Conspiracy Theories Have Much in Common. Their Differences May Hold the Key to Identifying When They Turn Violent.
Despite their apparent complexity, conspiracy theories often share a common, straightforward story. Conspiracies universally posit a shadowy “they” (the government, the military, the deep state, etc.), working towards a nefarious end. This common structure is key to their appeal and appears across disparate and even conflicting theories.
The differences between theories, however, can also be illuminating, potentially allowing us to one day better identify emergent conspiracy theories and when conspiracy theories are likely to motivate real world violence.
In a study led by researchers at the RAND Corporation, the team conducted a modeling effort to analyze conspiracy theory discourse at scale, ultimately analyzing 150,000 social media posts covering four popular conspiracy theories — alien visitation, white genocide, COVID origin, and anti-vaccination.
The study was developed to determine the distinctive features of these four conspiracy theories rather than to detect conspiracy theory related posts more generally. These four theories allowed us to investigate a range of conspiracy-related discourse, including harmful and potentially violent theories, like “anti-vaccination” and “white genocide,” as well as more neutral and less obviously pernicious ones, like “alien visitation.” The inclusion of “COVID origins” further offered us insight into the distinctive aspects of nascent conspiracy theories.
We combined Google’s BERT language model, which allowed us to determine the conspiracy theory discussed, with 119 “stance” variables to create a hybrid model. These stance variables, originally developed at Carnegie Mellon University, cover a wide range of the more subjective aspects of language, including personal roles (references to an individual’s occupation or identity), public vice (references to inappropriate social behavior like injustice or unfairness), and references to popular opinion (language including phrases like “some believe,” or “many hold that”).
While BERT was able to help capture what was being discussed in a given post, the inclusion of stance variables allowed us to further determine how it was being discussed, to reveal the attitudinal and cognitive dimensions of conspiracy-related social media posts. These stance features may help us better understand the persuasive and identity-forming functions of conspiracy discourse in the future. Although we were able to accurately distinguish these specific discourses and their unique dimensions, we believe there’s still more work to be done exploring this space.
Using scalable analytics to study the rhetorical components of online conspiracy theories revealed some important commonalities, as well as key differences, across theories. For example, all of the conspiracy theories we examined feature authority sources, e.g., “the government,” or “the military.” This matched our understanding of conspiracy theories. If a person (I) believes shadowy forces are hiding the truth, it makes sense to talk about that authority source (them). It also makes sense for that individual to invoke their own countervailing authorities (“a study” they read, or their brother who is “a doctor”) and their perceived audience (we) who, like that individual, believes that something isn’t quite right.
This combination of the first person with the social closeness of “we,” while pointing to the socially distanced “them” works to establish a new reality: “they are hiding the truth.” That in turn works to develop personal conviction and commitment to contain the perceived threat: “we need to band together to fight them.”
Not all “us vs. them” discourse is antisocial
The use of social distancing — defining an outside other — in conspiracy theories commonly works to direct anger and resentment towards “them” through expressions of negative sentiment. It can only be expected when analyzing discussion of nefarious plots to find negative talk (“lying” and “worst-case” scenarios) and references to public vices (e.g., “corruption,” or “bribery”) directed towards those imagined to be behind the scheme. But linguistic differences between the conspiracy theories we analyzed produced an unexpected and important insight: not all “us vs. them” talk is antisocial.
In our data, “alien visitation” conspiracies used social distancing much more than ordinary discourse. While we expected to find social distancing language in the deeply racist, antisemitic, xenophobic, and misogynistic “white genocide” discourse, we did not anticipate this in the context of “alien visitation.”
Rather than “othering” as a mechanism to prime ill will (as with the plotters, race traitors and non-whites in the context of “white genocide”), in “alien visitation,” there is largely positive sentiment towards the socially distant “them” — the aliens travelling across the stars to visit “us.” Alien visitation discourse often displays a marked excitement about the possibility of meeting aliens. The “other” of conspiracy theories is not always threatening; they can also be intriguing and alluring.
Measuring the Intensity of Antisocial Discourse
Two of the other theories we studied, “white genocide” and “anti-vaccination,” were distinguished by their antisocial discourse, combining fear, anger and discussion of public vice with a focus on othered personal roles and identities. Despite their commonalities, the antisocial aspects of these two theories differed significantly both in kind and severity.
“Anti-vaccination” theory works in large part by contrasting fear of corporate malfeasance with concerns over public safety and the health of children. Worrying about one’s child’s or their own health is not, in itself, antisocial behavior, the antisocial aspect is higher context: coping with worries about your child’s health by undermining public health initiatives, impeding access to vaccination for others, and, as occurs on the fringes of the movement, coordinating harassment campaigns against vaccination advocates.
In contrast, those who believe in “white genocide” theory regard other human beings as an existential threat, and then dehumanize them on the basis of skin color, religion or ethnicity. Black men, in particular, were the target of fear, as well as local and international institutions like the US Government and the United Nations. This antisociality was also visible in a patriarchal toxicity surrounding personal roles: fears about “white women” having sex with “black/brown men.” In previous research we’ve found that the more clearly specified the “other” is, as it is in “white genocide” conspiracy discourse, the more likely believers are to engage in violence. Indeed, the manifestos of white supremacist lone shooters almost invariably demonstrate belief in this theory.
Identifying Emergent Conspiracy Theories
In contrast to “white genocide” and other more mature conspiracy theories, “COVID origins” conspiracies, a relatively new set of theories, notably lacked a coherent story.
Social media posters attributed the origin of COVID-19 to a wide variety of man-made causes: China, 5G towers, Bill and Melinda Gates, and the “deep state” were all blamed. Because “COVID origins” conspiracy theories were new and rapidly evolving — news of the virus first emerged nearly one month into our three month sample — adherents floated new ideas and offered provisional explanations. One such explanation, the so-called “lab leak hypothesis” that the virus escaped from a lab in the Wuhan Institute of Virology, initially rejected out of hand, has subsequently received critical re-evaluation and been deemed possible. Conspiracies do occur, and sometimes conspiracy theories are borne out by evidence.
One useful and potentially portable linguistic insight here was in the significantly greater presence of uncertainty language and the relative absence of certainty markers relative to the three other theories in our dataset (“anti-vaccination,” “white genocide,” and “alien visitation”). Discussion of “alien visitation” provides a vivid contrast. People discussing “alien visitation” opened social media posts citing “more confirmation” and asserted that “the evidence is overwhelming,” while posts questioning the official account of COVID’s origin often began with phrases like “I don’t know how accurate this is” or “there are some people who say.” Our research efforts suggest that features like uncertainty may uniquely mark nascent conspiracy theories.
While our results are provisional, focusing on the rhetorical dimensions of these four conspiracy theories has allowed us to identify key markers that may correlate with antisocial discourse and emerging conspiracy theories. If these results are generalizable, such a nuanced understanding of hate and violence within some conspiracy discourse might help platforms and practitioners design policies and prioritize interventions in the future.
Authors: Yasmin Green, Director, Research & Development at Jigsaw and Bill Marcellino, Senior behavioral scientist at the RAND Corporation.
