Why We Should Be Suspicious of Conspiracy Theories: A Novel Demarcation Problem

Abstract

What, if anything, is wrong with conspiracy theories (CTs)? A conspiracy refers to a group of people acting in secret to achieve some nefarious goal. Given that the pages of history are full of such plots, however, why are CTs often regarded with suspicion and even disdain? According to “particularism,” the currently dominant view among philosophers, each CT should be evaluated on its own merits and the negative reputation of CTs as a class is wholly undeserved. In this paper, I defend a moderate version of “generalism,” the view that there is indeed something prima facie suspicious about CTs, properly defined, and that they suffer from common epistemic defects. To demarcate legitimate theorizing about real-life conspiracies from “mere conspiracy theories” (in the pejorative sense), I draw on a deep asymmetry between causes and effects in the natural world. Because of their extreme resilience to counterevidence, CTs can be seen as the epistemological equivalent of black holes, in which unwary truth-seekers are drawn, never to escape. Finally, by presenting a generic “recipe” for generating novel CTs around any given event, regardless of the circumstances and the available evidence, I rescue the intuitions beneath colloquial phrases like “That's just a conspiracy theory.”

Just because you're paranoid doesn't mean they aren't after you.

(Yossarian in Joseph Heller's Catch 22)

1. Introduction

In polite company, it is frowned upon to profess belief in certain so-called “conspiracy theories” (CTs). Politicians, pundits, journalists, and scientists widely employ the concept as a term of abuse. If something is branded as a conspiracy theory, that is sufficient reason to regard it with a healthy dose of skepticism, if not to dismiss it out of hand. This negative connotation is widespread in the media and in the public arena more generally (Napolitano and Reuter 2021). And yet, few people would deny that “conspiracies” – in the sense of secret plots hatched by a small group of people to achieve some nefarious goal – abound in history. It is not for nothing that “criminal conspiracy” is a category of law in many legal systems. So where does the strong presumption against CTs come from? And where exactly should we draw the line between legitimate hypotheses about conspiracies and unfounded CTs? Or is the bad reputation of CTs wholly undeserved?

Perhaps surprisingly, the majority view among philosophers now holds that, despite their unsavory reputation, there is nothing epistemically suspect about CTs (Pigden 1995; Coady 2007; Basham 2018; Dentith 2018b; Pauly 2020). Indeed, some philosophers have expressed rather strong opinions on the matter. According to Pigden, the belief that there is something suspicious about CTs is “one of the most dangerous and idiotic superstitions to disgrace our political culture” (Pigden 2006: 139). In Dentith's recent edited volume, which makes a rousing case for taking CTs seriously, they and Lee Basham talk about an “anti-conspiracy theory panic,” which is “not only anti-rational and non-historical” but also “unethical and foolish” (Dentith and Basham 2018: 91).¹ In this paper, however, I feel obliged to revisit this generalized suspicion towards CTs. In particular, I intend to rescue the widespread intuitions undergirding the pejorative meaning of “conspiracy theory” and its derogatory use in the public arena. Even though conspiracies really do occur and a blanket dismissal of all explanations involving conspiracies would be as irrational as the must lurid conspiratorial fantasy, I hope to show that a prima facie suspicion of CTs, suitably defined, is indeed epistemically justified.

The structure of this paper is as follows. First, I examine different conceptual strategies for defining “conspiracy theory” (narrow vs. broad, neutral vs. pejorative), drawing some instructive parallels with the traditional demarcation problem (sections 2.1 and 2.2). Next, drawing on a deep asymmetry between causes and effects in the natural world, I try to pinpoint exactly where legitimate theorizing about possible conspiracies ends and where we enter the realm of deficient CTs (3.1 and 3.2). In section 3.3, I analyze CTs as the epistemic equivalent of a black hole, in which unwary truth-seekers are swallowed up, never to be seen again. Building further on these points, I present a simple and generic recipe for fabricating novel CTs about any given historical event (4.1). This recipe shows that, unlike legitimate accounts of genuine conspiracies, CTs are radically underdetermined by any available evidence (4.2). Finally, I show how my analysis undergirds the deeply evaluative meaning of “CT” implicit in common phrases like “That's just a conspiracy” and “I'm not a conspiracy theorist, but …” (4.3).

2. Definitions of CT

2.1. Neutral vs. pejorative definitions

In the philosophical literature around CTs, the distinction between “generalists” and “particularists” has gained wide currency (Buenting and Taylor 2010; Dentith 2018a, 2019, 2022; Harris 2018; Keeley 2018). In Buenting and Taylor's (2010: 568) original definition, generalists believe that “the rationality of conspiracy theories can be assessed without considering particular conspiracy theories” and that “conspiratorial thinking qua conspiracy thinking is itself irrational.” Conversely, particularists argue that “assessing the rationality of a conspiracy theory should be done on a case-by-case basis” (2010: 570). Since there are no general reasons to discard or distrust CTs as a class, particularists believe that their bad reputation is wholly undeserved. Rather than adopting a uniformly negative (or positive) attitude towards CTs, we have to evaluate them on their individual merits (Dentith 2018b).

Particularists tend to favor a neutral and broad definition of CT: a conspiracy theory is simply a theoretical account of history or current events that somehow involves a conspiracy. For example, according to Keeley (2018: 423), a conspiracy theory denotes “a proposed explanation of some historical event (or events) in terms of the significant causal agency of a relatively small group of persons – the conspirators – acting in secret.” Perhaps the only descriptive element missing from this definition is that those who engage in a conspiracy usually (though not invariably) have criminal or nefarious intentions.² Defined as such, the first and obvious thing to note about conspiracies is that they are all too real. Being social primates, we occasionally form secret and hostile coalitions against other persons or groups, or against the public interest at large. The pages of history are replete with such episodes. The murder of Julius Caesar was the result of a successful conspiracy, as was the October Revolution in 1917 in Russia and the terrorist plot of 9/11. Indeed, the majority of government putsches, revolutions, political assassinations, and terrorist attacks in the historical record qualify as “conspiracies” under any reasonable definition. If we understand “conspiracy theory” to simply refer to a historical account of just such an episode, then no sensible person would deny that it is rational to believe some CTs. As Cassam (2019: 3) writes, “in that sense we are all conspiracy theorists.”

However, as Coady (2003) and others have noted, this is not how the term CT is typically used in everyday life and in political discourse. Most historians would deny that these accounts of conspiracies are “conspiracy theories” precisely because they believe them to be true, confirmed, or substantiated. When using the term CT, academics commonly refer to a category of unofficial or unconfirmed theories that are rejected or ignored by reputable historians, journalists, and other authorities. For example, this includes the theories that the moon landing never happened but was faked in a Hollywood studio, that Lee Harvey Oswald was just a patsy of a larger conspiracy against JFK, that 9/11 was an inside job carried out by the Bush administration, or that the coronavirus Sars-Cov-2, which unleashed the 2020 pandemic, was deliberately designed as a bio-weapon in a lab.

Some scholars have suggested that this disparaging use of “CT” is confined to academic philosophy or to cultural elites more generally. For example, Basham writes that the term “possesses no negative connotation except as a residue among certain academic, media and political elites” (Basham 2018: 40; see also Dentith and Basham 2018). However, recent experimental evidence suggests that this negative valence also permeates everyday use of the term (Napolitano and Reuter 2021; Douglas et al. 2022). In common parlance, the label “conspiracy theory” seems to reflect prior disbelief of the speaker in the explanation or hypothesis in question. In other words, people press the label into service to disparage conspiratorial beliefs they don't endorse, while rejecting it for beliefs they accept.³ For instance, people who believe that the moon landing was faked would not describe this account as a CT precisely because they believe it to be true, or at least credible enough to be taken seriously. Indeed, when expressing their beliefs, people will often use disclaimers such as “I'm not a conspiracy theorist, but …” or “This is not just a conspiracy theory,” in an attempt to preempt accusations to that effect (Husting and Orr 2007). By using such popular tropes, as Harambam and Aupers (2017: 126) write, “they actively resist their stigmatization as ‘conspiracy theorists’ by distinguishing themselves from the mainstream as ‘critical freethinkers’.”

Even those people widely regarded as conspiracy theorists often seem to endorse the pejorative meaning of “CT” – they just quarrel about where to draw the line. As experimental philosophers Napolitano and Reuter have recently concluded in a study on CTs, the concept is deeply evaluative because it “encod[es] information about epistemic deficiency” (Napolitano and Reuter 2021: 1). For instance, a recent French documentary exposing the (alleged) hidden agenda of medical experts and politicians during the COVID-19 pandemic bears the title “Ceci n'est pas un complot.”⁴ While the documentary makers go to great lengths to disavow the term “CT” for their own views, they have no qualms about wielding it to disparage other, more extreme theories surrounding the pandemic (e.g., beliefs about QAnon and the New World Order). Such a strategy “reclaim[s] rationality by labelling others within the conspiracy milieu the ‘real’ conspiracy theorists” (Harambam and Aupers 2017: 113).

To approximate the everyday notion of CTs, some CT scholars have baked “unofficial-ness” into the definition of CTs, distinguishing them from genuine conspiracies that are acknowledged by mainstream historians and journalists (e.g., Levy 2007; Uscinski 2020). As a first approximation, this definition indeed holds water. As a demarcation criterion between good and bad theories, however, this sociological criterion is hardly satisfying. To understand why, it is instructive to draw some parallels with the traditional demarcation problem in the philosophy of science, i.e., the challenge of separating genuine science from pseudoscience (Pigliucci and Boudry 2013). Even more so than with “conspiracy theory,” the word “pseudoscience” is widely recognized as a term of abuse. If perhaps some conspiracy buffs have appropriated the label “conspiracy theorist” as a badge of honor, no one has proudly proclaimed to be a “pseudoscientist.” As a first approximation, a pseudoscience can be defined as a theory or activity that aspires to scientific status but that is widely rejected by the scientific community. Of course, such a sociological criterion is not a genuine solution to the demarcation problem or else philosophers would have declared victory long ago. Indeed, it appears question-begging: we want to find out why the scientific community regards certain theories and doctrines as so defective as to be outright “pseudoscience.” Are they doing so for a variety of reasons and on a purely case-by-case basis, or are they relying on certain general criteria (Fasce 2020; Boudry 2022)?

In an analogous way, the sociological definition of CTs merely postpones the answer to the epistemological question of interest: why is it that our experts accept some conspiratorial accounts of history, while rejecting others as “mere conspiracy theories”? Furthermore, could it conceivably happen that those authorities are in error, and how do we determine if this has occurred? Strictly speaking, as Dentith (2018a: 101) has argued, “the fact some theory has official status tells us nothing about its epistemic merits.” As with the old demarcation problem, the sociological criterion appears to beg the epistemological question. For instance, if the main thrust of the “inside job” hypothesis about 9/11 is that the official authorities are lying to us and covering up the true identity of the perpetrators, it won't do to just criticize this view on the grounds that it has been rejected by official authorities. That is precisely the crux of the CT!⁵ This point generalizes to most popular CTs: the claim that official authorities are covering up the true nature of events is often an intrinsic part of the theory, so merely citing those authorities in support of a non-conspiratorial view seems to beg the question at hand.

2.2. Reframing particularism vs. generalism

If one adopts a neutral definition of CT, it seems obvious that generalism must be false. Indeed, given the ubiquity of documented and uncontroversial conspiracies in the historical record, it is doubtful whether any sensible person would embrace such a construal of “generalism” after even minimal reflection. All parties agree that some conspiracies are real and that it is rational to believe in them. A more interesting question is whether the class of bad or unfounded CTs suffer from general epistemic defects, which go beyond the particulars of each individual theory. Is there a general way to diagnose and spot bad CTs, or should we evaluate them on a purely case-by-case basis?

On this construal, the distinction between particularism and generalism mirrors the philosophical controversy around the old demarcation problem (Pigliucci and Boudry 2013). Generalists such as Karl Popper claimed that all pseudosciences share similar characteristics (e.g., unfalsifiability) and can be diagnosed on general grounds. By contrast, particularists such as Larry Laudan believe that bad theories come in all forms and shapes and that there are no short-cuts to evaluate and dismiss them as a class (Laudan 1983). If we adopt this approach, particularism and generalism emerge as opposite ends of a spectrum, rather than as a strict dichotomy.⁶ A moderate particularist need not commit herself to the view that every CT is radically different from the next one and that no two CTs are alike. She may admit that there are some recurring patterns and tropes in the class of CTs (e.g., Dentith 2018b; Pigden 2018), but warn against hasty generalizations and insist on evaluating each CT on its individual merits.⁷ Conversely, a moderate generalist need not commit herself to the view that there is a single and perfectly general silver bullet to demolish all bad CTs. She may admit that we should be somewhat attentive to the particulars, but insist that at bottom virtually all bad CTs suffer from the same epistemic defects, and that if we spot the relevant symptoms, our suspicion will be justified.

In what follows, I will defend such a moderate generalism, pushing back against non-trivial versions of particularism. Even though it would be (obviously) irrational to reject all hypotheses involving a conspiracy, I will argue that the category of beliefs that are widely regarded as CTs suffer from similar epistemic defects and display recurring symptoms (to wit, the postulation of preternaturally smart conspirators and the ensuing radical underdetermination by evidence). As a rule of thumb, any theory exhibiting these tell-tale signs deserves prima facie suspicion, and is most likely false.

Epistemically dubious CTs, I will argue, are not just a miscellaneous bunch of theories with nothing in common except for the fact that they are unwarranted. They are birds of a feather. In the rest of this paper, I will therefore adopt the narrow and pejorative definition of “CT,” because it is closest to the folk notion, and I will try to demonstrate that this label is indeed appropriate for (at least) the most popular and common such theories. By identifying the recurring flaws of CTs (section 3) and connecting this diagnosis to the proliferation and persistence of bad CTs (section 4), we can rescue the intuition behind phrases like “That's just another conspiracy theory.”

3. Demarcation of CTs

3.1. Asymmetry of causes and effects

A striking feature of hypotheses centering around a conspiracy, which has been noted by a number of authors (Keeley 1999; Lewandowsky et al. 2015), is that they lead us to expect an absence of evidence in their favor, or even the discovery of counterevidence. If people with nefarious intentions are conspiring behind the scenes, we would reasonably expect them to try to cover up the evidence for their schemes. To the extent they have been successful in this effort, we will not readily find any positive evidence for the conspiracy. Stronger still, CTs may give us some reason to expect evidence contradicting the conspiracy. After all, one strategy the alleged conspirators could use to throw us off the scent is to fabricate evidence for a rival account that is either non-conspiratorial or involves a different group of conspirators.

To make sense of this feature, I would like to draw attention to a deep asymmetry between causes and effects in the natural world, which was first described by Lewis (1979): effects almost always overdetermine their causes, in the sense that a single past event “radiates” into the future, leaving multiple and widespread effects, all bearing the fingerprints of the originating cause. Whereas events that are overdetermined by earlier causes are relatively rare (e.g., death by firing squad), the opposite is very common: “Extreme overdetermination of earlier affairs by later ones … may well be more or less universal at a world like ours” (Lewis 1979: 474). According to Cleland (2002), all historical sciences exploit this fundamental causal asymmetry. An event in the past leaves a multitude of independent traces in the present, any small subset of which is sufficient to infer that the event has taken place. Of course, this does not mean that it is always possible in practice to infer a past cause from its present effects: the causing events may simply have happened too long ago, it may have been too subtle and localized, and its effects may have slowly dissipated and disintegrated over time. Still, the asymmetry principle gives historians, archeologists, and paleontologists a huge advantage: with diligent investigation, it is possible to reconstruct even very distant causes, using only the smallest fraction of subsequent effects.

In a paper that builds on Lewis's thesis about the asymmetry of determination, Cleland (2002: 487) explains why it is so difficult to commit the perfect crime, in which the perpetrator manages to erase all the effects of his actions, making subsequent exposure impossible:

Footprints, fingerprints, particles of skin, disturbed dust, and light waves radiating outward into space must be eliminated. Moreover, it isn't enough to eliminate just a few of these traces. Anything you miss might be discovered by a Sherlock Holmes and used to convict you. Finally, each trace must be independently erased. You cannot eliminate the footprints by removing particles of skin or, for that matter, one footprint by removing a different one.

Consider now conspiracies, which are complex criminal or covert activities involving (by definition) coordination and planning between several people over a certain span of time. At first blush, any such events must leave a huge multitude of traces, radiating in many different directions. A successful cover-up of a conspiracy, in which the conspirators manage to erase all their traces, is therefore significantly more complicated than for a single person to commit the perfect murder. This is not to say, of course, that it is impossible. If enough time has elapsed before the start of our investigation, the traces may have become simply too attenuated and far-flung to allow investigators to infer the existence of a conspiracy. Moreover, even an imperfect crime can remain undetected if there are simply no diligent investigators motivated to solve it.

It is significant, however, that even the most powerful organizations in recent history, such as the U.S. intelligence agencies and the Soviet secret police, have failed to keep the lid on some very embarrassing conspiracies, despite their best efforts at concealment. In a paper calling on historians to take genuine conspiracies more seriously, Bale (2007: 59) argues that, in several documented cases, historians have managed to unearth previously unknown conspiracies by “a fortuitous combination of human blunders, factional infighting that generates information leaks and the onset of unanticipated historical events.” None of these fortuitous leaks can be guaranteed to happen in any given case, but the larger and more complicated the conspiracy, and the more stages of preparation and execution are required, the more opportunities for investigators seeking to unveil it. This, in essence, follows from the asymmetry of overdetermination between causes and effects.

Estimating the probability of any given conspiracy being discovered obviously suffers from a selection bias: we only know about the conspiracies that failed. However, if we look at some of the uncontroversial conspiracies acknowledged by historians and government authorities (e.g., the October Revolution, the Watergate affair, and the 9/11 plot by Al-Qaeda), it is remarkable that they are supported by a surfeit of different traces, rather than by a single smoking gun or a small number of tantalizing clues.

Consider the 9/11 attacks, for which we have both an “official” conspiracy explanation (centered around Al-Qaeda) and an unofficial CT, which alleges an inside job carried out by the U.S. government and/or its associates. Even in the run-up to the attacks, according to the official version, Al-Qaeda barely managed to avoid detection. By the start of the new millennium, intelligence agencies were well aware of the existence of the organization and its intention to carry out terrorist attacks against U.S. targets, as evidenced for example by the bombing of the World Trade Center in 1993 and the bombing of the U.S. embassy in Nairobi in 1998. Well before the attacks, as Wright (2006) has documented, and despite Al-Qaeda's careful attempts to erase all traces, evidence of the planning and preparation was “radiating” in all directions and had already reached the two main intelligence agencies of the U.S. federal government. Indeed, as Wright's account convincingly demonstrates, between the two of them the CIA and FBI had amassed enough intelligence to put the pieces of the puzzle together and figure out what exactly Al-Qaeda was up to. The main reason why they failed to uncover and foil the plot was that both agencies were involved in a bureaucratic struggle for power, working at cross-purposes and refusing to share crucial pieces of information (Wright 2006). In addition, both intelligence agencies were flooded with so much information that the “signals” were lost in the noise.

But the signals were multifold and had always been there. In the aftermath of the attacks, and with the benefit of hindsight, the copious evidence for the plot immediately jumped out. Once the CIA knew what to look for, it became frustratingly obvious who was behind the attacks and how exactly they had gone about staging them. Indeed, on the evening of the day of the attacks (around 9:30 pm EST), George Tenet, the then director of the CIA, could already inform the president that the Counterterrorist Center had gathered enough evidence to establish that Al-Qaeda was behind the attacks (Jacobsen 2019: 294). In the following days and weeks, the identity of every hijacker was established and linked to the organization. Ever since, this conclusion has become “overdetermined” by ever more evidence. Every stage of the plot (planning, preparation, and execution) has been meticulously documented based on many different sources, all of which reinforce the same conclusion about the identity of the perpetrators and the nature of their plot (Wright 2006). By the time Osama Bin Laden officially claimed responsibility for the attacks in 2004, it was already widely accepted knowledge that he masterminded the attack, so his confession was superfluous.⁸ Because of the asymmetry between causes and effects, an elaborate scheme such as this one simply left too many traces to remain hidden. Even if we possessed only a minute fraction of the available evidence, it would still be sufficient to figure out who was behind the 9/11 attacks.

In the case of the unofficial CT about 9/11, however, such direct and positive evidence for the alleged conspiracy is entirely absent (Cassam 2019), even though thousands of people are highly motivated to find it. Defenders of the unofficial CT point to various anomalies and alleged contradictions in the official story, as well as a variety of unexplained eyewitness accounts and tantalizing clues (e.g., Griffin 2008); however, there is no smoking gun incriminating the U.S. government in the form of documents, email conversations, or phone records that attest to preparation for the plot. Nor are there any testimonies or confessions from any of the accomplices. On the face of it, however, the principle of asymmetry teaches us that this is extremely unlikely. Even if we take into account that the U.S. government is a far more powerful organization than Al Qaeda, having more resources at its disposal to cover up evidence and prevent information leaks, we should also consider that the plot envisaged by the unofficial CT is far more elaborate and complicated than the official plot involving Al-Qaeda.

For instance, according to the most popular CT account, the Twin Towers were brought down by controlled demolition rather than by the impact of the planes and the ensuing fires. According to demolition experts, however, preparing a large building for controlled demolition is a very complicated and laborious procedure that takes weeks or even months. Evidence for such a complicated string of events, requiring the cooperation of at least dozens or even hundreds of people, would be expected to radiate in all directions, proffering a wide array of incriminating evidence. However, the unofficial CT requires us to believe that these huge buildings had been prepared for demolition in the preceding weeks without a single employee or visitor noticing anything suspicious and without a single person spilling the beans afterward. Even the U.S. government, or any of its agencies, would be impotent to cover up so much evidence, and to silence so many possible defectors.

3.2. Explaining absence of evidence

According to a well-known dictum, absence of evidence does not constitute evidence of absence. From a probabilistic point of view, however, this is not entirely correct. Absence of evidence does constitute evidence of absence when the prior likelihood of finding such evidence (given the hypothesis under investigation) is high, in conformance with Bayes’ theorem (Pigliucci and Boudry 2014). The reason why the missing evidence of controlled demolition undermines the “inside job” hypothesis is that the asymmetry criterion gives us a very strong reason to expect such evidence, conditional on the actual occurrence of controlled demolition.

The example of 9/11 is clear-cut and – in this author's view – leaves no room for reasonable doubt: the official conspiracy hypothesis is the only one that is even remotely plausible. However, not all alleged conspiracies require the level of complexity and scale that is necessitated by the “controlled demolition” CT about 9/11. CTs about the assassination of JFK, for example, may be comparatively small in scale and complexity: if JFK was assassinated on the orders of the Mafia or Fidel Castro, the plot could have been planned and executed by only a small number of people.

Even these conspiracy hypotheses, however, eventually run afoul of the asymmetry principle, and degenerate into baseless CTs (Clarke 2002, 2007). To see this, we have to add a temporal dimension to our demarcation criterion. In his seminal paper on CTs, Keeley (1999: 56) suggested that the problem of unwarranted CTs is the “increasing degree of skepticism” required by them. This point can be cashed out with the help of the asymmetry principle. In the immediate aftermath of an event, it is often entirely reasonable to consider different working hypotheses, including those involving conspiracies. Though it is doubtful whether the controlled demolition theory about 9/11 was ever a plausible candidate, even right after the attacks took place, this is different for many other events. In the early stages of investigation, one should indeed abide by the dictum that absence of evidence does not constitute evidence of absence. Evidence may be absent for the simple reason that no-one has properly searched for it yet. Even when the investigation is underway, it is rational for an investigator not to be too quickly deterred by an apparent absence of evidence. If the conspirators have been trying to erase their tracks, then it is hardly surprising that the evidence doesn't scream in our face. Rejecting any suggestion of conspiracy at such an early point can be as unreasonable as immediately jumping to the conclusion of a conspiracy. This is why some conspiracy accounts that were initially rejected by epistemic authorities – and indeed were sometimes branded as “mere conspiracy theories” in the pejorative sense – later ended up becoming part of textbook history (e.g., the Watergate scandal or the Iran-Contra Affair).

As time wears on, however, the opportunities for finding positive evidence multiply and the probability of not finding any such evidence, under the assumption of a conspiracy, dwindles. As the investigation proceeds and yet fails to yield evidence for a conspiracy, the principle of asymmetry leads us into a dilemma. Either we accept that, given the hypothesis of a genuine conspiracy, an honest and diligent investigation would have discovered positive evidence by now, and hence absence of evidence indeed translates into evidence of absence; or alternatively, we conclude that there has not been a diligent and honest investigation because the investigators were themselves involved in the cover-up. However, this would imply an even larger conspiracy than we initially assumed, which – by the principle of asymmetry – would have radiated even more effects for others to uncover. The fact that we do not find that evidence requires us to posit an even larger conspiracy, and so on. In other words, the principle of asymmetry leads to a vicious regress, which requires us to (1) attribute ever more intelligence and power to the alleged conspirators and (2) keep widening the circle of presumed conspirators.

Even though it is true, as some particularists have argued (Hagen 2018), that CTs are not necessarily “vast” and do not necessarily postulate “preternaturally smart” conspirators, they have a tendency to end up doing so as time goes by.⁹ By itself, the hypothesis that Lee Harvey Oswald acted on someone's orders does not automatically require a vast and complex conspiracy. However, to explain why the Warren Commission concluded that all the available evidence pointed to Oswald being the sole perpetrator, one does need a more elaborate cover-up involving the Warren Commission itself (or at the very least, a politically motivated effort to turn a blind eye towards unwelcome evidence). And to explain why that conspiracy has not been exposed by now requires an even grander one. As researchers have observed, almost every popular CT has the tendency to grow larger in scope over time, even if they start out small. As a result, in Keeley's words, “[w]hat began as a small conspiracy … invariably swells into a conspiracy of huge proportions” (Keeley 1999: 57). Because any counterevidence or missing evidence can in principle be explained away within a CT, attempts to refute it often backfire: the more arguments you garner against it, the more its proponents will be impressed by how powerful and devious the conspirators prove to be. As Byford (2011: 55) recounts, when the Protocols of the Elders of Zion, the most infamous conspiracy document in modern history, was exposed as a forgery concocted in Czarist Russia, “many of the book's admirers simply dismissed the evidence against it as a campaign by Jews to undermine the 'leaked' document which exposes so clearly their sinister secret.” Even academic researchers studying conspiracy theories have been accused of being government shills (Lewandowsky et al. 2015). It is this feature, I believe, that allows us to distinguish legitimate hypotheses about conspiracies from CTs in the derogatory sense, because it is characteristic of virtually all popular and unofficial CT circulating on the Internet and elsewhere. To summarize, any conspiracy hypothesis that requires us to assume preternaturally smart and powerful conspirators to be consistent with the available evidence (as necessitated by the principle of asymmetry) should be rejected as a baseless CT.

Note that the exact force of this probabilistic argument depends on the degree of openness of the society in which the conspiracy hypothesis is being investigated. In a closed and authoritarian society without free press and liberty of expression, it is comparatively easier (especially for those in power) to keep a lid on sinister malfeasances that hurt the public interest. Potential whistleblowers can be intimidated into silence or even gotten rid of, and meddlesome investigators can be sabotaged. In more open societies, such tactics may backfire by drawing more attention to the conspiracy, thus leaving more trails for other investigators to follow and thus to uncover the plot. In addition to facing lower costs of defection, whistleblowers may also receive higher rewards in open societies, in terms of reputation, fame, or money (see Pigden 2018 for an analysis of "defectibility"). Generally speaking, therefore, absence of evidence for a conspiracy is more informative in liberal democracies than in authoritarian societies, because the opportunities for finding incriminating evidence are more plentiful.¹⁰ To appreciate this point, we don't even need the chauvinistic assumption that politicians in ‘our’ liberal and democratic societies are less inclined to engage in conspiracies because they supposedly abide by higher moral standards; it suffices that in closed and autocratic societies they can just get away with it more often.

Another important factor to take into account is the increasing “transparency” of our digital age (Dennett and Roy 2015). Whereas powerful people and organizations in the past were able to keep secrets by restricting local information flows, the rise of the internet and telecommunication has created a much more open epistemological environment. As a result, as Dennett and Roy (2015) have argued, the “half-life” of secrets has been reduced significantly, and even the most powerful organizations have found it difficult to keep a secret for very long. In the current framework, one could argue that the rise of this “new transparency” has greatly accentuated the causal asymmetry principle: most social events leave far more traces than they used to.

If my proposed demarcation criterion makes sense, we can use it to separate legitimate hypotheses about conspiracies from unsubstantiated CTs. But it still does not show why CTs have such a bad reputation in the public arena, and why it is reasonable to be suspicious of many conspiracy claims, even before having properly investigated them.

3.3. Epistemic black holes

Law (2011) compared irrational belief systems to “intellectual black holes” into which unwary truth seekers may fall, never to escape again. I believe this metaphor is especially appropriate for CTs.¹¹ Fundamentally, this is because the evasive maneuvers according to which absent evidence or counterevidence is further proof of the conspiracy cannot be regarded as ad hoc. When we complain that a theoretical move is ad hoc, we are saying that it lacks an independent theoretical justification and is being deployed with the sole purpose of rescuing the theory or hypothesis from refutation (Bamford 1993; Boudry 2013). In the case of CTs, however, the rescuing auxiliary assumptions (“the conspirators must have fabricated this apparent counterevidence”) flow quite naturally from the theoretical core. If we are investigating the hypothesis that duplicitous agents are plotting behind the scenes, it is only natural to expect them to cover up their tracks and even to fabricate evidence. As Keeley has pointed out: “it is not ad hoc to suppose that false and misleading data will be thrown your way when one supposes that there is somebody out there actively throwing the data at you” (Keeley 2018: 429). The theory itself engenders the evasions and defense mechanisms that we rightly regard as suspicious (Boudry and Braeckman 2012). Once you are prepared to ascribe preternatural shrewdness and power to the conspirators in your CT, no possible evidence can shake you out of your conviction.

To feel the gravitational pull of such epistemic black holes, consider the following question: how can CT skeptics be so confident that perfectly executed conspiracies, in which the conspirators flawlessly cover up their every track, are fictitious? As pointed out above, this may be just a selection effect. By definition, we only know about the conspiracies that have failed, or at least have been exposed after completion. However, we do not know how many conspiracies were pulled off successfully because we will never hear about those. In this case, my argument based on the principle of asymmetry of causes and effects would be question-begging: I am already assuming that no group of people is clever and powerful enough to successfully cover up large-scale and complicated conspiracies.

In his famous critique of the “conspiracy theory of history,” Karl Popper argued that real history is dominated by unintended consequences and that, when a conspiracy is attempted, it “never – or ‘hardly ever’ – turns out in the way that is intended” (Popper 2002 [1963]: 166). In other words, there are always mishaps, slip-ups, misunderstandings, and unforeseen circumstances. Or are there? What if Popper is simply looking at a biased sample of history, oblivious to all those episodes in which things did go exactly according to some evil plan? There may be an untold number of genuine conspiracies that will never make it into history textbooks precisely because they were flawlessly executed. Even some academic scholars of CTs have been impressed by this argument and regard it as a reason to take CTs more seriously (e.g., Basham 2006).

Indeed, once you are lured into this conspiracist mindset, you can even give the known track record of failed conspiracies a conspiratorial twist. Perhaps “they” sometimes deliberately allow some minor plots to be exposed (such as the Watergate scandal), thus lulling us into a false sense of security. By spoon-feeding us evidence that suggests that real conspiracies are exposed sooner or later and that mishaps and leaks always occur, they divert our attention away from the bigger scheme of things (see Byford 2011: 154). In his philosophical defense of CTs, Basham argues that, despite all the critical arguments leveled against them by clever philosophers, CTs have emerged as “remarkably wily and resilient epistemic creatures” (Basham 2006: 133). He is quite right: CTs as a class have a remarkable degree of internal resilience and resourcefulness. But this is exactly what is so troubling about them. In fact, this extreme resilience leads to two related problems, as I will show in the next sections: (1) CTs are too easy to generate and (2) for any given historical event, multiple and contradictory CTs can be constructed, all equally compatible with the evidence. And this, finally, will allow us to explain the bad reputation of CTs.

4. A proliferation of CTs

4.1. A generic recipe for CTs

It is impossible to predict what the next major historical event will be, but one thing can already be stated with near-absolute confidence: whatever happens, and whoever (if anyone) is held responsible, the event will soon spark some unofficial CT, a hypothesis that will end up attributing preternatural intelligence and powers to some alleged conspirators. This is because every historical event can be turned into a CT, even the ones whose official version already involves a conspiracy. To see this, we can think of the following steps as constituting a “recipe” for generating novel CTs around any given historical event.

• – The official story. Take the official version of events accepted by mainstream media, governments, or scientists. Whatever this official version, this is not what actually happened. It is a cover-up invented and disseminated by the conspirators working behind the scenes to distract the public's attention from some far more sinister events. By definition, the efforts of the conspirators will always appear “successful” because the view they wanted to impose on the rest of society is exactly the official, mainstream one.

• – Official conspiracies. If the received version of events already involves a conspiracy (such as 9/11), one just has to invent a higher-order conspiracy, in which the officially accepted conspiracy is nothing but a false flag operation. The designated “conspirators” are never the real conspirators.

• – Refutation of the official version. To attack the official version, focus on any unresolved questions, apparent contradictions, or minor gaps in the official narrative. This will not be too difficult. No matter how well-documented, no historical account of any event is ever fully complete. Historical explanations are always “fuzzy around the edges” (Dentith 2019: 2248). Indeed, the more extensively documented a historical event, the more opportunities for finding such suspicious anomalies (the 9/11 attacks partly happened on live television). From these “errant data,” as Keeley (1999: 52) called them, one can derive the conclusion that the official narrative cannot possibly be true and that something dark and sinister is going on.¹²

• – The identity of the conspirators. Who is really behind the event? It suffices to find any party who has benefited in some way from the historical event, or could conceivably have benefited. Because major historical events will always happen to benefit someone or another, especially powerful groups or organizations, this step will not be too difficult.

• – Lack of evidence for the conspiracy theory. As explained, absence of evidence need never discourage you. Under the assumption of your CT, missing evidence is precisely what you would expect if the conspirators have been very scrupulous in covering up their tracks.

• – Evidence against the conspiracy theory. Any counterevidence can be turned on its head and presented as further corroborating your CT. It is reasonable, after all, to assume that the conspirators will fabricate evidence to shore up the official version and discredit those who might reveal their secrets.

• – Critics. Any critics of your CT can be accused of being complicit in the plot, of being stooges paid by the conspirators to thwart honest investigations, or of being “sheeple,” gullible dupes of official propaganda.

As should be clear, these are perfectly general steps, which will work regardless of the nature of the historical event and the received account accepted by mainstream media or historians. Even paradigmatic cases of genuine conspiracies in the historical record can be (and have been) subjected to unofficial CTs. The Watergate scandal, for example, an actual and documented example of a nefarious cover-up that one would think would satisfy conspiracy theorists, has been the subject of higher-order CTs, in which the hidden truth was more sinister still. According to this CT, Nixon was actually innocent and the whole alleged break-in and cover-up had been staged by Nixon's enemies to remove him from the White House (Brotherton 2015: 65).

4.2. Radical underdetermination

The above recipe for making CTs allows one not only to turn every historical event into a CT but also to create multiple CTs for any given historical event. In philosophical terms, CTs are radically underdetermined by the available evidence.¹³ What this means in practice is that the theoretical parameters of any given CT (the culprits, the plot, and the mechanisms) are partly arbitrary and can easily be substituted for one another. As Harris (2018: 256) recently put it, “any number of conspiratorial explanations will fit the data, and hence will be equally supported.” The epistemic structure of a CT works like a protective outer shell, in which virtually any content can be inserted.

This is not to say that the parameters of a CT are completely arbitrary or unmotivated. As noted above, conspiracy theorists pose the cui bono question to identify the conspirators behind the scenes: who benefited from this course of events? This places some reasonable constraints on the generation of plausible CTs: theorists need to be able to spin a plausible narrative in which the alleged culprits stand to gain from what happened. For instance, if you believe that the moon landing was faked, the most straightforward answer to the cui bono question will be NASA or the U.S. government. Perhaps NASA wanted a prestigious project to attract more funding, or perhaps the Americans just wanted to beat the Soviets at their game and win the Cold War. Psychological or social factors may also place some constraints on the identity of the perpetrators. According to Lipset and Raab (1978), even though the perpetrators in a CT are often hidden from view, most culturally successful CTs still have a visible target that can act as a placeholder for the invisible culprits. For instance, while a secret organization like the Elders of Zion is hard to pinpoint, the Jewish community provides a suitable proxy.

Even accounting for these constraints, however, it is still true that, for most historical events, one can imagine multiple answers to the cui bono question. As a consequence, communities of CT believers often struggle with internal disarray. Because it is always possible to come up with a rival CT that posits a different plot with different perpetrators and that explains the available evidence equally well, it is often difficult to settle on any single CT. A good example is the assassination of JFK. If one rules out the official version that Lee Harvey Oswald acted alone, then a whole list of potential conspirators has to be considered. In a 2013 Gallup poll, a clear majority (61%) of Americans claimed that Lee Harvey Oswald did not act alone in killing JFK, and thus the official account is false. When respondents were asked who was most likely behind the assassination, opinions were divided across a wide array of potential culprits: the Mafia (13%), the federal government (13%), the CIA (7%), Cuba and Fidel Castro (5%), JFK's own vice-president Lyndon Johnson (3%), the Soviet Union (3%), the Ku Klux Klan (3%), FBI director J. Edgar Hoover (1%), and various other actors (Swift 2013). Because of the self-sealing logic of CTs, it is extremely difficult to settle disputes about the identity of the conspirators, or about the nature of the plot. If you consider that any of these parties could have successfully covered up their tracks, and that any evidence incriminating perpetrator X may have been planted by perpetrator Y, it is impossible to rule out the involvement of any of them. As a result, there is still no consensus in sight after half a century of conspiracy theorizing.

In documented conspiracies, such as the Watergate scandal, we expect to see the opposite pattern, again because of the principle of asymmetry of causes and effects. Because actually occurring conspiracies radiate outward into the future, leaving a whole spectrum of effects, independent lines of evidence will usually converge on the same conclusion and incriminate the same perpetrators. For that reason, the proliferation of multiple and conflicting CTs may be regarded as indirect evidence that an epistemic black hole has opened up around a particular event.

4.3. The likelihood of “observing” CTs

Given that it is always possible to construct multiple CTs around any given historical event, and given that conspiratorial explanations of history are psychologically appealing for a number of reasons (Uscinski 2018; Douglas et al. 2019), we should not be surprised to find that every major historical event will spawn multiple CTs, even if no actual conspiracy has occurred. A good example is the template of a “false flag” operation for terrorist attacks. In recent years, every single terrorist attack or mass shooting has been the subject of CTs following the “false flag” playbook, often within hours or even minutes of the attack (Stokes 2018). It is a very safe prediction that the next terrorist attack, whatever the circumstances and whoever the (officially designated) culprit, will immediately prompt one or multiple CTs.

Moreover, because of their warped epistemology, CTs can persist in the teeth of any evidence or counterevidence. In his recent defense of particularism, Basham (2018: 44/48) assures us that “[P]eople are not long attracted to conspiracy theories sans any evidence … Conspiracy theories that are long lived are most always characterized by interesting, if not always conclusive evidence.” However, Basham does not present any evidence for this claim and indeed it is demonstrably false. For instance, in the absence of any shred of credible evidence that the Twin Towers were brought down by controlled demolition¹⁴ and despite detailed refutations by structural engineers and demolition experts, the 9/11 Truth movement has not petered out and the “controlled demolition” claim is still one of the core tenets of the movement. Even the belief in the authenticity of the Protocols of the Elders of Zion, perhaps the most notorious CT in recent history, is still rampant a full century after having been exposed as a fraud (in The Times in 1921). To this very day, the Protocols are being regularly reprinted and disseminated as an authentic document, now predominantly in the Islamic world.¹⁵

It is this proliferation and persistence of CTs in the teeth of any evidence, I believe, that explains the derogatory meaning of the phrase “that's just a conspiracy theory,” or the disclaimer of people who want to defend a conspiratorial hypothesis: “I'm not a conspiracy theorist, but …” (Husting and Orr 2007; Harambam and Aupers 2017). If we find that a recent historical event is the subject of unofficial conspiratorial explanations, which are believed by a substantial group of people, the odds are very high that we are dealing with an unfounded CT. This prima facie suspicion can be accounted for in Bayesian terms, as follows.

If C_x indicates that there is an unacknowledged conspiracy behind an event X and CT_x means that event X gives rise to popular conspiracy theories, then P(C_x|CT_x) is the probability that there is an actual unacknowledged conspiracy behind X given that many people are alleging a conspiracy, while P(CT_x|C_x) is the likelihood that an actual conspiracy will give rise to allegations of conspiracy. Because no historical event is immune from CTs and because many people are psychologically disposed to conspiratorial explanations, the likelihoods P(CT_x|C_x) and P(CT_x|¬C_x) will both be very high. In other words, even if there is no actual conspiracy, it is still very likely that there will be persistent allegations of conspiracy. Following Bayes’ formula, P(C_x|CT_x) = P(CT_x|C_x)P(C_x)/P(CT_x). In other words, whatever we think of the prior probability of an actual conspiracy behind an event X, the mere fact that there are certain persistent CTs around the event has hardly any probative value. It always happens. Moreover, there will typically be multiple and contradictory CTs, implicating different culprits and involving different plots. Those who are inclined to do so will never fail to come up with some CT regardless of what happened.

Suppose that, in the wake of the next terrorist attack (claimed by either jihadists or right-wing extremists), someone says: “Don't you see, this was just a false flag operation staged by the government with paid actors?” (Stokes 2018). Even before I have considered the evidence for this claim, an attitude of prima facie skepticism is warranted, because I know that those who are inclined to do so will always “see” false flag operations regardless of what actually transpired, and that such CTs can hold sway in certain circles even in the total absence of any credible evidence. Given all this prior knowledge, and given that we do not always have time to investigate every bizarre claim that someone throws our way, I think it would be justified to reply: “Well, that's just a conspiracy theory.”

Note that I am here assuming the perspective of someone who does not find himself in an environment that is seriously “epistemically polluted” (Levy 2021). For someone who already believes that previous and similar terrorist attacks were false flag operations carried out by the government, for instance because this person is immersed in an online echo chamber brimming with conspiracist misinformation, it may in fact be rational to believe that this novel attack was also a false flag operation. This would follow through simple induction or inference to the best explanation (“If they could pull off that other terrorist attack, why not this one? There are already plenty of precedents.”). This observation explains the empirical finding that the best predictor of belief in any given CT is belief in other CTs (Goertzel 1994; Swami et al. 2011). This is an instance of the more general epistemological point that, if an agent's evidence base is heavily distorted or biased, it may be “rational” to accept beliefs that would otherwise be irrational.¹⁶

4.4. Abusing pejorative concepts

Like all evaluative concepts, “conspiracy theory” is susceptible to abuse. Particularists such as Basham, Dentith, and others are right to worry that actual conspirators may exploit the bad reputation of CTs to obstruct unwelcome investigations into their nefarious doings. Dentith (2020) recently claimed that the Nixon administration initially brushed off accusations about the Watergate Affair as “just conspiracy theories,” and although they did not provide actual quotations, we may well take their word for it, since it would be the smart thing for any conspirator to do. However, any derogatory term is susceptible to abuse. If phrases like “that's just a conspiracy theory” are brandished to prematurely dismiss legitimate questions about possible conspiracies, we would expect the principle of asymmetry of causes and effects to eventually resolve the matter. It should also be noted that the same potential for abuse exists for the derogatory term “conspiracy theory phobia” or “conspiracy denialism,” which have been proposed by some particularists to characterize what they see as an unjustified suspicion of CTs (Basham and Räikkä 2018; Pigden 2018). Whether or not their concept has some merit, it should be clear that it too is open to abuse. For instance, antisemites may happily use it to pathologize the widespread skepticism about a global Jewish conspiracy for world domination (Brotherton 2015). However, that by itself is no reason to reject the concept of “conspiracy theory phobia.” To draw one last parallel with the old demarcation problem, the term “pseudoscience” has also been abused by, for instance, Intelligent Design creationists to dismiss Darwinian evolution (e.g., Johnson 2002). But just because “pseudoscience” has been hijacked by pseudoscientists does not now show that it is a useless concept. What it does show is that every concept should be handled with care and that there may be reasonable (or unreasonable) boundary disputes about its proper application.

5. Conclusion

What is it about “conspiracy theories” that makes the very concept into a term of abuse, hurled around in the public arena to discredit the views of opponents? In this paper, I have argued that an intellectual presumption against CTs – provisionally defined as unofficial and conspiratorial explanations of historical and current events – is indeed justified. Granted, nefarious conspiracies abound in the historical record, but investigators should expect to encounter incriminating evidence as time wears on, because of a deep asymmetry between causes and effects in the world. This point applies especially to the kind of large and complex plots that are characteristic of virtually all popular CTs. And it applies with even greater force in open and democratic societies characterized by freedom of expression and (relatively) unfettered access to information, and in the “new transparency” of our digital age.

Although there is no crisp line dividing respectable hypotheses about conspiracies from CTs (in the derogatory sense), the same is true with respect to science and pseudoscience. However, this does not stop us from applying these terms to prototypical and uncontroversial instances on both ends of the spectrum. My proposal to solve the demarcation problem for CTs can be boiled down to a simple rule of thumb for anyone investigating possible conspiracies: if at some point your conspiracy hypothesis can only be rescued from refutation by making the alleged conspirators preternaturally intelligent and powerful, then you have entered the realm of unfounded CTs.¹⁷ Since I believe almost all popular beliefs widely labeled as “conspiracy theories” fall foul of this criterion, and since I have tried to rescue the evaluative component of the everyday concept, I have indeed “bake[d] in the irrationality” (Dentith 2018a: 104) into my definition itself. But this is not a problem, as long as this is unambiguously stated at the outset.

Semantics asides, the main argument of this paper is that theoretical explanations of history that fall foul of my demarcation criterion open up an epistemic “black hole.” Once you start attributing preternatural foresight, power, and intelligence to the alleged conspirators, no possible evidence can shake you out of your conviction. Based on this analysis, I presented a generic recipe for constructing novel CTs that, I claim, will never fail to work. Given that the human mind is generally disposed to conspiracist ideation (van Prooijen and Van Vugt 2018), we should not be surprised that every major historical event will spawn various unofficial CTs, regardless of the available evidence and no matter what actually happened.

This leaves us with the question: is the current account a vindication of “generalism”? What it does not give us are general grounds for dismissing any hypotheses involving conspiracies. It is obvious that some conspiracies are real and it is rational to believe in them. What my analysis of “epistemic black holes” offers is a description of the common epistemic defects from which virtually all unofficial conspiratorial explanations of history suffer. These so-called “CTs” are not just a varied and miscellaneous bunch of theories which have nothing in common apart from the fact that they involve conspiracies and are unofficial or unwarranted. They follow the same explanatory recipe, and once you start recognizing the steps, you are right to be very suspicious.

In this way, we can also rescue the pejorative connotations of the concept “CT” in the public arena, exemplified by widely used phrases such as “That's just a conspiracy theory” and “Look, I'm not a conspiracy theorist, but …” The intuition behind these everyday phrases is that CTs are gratuitous and facile, that anyone can dream up such fantastical notions that are conveniently shielded from refutation. Because, have you heard that even academic journals are now being infiltrated by Deep State agents posing as real philosophers in an attempt to shore up all the official lies we have been told about 9/11?¹⁸