Why Are Conspiracy Theories So Compelling?

In my last post, I described the first phase of my research, in which I am attempting to develop an empirically informed definition of ‘conspiracy theory’. In this post, I want to discuss the second focus of my research: why it is that conspiracy theories are so compelling for so many people.

Newspaper article with headline "Kennedy Slain by CIA, Mafia, Castro, LBJ, Teamsters, Freemasons"Although the specifics can be debated, it is clear that conspiracy theories are very popular. In a recent survey, 61% of participants claimed belief in some form of a conspiracy theory about the assassination of John F. Kennedy. This could possibly be attributed to increased publicity about the event due to its impending fiftieth anniversary and coverage of the release of some previously classified documents regarding it. But in an even more wide-ranging study four years ago, the number was 51%. At the very least, it looks plausible that more than half of Americans believe in this particular conspiracy theory, and there are plenty of other theories out there. For example, approximately 40% of respondents endorsed the conspiracy theory that the FDA is withholding a natural cancer cure.

Conspiracy theories are often treated dismissively as the ravings of deranged paranoiacs. Yet, we have good reason to believe that a majority of Americans believe in at least one conspiracy theory, and we can’t dismiss all of them in this way. Why, then, are conspiracy theories so compelling? There are a number of predictors for belief in conspiracy theories. The best is belief in other conspiracy theories: if someone believes one conspiracy theory, the likelihood that they believe another goes up. Other predictors are useful for predicting if a subject believes in a particular conspiracy theory, but not for the likelihood that they believe in conspiracy theories generally. Belief in conspiracy theories is common regardless of race, but white Americans are more likely than African-Americans to believe in Sandy Hook conspiracy theories (in which the government supposedly faked the Sandy Hook shooting in order to initiate more stringent gun control laws), while African-Americans are more likely than white Americans to believe that the CIA developed AIDS in order to kill African-American populations. Similarly, political liberals are more likely to endorse GMO conspiracy theories, while political conservatives are more likely to endorse climate change conspiracy theories. Evidence does suggest that people are less inclined to believe in conspiracy theories the more educated they are, but exactly why this is the case is still unclear. Higher education is correlated with a complex of many other facts and it remains to be seen whether the education itself is the cause of decreased belief.

My own suspicion is that an important part of the appeal of conspiracy theories is that we tend to find appeals to coincidence unconvincing. This is often perfectly reasonable. If a recently-elected politician installs close friends and family to all important posts, insisting that, by coincidence, their friends and family were the most qualified individuals for the posts, we will be rightly suspicious. It can be a problem, however, when this suspicion transfers over to extraordinarily complex events. For example, there is a long-standing conspiracy theory that Bill Clinton arranged for the assassination of dozens of people with whom he had varying levels of contact. An enormous part of the appeal stems from the seeming unlikelihood of so many deaths that can be linked to Clinton. Of course, a president comes into contact with a staggering number of people, and some small number of these are bound to die in a variety of ways. It is not surprising that a number of people who met Clinton died; it is merely coincidental, and what would really be surprising is if no one who he met died. When a case is sufficiently complex (such as the network of everyone a United States president meets), coincidence will often be the explanation for events.

An image titled "The Clinton Body Bags" that lists people Bill Clinton came in contact with who are now dead.There are other cases where “conspiratorial thinking,” in which we are inclined to suspect agency is the cause of an event rather than coincidence, seems appropriate. It seems appropriate that homicide detectives presume agency was involved rather than coincidence when investigating an unexpected death, and that they ask questions like “Who would benefit from this?” in determining what agency was at work. On the other hand, it seems inappropriate that a voter should presume agency rather than coincidence was involved in explaining why a former member of the president’s staff died in a plane crash, and should not ask questions like “Who would benefit from this?” in order to discover who might have arranged the disaster.

Conspiratorial thinking, utilized in the appropriate circumstances, is a powerful tool that allows us to discount appropriately explanations that are, in other circumstances, much more plausible. When applied in inappropriate circumstances, on the other hand, conspiratorial thinking can metastasize and overwhelm our rational thinking. For instance, someone nearly always benefits from any event, so that asking “Who would benefit from this?” will nearly always yield a suspect. Without compelling reason to suspect agency in the first place, it is important to refrain from asking the question. My hope is to run a series of psychological studies to see whether people who believe in conspiracy theories are also more suspicious of coincidence as an explanation in general.

In my next post, I’ll talk some about some difficulties I’ve had running the initial portion of this study, as well as talk a bit about the digital tools I’m using.

Announcing Topic Modeling – Theory & Practice Workshops

An example of text from a topic modeling project.We’re happy to announce that Scholarly Commons intern Matt Pitchford is teaching a series of two Savvy Researcher Workshops on Topic Modeling. You may be following Matt’s posts on Studying Rhetorical Responses to Terrorism on Twitter or Preparing Your Data for Topic Modeling on Commons Knowledge, and now is your chance to learn the basics from the master! The workshops  will be held on Wednesday, December 6th and Friday, December 8th. See below for more details!

Topic Modeling, Part 1: Theory

  • Wednesday, December 6th, 11am-12pm
  • 314 Main Library
  • Topic models are a computational method of identifying and grouping interrelated words in any set of texts. In this workshop we will focus on how topic models work, what kinds of academic questions topic models can help answer, what they allow researchers to see, and what they can obfuscate. This will be a conversation about topic models as a tool and method for digital humanities research. In part 2, we will actually construct some topic models using MALLET.
  • To sign up for the class, see the Savvy Researcher calendar

Topic Modeling, Part 2: Practice

  • Friday, December 8th, 11am-12pm
  • 314 Main Library
  • In this workshop, we will use MALLET, a java based package, to construct and analyze a topic model. Topic models are a computational method of identifying and grouping interrelated words in any set of text. This workshop will focus on how to correctly set up the code, understand the output of the model, and how to refine the code for best results. No experience necessary. You do not need to have attended Part I in order to attend this workshop.
  • To sign up for this class, see the Savvy Researcher calendar

What is a Conspiracy Theory?

Part of my internship at the Scholarly Commons will be a series of blog posts to describe my research and the different tools that I’ll be using to pursue it. In this first post, I’ll begin to give an account of my overall research project. Future posts will deal with other parts of the research project, what sorts of tools I will be using, the ways I’m gaining facility with those tools, and the progress of the research itself.

What Is a Conspiracy Theory?

The first phase of my research involves developing an empirically informed definition of ‘conspiracy theory’. A naive definition might be “a theory that involves a conspiracy.” This leads to many things being called conspiracy theories that would not ordinarily be understood as such. For example, the official account of 9/11 would be a conspiracy theory: Al-Qaeda, working in secret (i.e., as a conspiracy), planned and carried out the attack. While such a capacious definition of ‘conspiracy theory’ might be appealing, it runs counter to many people’s sense of what the term means.

In the philosophical literature on conspiracy theories, several definitions have been floated, but there is no agreed upon way of understanding the term. As a result, it can be difficult to know whether there is a connection between what the philosopher in question is discussing and what is commonly taken to be a conspiracy theory. In the psychological and sociological literature on conspiracy theories, much less attention is paid to questions of definition, with certain “paradigmatic” theories normally being presented as conspiracy theories. In these cases, it is reasonable to wonder if the theories presented as “paradigmatic” are actually atypical in some respects, barring some evidence that they actually are typical. In both the philosophical and psychological/sociological cases, I am concerned that choices of particular conspiracy theories might be the result of unintentional “cherry-picking” of examples, which would threaten to skew accounts.

To solve this problem, I am inspired by Paul Thagard’s study presented in “Creative Combination of Representations: Scientific Discovery and Technological Innovation,” in his collection “The Cognitive Science of Science: Explanation, Discovery, and Conceptual Change.” In that study, Thagard investigates two texts, one an anthology of important scientific discoveries, the other an anthology of important inventions. For each text, he goes through each entry, coding for the presence of certain features. This allows him to give an empirically informed account of typical features of both scientific discovery and technological innovation (specifically with regard to their use of representational combination). While there are still reasons to be wary of treating these features as characteristic (e.g., it might be that the most important scientific discoveries are actually atypical cases of scientific discovery), this is at least a good effort at moving away from cherrypicking examples.

In my own study, I have selected an anthology of various conspiracy theories. The text is “Conspiracies and Secret Societies: The Complete Dossier” by Brad and Sherry Steiger.

I have selected several features to look for in the entries. In particular, my own hypothesis is that conspiracy theories typically utilize appeals to coincidence in order to motivate their own acceptance. An appeal to coincidence occurs when a theory criticizes an alternative theory for containing an explanation that involves coincidence. For example, some 9/11 conspiracy theories observe that a number of unusual stock market behaviors with regard to the airlines involved were exhibited in the days leading up to the attack, and that this led to a great deal of profit on the part of the investors. One way to explain this would be to say it was a coincidence. The conspiracy theorists insist instead that it is evidence of insider trading among people who had knowledge of the planned attack. This substitution of conspiracy for coincidence is, I predict, typical of conspiracy theories in general.

Two lab assistants and I are working through the book and coding for the presence of the chosen features. The hope is that we will be able to make some empirically informed judgments about what features are typical of conspiracy theories. In addition to this strategy, I will utilize some text mining strategies in order to both check our own conclusions and look for other typical features we may have missed. Although the amount of text in the book is fairly small, the hope is that a meaningful topic model might be developed in order to see if the groupings that we notice ourselves emerge in the model as well. This would give us some additional evidence to be satisfied with our own coding. It could also be the case that the model could reveal certain other groupings based around features we had not coded for that we could then independently check. In the end, the hope is that we will be able to give examples of paradigmatic conspiracy theories and have some empirical backing for our choices.

In my next post, I will discuss the second component of my research project: an investigation into why conspiracy theories are so appealing to people.

Topic Modeling and the Future of Ebooks

Ebook by Daniel Sancho CC BY 2.0

This semester I’ve had the pleasure of taking a course on Issues in Scholarly Communication with Dr. Maria Bonn at the University of Illinois iSchool. While we’ve touched on a number of fascinating issues in this course, I’ve been particularly interested in JSTOR Labs’ Reimagining the Monograph Project.

This project was inspired by the observation that, while scholarly journal articles have been available in digital form for some time now, scholarly books are now just beginning to become available in this format. Nevertheless, the nature of long form arguments, that is, the kinds of arguments you find in books, differs in some important ways from the sorts of materials you’ll find in journal articles. Moreover, the ways that scholars and researchers engage with books are often different from the ways in which they interact with papers. In light of this, JSTOR Labs has spearheaded an effort to better understand the different ways that scholarly books are used, with an eye towards developing digital monographs that better suit these uses.

Topicgraph logo

In pursuit of this project, the JSTOR Labs team created Topicgraph, a tool that allows researchers to see, at a glance, what topics are covered within a monograph. Users can also navigate directly to pages that cover the topics in which they are interested. While Topicgraph is presented as a beta level tool, it provides us with a clear example of the untapped potential of digital books.

A topic graph for Suburban Urbanites

Topicgraph uses a method called topic modeling, which is used in natural language processing. Topic modeling will examine text, and then create different topics that are discussed in that text based on the terms being used. Terms that are used in proximity to one another at a frequent rate are thought to serve as an indicator that various topics are being discussed.

Users can explore Topicgraph by using JSTOR Labs’ small collection of open access scholarly books that span a number of different disciplines, or by by uploading their own PDFs for Topicgraph to analyze.

If you would like to learn how to incorporate topic modeling or other forms of text analysis into your research, contact the Scholarly Commons or visit us in the Main Library, room 306.