2022 blog post, “10 Factors That Shape a Rumor’s Capacity for Online Virality,” written by the Election Integrity Partnership, a nonpartisan consortium of researchers co-led by the Stanford Internet Observatory and the University of Washington’s Center for an Informed Public. The original article was co-authored by Kate Starbird, Mike Caulfield, Renee DiResta, Emma Spiro, and Madeline Jalbert.
“Rumors are the oldest form of mass media,” Jean-Noël Kapferer wrote in the 1990 book Rumors: Uses, Interpretations, and Images.
Reporters and fact-checkers are familiar with the challenges posed by rumors: They tend to be persistent. They are often entertaining. And they sometimes turn out to be true. Today, information often flies faster than facts can be known, as audiences on social networks share the claims, links, and memes that intrigue or outrage them. This is not limited to political content, of course, but it can be particularly impactful in certain domains — such as elections, where confidence in the process, and the outcome, is critical to democracy. Decades of research provide insight into how and why rumors spread, and this knowledge can help us anticipate what kinds of rumors might emerge and which rumors have the potential for virality.
At the Election Integrity Partnership, we have identified 10 factors that help determine a rumor’s potential to gain traction:
- Uncertainty/ambiguity
- Diminished trust in media and authoritative sources of information
- Significance/impact
- Familiarity/repetition
- Compellingness of evidence
- Emotional appeal
- Novelty
- Participatory potential
- Origins and amplification in the social network
- Inauthentic amplification or manipulation
Previously, we introduced a threat framework around these 10 factors and applied it to the perspective of two perspectives: (1) election officials who need to anticipate rumors that may undermine the public’s confidence in the voting process; and (2) analysts and crisis communication teams who need to assess the potential virality of an emerging rumor. For Nieman Lab, we present a condensed version of these insights specifically for journalists and news organizations assessing voting- and election-related rumors in the weeks leading into the 2022 U.S. midterm elections.
Rumors emerge and thrive under uncertainty. When people feel a sense of uncertainty about a particular topic — perhaps due to a lack of timely information — they come together to try to resolve that uncertainty, participating in what’s called “sensemaking.”
There’s a fair bit of uncertainty in an election. Many states and locales offer both day-of ballots and mail-in ballots, and it can take a few days for results to come in. Sometimes the candidate who looks as if they’re leading early ultimately loses when the final votes are counted. Delays in election results, though expected and often necessary, can catalyze the spread of rumors and provide windows of opportunity for motivated actors to exploit.
As election officials assess a rumor’s capacity for online virality in the coming weeks, journalists and news organizations should be asking many of the same questions about uncertainty:
- Is this rumor taking place during a period of high uncertainty, including times before voting results or other official announcements are released?
- Are authoritative sources of information available and timely?
- Has the rumor changed over time as new information is introduced?
- Is a particular narrative related to a part of the voting process where it is inherently difficult to identify what the “truth” is or what should be happening?
- Is the core claim unfalsifiable?
Another factor that mediates the spread of rumors is the availability of timely, quality information from trusted sources, including news organizations, government agencies, and public officials. In informational environments where the official sources are not seen as trustworthy, either due to their own failures, bad-faith efforts to undermine confidence, or a combination of the two, rumors are more likely to spread.
Journalists and news organizations covering the upcoming elections should be thinking about these questions:
- Are local and state election officials experiencing diminished trust, generally or among specific audiences?
- Have election officials made previous errors that received public criticism?
- Have errors related to this part of the election process been made in the past?
- Are there local media outlets that are trusted — or do residents predominantly rely upon national news and hyperpartisan media outlets?
- Are authoritative sources of information generally timely and accurate?
Certain contextual features of an existing or potential rumor — significance/impact and repetition/familiarity — can set it up for “success” in terms of spread. In the elections context, these factors focus attention onto processes, procedures, or specific locales that may be vulnerable to rumoring, including claims that the results would have significant impact on election outcomes or because those election elements have been the focus of prior rumors.
If the eventual outcome or facts of the matter have the potential for great impact — for example, on a close election in a swing state — then we can expect enhanced anxiety and strategic attention to be placed on those elections. Both can lead to higher levels of rumoring, and together the effect may be multiplicative. On the other hand, a rumor about a small or isolated issue with the voting process may not have viral potential, unless its adherents can convince others that it reflects a larger pattern.
Another factor is familiarity, which can be created and reinforced through repetition. In their foundational 1945 work on rumoring, Floyd Henry Allport and Milton Lepkin found that the best predictor of whether a rumor was believed was the number of times it had previously been heard. Indeed, the mere repetition of information is enough to increase its acceptance, a phenomenon termed the “illusory truth effect.”
Relatedly, many rumors have common elements with previously successful rumors. Researchers explain that new rumors often rely upon “narrative templates” (or tropes) which are recycled with novel elements as new events unfold, as Gary Alan Fine and Bill Ellis wrote in their 2010 book The Global Grapevine: Why Rumors of Terrorism, Immigration, and Trade Matter. These familiar story elements — discarded ballots, rigged machines, or dead voters casting ballots — form a personal universe of “things that have been known to happen before,” and can confer a sense of plausibility to a rumor. And plausibility is often enough: transmission of a rumor does not require full belief in the rumor, but a level of believability. People are likely to forward rumors of which they are not fully convinced if they cohere with an individual’s sense of what happens in the world and the sense of what happens in the world is determined by things which they have heard before.
For election-related rumors, we can expect to see some kinds of contagion, where a rumor that spread in one part of the country may spread again in another jurisdiction that uses similar voting procedures. This was seen in 2020, for instance, in Maricopa County, Arizona, where unfounded claims that the use of Sharpie pens on ballots used at in-person voting locations invalidated those ballots, leading to confusion and conspiracies elsewhere.
How such evidence comes to light can matter. First-person accounts can be compelling on their own, particularly where the source is seen to lack malicious motive, or the situation is imbued with a sense of immediacy. In our own work we have seen that credibility amplified through such tropes as the “accidental witness” who stumbles on malfeasance without seeking it out, therefore lacking motive. Once established, such claims can begin to echo through both organic and coordinated repetition, sometimes called “copypasta,” and in the process, develop more complex theories around them as “interpreters” shape explanations to be more convincing and aligned to community concerns, as Kapferer wrote in Rumors. In a short period of time, a piece of evidence can be introduced, framed, refined, and integrated into existing narratives in ways difficult to undo. On the other hand, if underlying evidence is easily and quickly refuted by a trusted source, that may mitigate the spread of a rumor.
As election officials think about the types of “evidence” that could be used to undermine trust in the voting process — like screenshots from websites or TV graphics showing “vote dumps,” surveillance videos of vote-counting processes, or a public-facing website where voters can check the status of their ballot — journalists and news organizations in their reporting should be diligent in assessing the “evidence” that provides the basis for the claims and asking other questions:
- Is the evidence easy to find?
- Is it compelling and memorable?
- Is there photo or video evidence?
- Is there data or statistical evidence?
- Is there a first-person account or is the claim based on a second-person or “friend of a friend” account?
- Relatedly, is that evidence difficult to refute?
- Is there a clear fact-check?
A significant part of a rumor’s engagement potential is its capacity to stimulate an emotional response. Our emotional responses can be a major factor in the sharing of rumors. Emotions can activate us to do something — and in online environments that often means engaging with content. Rumors that invoke a strong emotional response will therefore likely spread further and faster than other rumors, including those with a humorous quality. There are reputational rewards for making other people laugh, and this can motivate the sharing of rumors that are funny.
But there are darker aspects to emotional appeal. In early work on rumoring, Robert H. Knapp wrote in 1944 that rumors thrived on “wish, fear, and hostility.” Similarly, Terry Ann Knopf, in her 1975 book Rumors, Race and Riots, saw “hostile belief systems” as a prime determinant of the impact of rumor. When such hostility is present, rumors can become a tool to justify and intensify hostile beliefs by linking them to actual events.
Election-related rumors that villainize specific individuals or groups, like poll workers, judges, law enforcement officers, or members of a political party, in ways that evoke feelings of anger and/or disgust have the potential to spread widely among an “in group” that shares a particular demographic or political identity.
When assessing a rumor, journalists and news organizations should ask:
- Does this particular rumor make an explicit emotional appeal?
- Does it invoke anger, outrage, disgust, or self-righteousness?
- Does it villainize a particular individual or group?
- Do posts spreading this rumor make an explicit mention of an out-group political party?
- Alternatively, is the rumor humorous?
People can also contribute to shaping the content of a rumor as it spreads. Foundational research on rumoring describes how rumors are reshaped in their retelling through sharpening, where certain details are added or enhanced, and leveling, where some elements are removed to make the story less complex. In online environments, participation includes adding new evidence (e.g., a person sharing their own voting experiences that align with a rumor’s claims); providing interpretations of evidence (e.g., a statistician interpreting vote count data), synthesizing related rumors into broader narratives (e.g., connecting rumors about voting issues in different locations to larger claims of fraud); adapting a rumor’s core claims to conflicting evidence; and even correcting false claims. Rumors that provide easy avenues for a large number of people to participate will likely spread fast and far.
It’s also important to understand how the location of a rumor within the “social network” can play a role in how quickly and how far it spreads. Rumors spread through social networks, both online and off — and the structure of those networks shape their spread. Rumors that begin at the margins of social networks — for instance, social media accounts with small numbers of followers and therefore not well connected — may have a harder time reaching the center of the conversation than rumors that begin with influencers with large audiences that include journalists, political leaders, celebrities, and an emergent class of social media all-stars who have gained audiences primarily through their social media activity. A rumor that begins with an influencer in a central position of a network is likely to spread rapidly.
But even for rumors on the margins of a network, as they spread, they can accumulate exposure, especially through amplification by influencers. Recent research by University of Pennsylvania sociologist Damon Centola suggests that these large accounts may serve a gatekeeping function in rumor “contagion.” Influencers, for example, may see claims percolating within their communities — they are often tagged by their followers, who wish to generate lift for the claim — but must decide whether to stake their reputation on a rumor by pushing it along. In tightly connected social networks, once a rumor reaches some of those influential accounts, it can quickly echo across the entire community. And rumors will see additional bursts as they jump from one community to the next. Thus, to estimate the potential spread of a rumor, it can be valuable to measure the current spread within and across distinct communities as well as different social media platforms.
With a rumor, journalists and news organizations should ask:
- Where did this rumor originate (a social media account, website, celebrity, or elected official)?
- Is the rumor currently limited to just a few posts or statements by its original source? Or is it moving beyond that original source to other social media accounts, websites, or other sources?
- If spreading, how much engagement has the rumor received thus far? Has the rumor reached nano- or micro- influencers (social media all-stars with 5,000 to 50,000 followers) within specific social networks? Has it reached the megaphones of high-follower social media accounts or media outlets with substantial audiences? Has its spread been mostly limited to a specific community within one platform? Is it spreading widely across many communities within one platform? Is it spreading across multiple platforms and communities?
And a final factor that’s important to highlight: the role of coordinated, and often inauthentic, amplification. Social media environments are vulnerable to manipulation — both through infiltration and intentional shaping of the social networks and through gaming of their recommendations algorithms. Though social media companies have made an effort to address some of these vulnerabilities, there are still pervasive issues with manipulation, whether it be automated “bot” accounts, astroturf campaigns, copypasta, and other techniques. Rumors produced or opportunistically picked up by actors who employ these techniques or benefit from past use of these techniques to build large followings are likely to spread rapidly.
As we approach the final weeks of the 2022 midterm elections, election officials, analysts, journalists and news organizations should all be aware of these fundamental dynamics of how certain rumors and claims can spread and take hold in our information environments.
Kate Starbird is an associate professor at the University of Washington and cofounder of UW’s Center for an Informed Public (CIP). Mike Caulfield is a misinformation researcher at CIP. Renée DiResta is research manager at the Stanford Internet Observatory. Emma Spiro is an associate professor at the University of Washington. Madeline Jalbert is a postdoc at CIP. Michael Grass is the communications director at CIP.
