1 Introduction

The rationality of belief in conspiracy theories is a hotly debated issue in contemporary philosophy. There are two main views in the debate. There are the generalists who argue that belief in conspiracy theories is prima facie irrational (Cassam, 2019; Napolitano, 2022; Napolitano & Reuter, 2021; Sunstein and Vermeule, 2009). By contrast, the particularists understand conspiracy theories to be putative explanations and argue that each conspiracy theory ought to be examined on its own merits. According to the particularist, we cannot assess conspiracy theories as a class but, rather, conspiracy theories should be assessed on a case-by case basis (Basham, 2018; Coady, 2012; Dentith, 2016, 2019; Keeley, 1999). On this view, there is nothing inherently irrational about belief in conspiracy theories; after all, conspiracies unravel every day, and most persons would agree that history is full of secretive plots, political and otherwise (Pigden, 1995).

However, empirical research indicates that “conspiracy beliefs emerge as ordinary people make judgments about the social and political world” due to situational triggers and subtle contextual variables (Radnitz & Underwood, 2015, p. 113). Reviewing recent scientific findings, Brotherton (2015) concludes, in the same vein, that conspiracy belief is far from being generally bizarre, but that “[w]e have innately suspicious minds” and we are all “natural-born conspiracy theorists” (ibid., p. 17). Poth and Dolega (2023) show that conspiracy belief can be Bayes-rational. It is true that the fact that people are naturally-born conspiracy theorists is logically compatible with generalism about conspiracy theories. However, there is a tension between these ideas. How come people are so prone to conspiracy belief if such belief is prima facie suspect? The tension is less obvious between people being naturally-born conspiracy theorists and particularism about conspiracy theories. If some conspiracy theories are plausible, it comes as less of a surprise that people are inclined to believe in them. However, this raises the problem of explaining what separates rational from non-rational belief in conspiracy theories for the particularist.

In this paper I will contrast three strategies to this end. First, I will present the objectivist strategy, which assess the objective probability of conspiracy theories. Second, I consider the subjectivist approach, which focuses on the rationality of conspiracy belief from the point of view of the believer. There are two standard perspectives: under risk and under uncertainty. First, I will consider the decision to believe in a conspiracy in the framework of decision making under risk using determinate probabilities and epistemic utilities. However, relying on determinate probabilities seriously limits the applicability of both these approaches. Moreover, focusing solely on epistemic utilities fails to do justice to the particular character of conspiracy beliefs, which are not exclusively epistemically motivated. Thus, lastly, I explore the second perspective of the subjectivist approach, viewing belief in conspiracy theories under uncertainty about probabilities. In this context, I investigate the rationality of conspiracy belief under a number of standard decision rules and sources of non-epistemic utilities.

2 The Objectivist

The objectivist view holds that external factors of a conspiracy theory determine its rationality. Thus, the objectivist places his or her reliance on empirical data to determine the viability of conspiracy theories. For instance, a large conspiracy may be more likely to be revealed as time passes, and the corresponding theory becomes increasingly unlikely when the conspiracy is not revealed. This is precisely what David Robert Grimes argues (Grimes, 2016, 2021). Following his definition, a conspiracy theory is an explanation of some event or practice that references the machinations of a (usually powerful in some sense) small group of people, who attempt to conceal their role (at least until their aims are accomplished).Footnote 1 Grimes acknowledges that there are conspiracies that are viable and therefore believing in them would not be irrational, writing that it would be “unfair [to] dismiss all allegation of conspiracy as paranoid where in some instances it is demonstrably not so” (2016, 2). Grimes goes on to suggests that there “must be a clear rationale for clarifying the outlandish from the reasonable”. His suggestion is that we can determine the likelihood of the viability of a claim of conspiracy by looking at the number of conspirators involved, as well as the amount of time passed since the conspiracy occurred.

To this end, Grimes puts forth a simple mathematical model for estimating just such viability. The parameter p is the probability of an intrinsic leak or failure, and is estimated from known conspiracies, what Grimes calls “exposed examples” (p. 6). The three examples he uses are the National Security Agency PRISM affair, secretly conducting excessive surveillance of internet traffic; the Tuskegee syphilis experiment, where the US Public Health Services made unethical observational studies on African-American men; and the FBI forensic scandal, the pseudoscientific nature of the bureau’s forensic testing resulting in many innocent men being incarcerated. Taking the number, N, of conspirators and the life-time t of the conspiracy and calculating the failure odds, Grimes suggests that we can predict “a best-case scenario” for a conspiracy theory’s viability. In his model, Grimes assumes that everyone involved is in on it. For example, in the NSA case, Grimes assumes that all employees at NSA at the time, approximately 30,000, were conspiring. He makes a number of further assumptions: (1) that the conspirators are dedicated to kee** their conspiracy a secret, (2) that leaks by the conspirators expose the conspiracy, (3) that the number of conspirators over time, will depend on one of the following factors: If the conspiracy requires constant upkeep, then the number of conspirators remains constant; if the conspiracy is a single event and no new conspirators are required, then the number of conspirators involved will decay over time; and, finally, if the conspirators are rapidly removed for some reason, there is an exponential decay. I will return to these matters in a moment.

With data from these three cases above, Grimes (6) estimates the probability of failure due to internal leaks, with the lower bound given byFootnote 2

$$p > 1 - \sqrt[{N(t)}]{{1 - \frac{\ln \;2}{t}}}$$

Thus, the probability of failure is seen as a function of N and t, such that a high probability yields a low chance of a successful conspiracy.

With Grimes’ model in mind, consider the following example. The Swedish Prime Minister Olof Palme was assassinated in central Stockholm on February 28, 1986, while walking home one night from the movies. The assassination and subsequent police investigation have caused one of the greatest controversies in Swedish history. The case was considered closed by the authorities after more than 35 years of investigations, a decision that faced heavy criticism. Many explanations as to what really happened are still floating around, among them conspiracy theories per the Grimesian definition. For example, a version of the so-called SÄPO-conspiracyFootnote 3 identifies the Swedish Security Service (SÄPO), together with its American counterpart, CIA, as colluding to assassinate the prime minster. According to Grimes’ analysis, we should be able to predict the intrinsic failure of this theory by estimating the number of actors involved after a certain time, in this case about 1400 working for SÄPO and roughly 21,000 for the CIA.Footnote 4 Using the above equation, the estimate time for such conspiracy to be revealed would be approximately 34 years.Footnote 5 Since more than 36 years have passed since the murder took place, we can rule out the SÄPO-conspiracy as a non-viable conspiracy theory.

The objectivist view, then, seems to offer a clear-cut way of determining a viable conspiracy theory. All we need is to plug in some figures and we can end the discussion once and for all. However, even though this approach is indeed important in many ways, and Grimes may very well be correct in proposing that the number of conspirators has an effect on it potentially being leaked to the public, it suffers from a number of challenges that are not so easily overlooked.

One criticism raised by Dentith (2019) targets the examples that Grimes uses for the parameter estimation, more specifically the manner in which the three conspiracies were exposed. Grimes has it that a leak generally comes from a conspirator either accidentally (where a conspirator fails in some way to cover up the conspiracy) or intentionally (where a whistleblower purposefully leaks the information). However, Dentith argues that the whistleblower Edward Snowden, for instance, in the NSA case was an outsider, i.e., someone who discovered the conspiracy without having been part of it. Similar observations apply to the two other exposed conspiracies. Dentith shows that in all three examples, the leak originated from outsiders. Thus, Dentith concludes that there is “a mismatch between Grime’s chosen examples, and his theory about how leaks over time revealed and made these conspiracies redundant; his examples fail to capture the very thing he wants to measure” (p. 17). Dentith further criticize Grimes’ account on the grounds of its failure to distinguish between conspirators and whistleblowers, stating that it is even possible to be involved in a conspiracy without realizing you are conspiring, in which case you would see no need to become a whistleblower.Footnote 6

While I agree with Dentith’s criticism, in the present context I believe there is another fundamental concern, one that I take Grimes himself anticipates. He writes: “There is also an open question of whether using exposed conspiracies to estimate parameters might itself introduce bias and produce overly high estimates of p—this may be the case, but given the highly conservative estimates employed for other parameters, it is more likely that p for most conspiracies will be much higher than our estimate, as even relatively small conspiracies (such as Watergate, for example) have historically been rapidly exposed” (p. 9). However, Grimes severely underestimates the significance of this problem. I will briefly elaborate this point.

The working definition of conspiracy theories does not itself a priori dismiss conspiracy theories as inherently false. Let us call the class of all conspiracy theories, CT. Even so, there are conspiracy theories that we may dismiss, because, according to Grimes, they are “demonstrably nonsensical” and outlandish. Typically, they contradict scientific thinking or are based on evidence incompatible with science. Then there are those conspiracy theories that, according to Grimes are reasonable and whose viability is still an open question. These conspiracy theories “would be unfair to simply dismiss”, because they reflect real concerns of possible collusion. Importantly, they build on a narrative that can be challenged by scientific evidence. We can then estimate using Grimes’ model which of these are viable after time t. Then there is the class of conspiracy theories that are considered reasonable, but are not viable after time t. Finally, there is a class of conspiracies already mentioned that are exposed conspiracies.

According to Grimes, the model provides an estimate for the intrinsic failure of reasonable conspiracies after some time t. Importantly, these conspiracies may or may not be exposed in the future. However, the parameter estimates in the model are based solely on exposed conspiracies. Now, this matters because, as Grimes himself indeed emphasizes, “conspirators are in general dedicated for the most part to the concealment of their activity” so that “the act of a conspiracy being exposed is a relatively rare and independent event” (3–4). In fact, one of Grimes’ assumption (3), about a decaying group of conspirators, lend further support to the latter claim. In other words, Grimes’ parameter estimates introduce not only “overly high estimates” of p, but quite possibly intolerably high estimates. Another way of putting it, is that the model does not estimate the probability of failure per se, but the conditional probability of failure given that the conspiracy will be exposed at some point.

In order to make probability estimates about the intrinsic failure of conspiracy theories generally, a model must be based on a representative sample from the class of reasonable CT, one that is not skewed toward conspiracies that have been or will be exposed. Granted, this is very difficult, and most likely an unrealistic enterprise. In effect, this is a dilemma for Grimes: either the model quite possibly grossly overestimates the probability of intrinsic failure or it provides no guidance whatsoever as to the viability of a given reasonable conspiracy theory.

Nevertheless, important lessons are still to be had from the objectivist account, for example concerning the impact of the size factor when estimating the likelihood of conspiracy theories. But just how size matters is still an open question. Since Grimes’ model in the end cannot tell us much about the rationality (or lack thereof) of believing a conspiracy theory, I now turn to an alternative, subjectivist account.

3 The Subjectivist: Believing in Conspiracy as a Decision Under Risk

The subjectivist framework looks at belief in conspiracy theories as they arise from rational Bayesian cognition and considers the individual believer’s perspective, assuming that a decision-maker seeks to maximize the payoffs of their decision. I will use Gabriel Doyle (2021)’s account for detailing this framework. The account builds on previous research by Olsson (2020) indicating that seemingly irrational belief polarization can arise from ideally rational Bayesian reasoning, where two ideal Bayesian agents can end up divided on an issue given the exact same evidence to consider. Doyle believes that something similar is true for belief in a conspiracy theory: such seemingly irrational belief can arise from impeccably rational processes of belief formation.Footnote 7

In the spirit of Bayesianism, Doyle assumes that a person seeks the most satisfying explanation H for a set of observed phenomena D based on the posterior probabilities of the relevant hypotheses. These probabilities are calculated as the product of the likelihoods and priors for each hypothesis. To determine which hypothesis is preferred, Doyle looks at the epistemic utility of choosing a high-posterior hypothesis together with other additional utilities derived from the hypothesis themselves. In his framework there are two primary explanations available: the conspiracy hypothesis, HCT, and the non-conspiracy theory hypothesis, HNC. The posterior probability for each hypothesis is proportional to the likelihood of seeing observed data under a given hypothesis, times the prior belief in that source. The utility is in Doyle’s framework “primarily derived by identifying the true source of the data” (p. 3). This “epistemic” utility is a function of the posterior probability of the hypothesis given the data:

$$U_{epistem} \left( {H|D} \right) = {\text{log}}p\left( {H|D} \right)\propto{\text{log}}p\left( {D|H} \right) + {\text{log}}p\left( H \right)$$

Doyle considers that there may be other utilities factoring into the decision-making, such as personal satisfaction from having a belief. I will return to this topic in the next section. Further, he elaborates on a number of ways in which normatively optimal reasoning could lead someone to support or commit to a conspiracy theory, such as having elevated prior beliefs in conspiracies; having different likelihood functions, and having missing hypotheses. I will summarize each of these in turn.

According to the elevated prior hypothesis, strong beliefs in conspiracies raise the individual’s prior belief in a conspiracy theory explanation for future phenomena that may present themselves. According to Doyle, the hypothesis coincides with other findings about conspiracy beliefs, for example that conspiracy theory beliefs “appear to cluster” (Doyle, 2021, p. 4), (which I take to mean that people tend to believe more than just one conspiracy theory that are structurally similar).Footnote 8 If two people agree on a given CT’s likelihood and non-epistemic utility, the elevated conspiracy theory prior may be enough to yield a higher posterior for a conspiracy theory believer. Table 1 from Doyle (2021, p. 4) indicates that with a change from 5 to 20% in prior expectation of conspiracy, a rational non-conspiracy theory believer (N) is converted into a conspiracy theory believer (B).

Table 1 Dayle’s table of elevated priors. B is CT-believer’s probability, N is non-believer. HCT is the CT, HNC is the non-CT explanation
Full size table

Doyle next consideration is the different likelihood function, which state that even if two people agree on the set of hypotheses, the observed data and the conditional probability distribution, they may still end up with different posteriors, if they disagree on the value of another factor, for example which experts they trust. Take climate-change denial as an example. A person’s likelihood function for the scientific consensus on climate change depends, according to Doyle on a third variable, namely the trust placed on scientists. This factor may be enough to change the posteriors. In such cases, Doyle argues, the probability of the conspiracy theory hypothesis is higher for the conspiracy theory believer, who thinks scientist would lie to maintain a desired consensus than for the non-believer who considers this possibility unlikely. The third explanation of conspiracy belief is the missing hypothesis explanation. In this case the source of accepting conspiracy theory beliefs is that the true non-conspiracy explanation is not part of the conspiracy theory believer’s arsenal of potential explanations.

While Doyle should be credited for the subjectivist approach to conspiracy belief and for noticing, without further elaboration, the potential role of non-epistemic utilities, I believe this framework needs to be further developed, in order to better understand conspiracy beliefs. My reasons for this are threefold: First, more can be done about the fact that a decision is made under conditions of known probabilities, which makes a framework such as Doyle’s vulnerable to the same problems as the objectivist account, namely that such known probabilities are difficult to come by (see previous section), which is of course also a worry for decision theory generally; second, it relies on numerical utilities, which again, calls for speculation as to what to base those numbers on; and third, I argue, it doesn’t capture the non-epistemic utilities involved in many conspiracy beliefs. These remarks suggest that an alternative view is worth develo** in addition to the already existing frameworks, namely accounting for conspiracy belief within decision theory under uncertainty about probabilities, focusing on non-epistemic utilities, which I will develop next.

4 The Subjectivist: Believing in Conspiracy as a Decision Under Uncertainty

Belief in conspiracy theoriesFootnote 9 can sometimes be primarily psychologically motivated at the expense of a primary concern for truth. As such, they can fill the need to explain unusual or stressful events such as terrorist attacks, disease outbreaks, scandals, illegal mass surveillance and sudden, complex deaths (van Prooijen & van Vugt, 2018). They may reduce the complexity of uncertainty into focused fear, and they may also motivate collective action, when used in the context of relationships between in- and out-groups (Franks et al., 2013). Conspiracy theories can serve multiple purposes, including various political grievances (including legitimate ones) in contemporary society (Doyle, 2021; Gosa, 2011). As such, the conspiracy theorist emerges as a social phenomenon, and not a cognitively challenged or epistemically vicious outlier. This should arguably be reflected in a decision theory about our ordinary choice situations. Typically, we are not handed probabilities about the possible states the world might be in. Rather, as decision makers we often have to make our decisions under uncertainty. Considering the notion of non-epistemic utility coupled with the fact that we do not often have posterior probabilities for our choices, can provide a broader framework for understanding conspiracy theories, and why rational people believe in them.

I will consider two potential sources of non-epistemic value, which I call puzzle hunt and social power. Puzzle hunt has two aspects: one is the human instinct to solve puzzles, and another is the social effect that follow. For example, one might gain increased social status, social credit or other form of recognition, by solving a mystery or puzzle. Social power involves ways in which a belief in a particular conspiracy theory may contribute to a social power struggle.

It may manifest itself in protesting against oppressors, undermining authority or legitimacy, by challenging the reputation of those believed to have conspired.Footnote 10 I will discuss each in turn and I will apply a number of decision rules to formalize the circumstances under which a rational decision-maker may come to accept a conspiracy theory over a non-conspiracy theory. Of course, these are only two sources of non-epistemic value, and not meant to be an exhaustive list.

4.1 Puzzle Hunt

Humans like a good puzzle, and as the anthropologist Marcel Danesi argues, the need to solve puzzles, for mystery and suspense, is displayed in humans at a level not found in any other animal species. Danesi writes that “puzzles and mysteries are intrinsically intertwined in human life” since the dawn of history (Danesi, 2002, p. 2); a phenomenon I shall dub Puzzle hunt. The question of why people are fascinated by seemingly trivial puzzles that often require substantial time and mental effort to solve is beyond the scope of my discussion here. However, I would suggest that there is a social aspect to puzzle hunt, much like the concept of Counterknowladge and Treasure hunt described by Gosa (2011). Puzzle hunt is social in that it can be socially motivated, by the desire to gain appraisal for being the one who solved a mystery.

Puzzle hunt, then, as characterized above, is arguably driving some conspiracy theorists. Considering again the case of the conspiracy theories surrounding the assassination of the Swedish prime minister. Many felt dissatisfied when years of police investigations only led to the wrongful arrest and conviction of a suspect who was later freed. In light of these events, a group was formed that became known as “the private investigators” (or “Privatspanarna”). Some of the private investigators organized regular meetings, discussing and debating the puzzle-pieces they felt were missing to solve the murder-mystery. Some of the private investigators were retired police officers and journalists that, even after retiring, continued working on the case. What they all had in common was a feeling that the mystery was not yet solved and that they wanted to find out what really happened that winter night in 1986. The concluding remarks on the website palmemordet.se leaves us with the same sentiment: “[the] murder of Olof Palme remains a mystery with no sure answers”.Footnote 11

Let us now consider the problem of deciding which hypothesis to believe from the perspective of puzzle hunting. To fix ideas, we consider the conspiracy theory (CT) to be about the Palme murder, and the decision-maker might be a member of the group of private investigators. There are two possible states of the world: either CT is true or CT is false. The decision maker can either believe in CT (BCT) or not believe in CT (BNC). Given a possible state of the world, the outcome of believing the one or the other, in terms of how the decision maker is likely to be perceived by her peers, will be different, as illustrated in Table 2.

Table 2 Potential utility preferences of a private investigator, with Puzzle hunt as a source for non-epistemic value
Full size table

According to Table 1, U1 is the most prioritized utility due to the great satisfaction of pursuing a puzzle hunt and being the one who was able to solve it, thus being right. However, if one believes the conspiracy theory and it is false, there is a greater negative loss, compared to not believing the conspiracy given the same outcome. There are a number of reasons for this, one of which can be a loss of social status when a theory is believed and otherwise acted upon (by telling others or in other ways expressed through behavior), and there is further risk that people will not believe or take the persons conviction seriously in the future. In the scenario where one does not believe the conspiracy theory and it is true, there is a lost opportunity of having been right, and solving the puzzle, as well as being duped by the conspirators; but this loss is not as costly as having pursued a false conspiracy theory, as shown in the table. The following preference order in ordinal ranking U1 > U4 > U3 > U2 over the outcomes can then reasonably be assumed (given that there is no other information about the utilities):

Being the first to solve the puzzle

is better than

being misled,

is better than

being naïve, and duped by the conspirators

is better than

being dismissed as a conspiracy theorist in the pejorative sense.

Let us now look at the outcome of applying some standard decision rules to the decision-matrix in Table 1. The maximax rule—maximizing the maximum outcome—is based on the decision maker being optimistic about the future. In the puzzle hunt case, the optimal choice under the maximax rule is to believe the conspiracy theory (B CT).Footnote 12

By contrast, the maximin rule is based on the decision-maker being pessimistic about the future. It recommends the decision maker to choose the alternative that has the maximal security level in the sense that it maximizes the minimum payoff. In our case of the puzzle hunt scenario, the matrix for the maximin rule indicates that the appropriate decision is to not believe the conspiracy theory in order to secure the best option out of the worst ones.

4.2 Social Power

Social power is a source of non-epistemic value of some conspiracy theories with practical value in exposing injustice or oppression, as well as a way of maintaining or gaining social power. The economist Péter Krekó argues for the importance of groups and social identity in relation to conspiracy belief, in particular how these can drive collective action. He writes that: “Conspiracy theories emerge in groups and are deeply rooted in the social identity of the group. These beliefs help the group and the individual as its member to understand and make sense of the world. They even help raw line between the in-group and the out-group (see for example “out-group paranoia,” Jost & Kramer, 2002). As these theories are anchored to the social identity of the groups, they can reveal the characteristics of the groups where they emerge, spread, and drive collective action” (Krekó, 2015, 64).

As such, one can utilize belief in conspiracy theories to either prevent, undermine or encourage social upheaval. It can be viewed as an expression of (desired) social power more generally. As van Prooijen and Douglas (2018) notes, some consequences of believing in conspiracy theories may be positive, depending on the social change that these movements pursue. For instance, conspiracy theories can inspire and justify protest movements.Footnote 13 Imhoff and Bruder (2014) consider the role of conspiracy mentality in motivating social action aimed at changing the status quo.

There are many historical examples that can illustrate this. One is Machiavelli (1532/1998) who was very aware of the danger of conspiracies for the nobles. He wrote in The Prince that a prince should avoid being hated by the commonfolk, so that he would not have to fear conspiracy (consequently theorizing about possible conspiracies, i.e. thinking about conspiracy theories). Edward Snowden highlighted the concept of the enemy as fundamental in a conspiracy theory, and that conspiracy theories can “erode civic confidence” (Snowden, 2021, p. 3). The way Snowden sees it, in democracies today it is not our rights and freedoms that are important, but rather “what beliefs are respected: what history, or story, undergirds their identities as citizens, and as members of religious, racial and ethnic communities”. Gosa (2011) likewise discusses the use of conspiracy theories as counterknowladge in hip hop culture. Conspiracy theory, as such are an alternative knowledge system in certain subcultures, and they are “intended to challenge mainstream knowledge producers such as the news media and academia” (Gosa, 2011, p. 190). According to Gosa conspiracy theories “give voice to inequality”. This does not mean that they always succeed. Rather, as Gosa notes, wrap** legitimate concerns in conspiracy theories can serve as a digression from the inequality it wishes to expose; and it may hinder “one’s ability to be a powerful cultural force for racial justice” (p. 201). In this sense there is also a considerable risk one must consider with conspiracy beliefs for social power. Consider the following concept to illustrate the social power as a source of non-epistemic value: the patriarchy as a conspiracy. Certain feminist movements characterize the patriarchy in terms of a conspiracy, as the systematic oppression of women by the patriarchy (Haraway, 1988; Hill & Allen, 2021; Knight, 2000). Haraway describes the identification of The Other in academic feminism in terms of a conspiracy. She writes: “We have used a lot of toxic ink and trees processed into paper decrying what they have meant and how it hurts us. The imagined “they” constitute a kind of invisible conspiracy of masculinist scientist and philosophers replete with grants and laboratories (Haraway, 2013).Footnote 14 Further, Knight writes about the role of conspiracy rhetoric in feminism: “In the struggle to give a name to—and find someone to blame for—what Berry Friedan famously called the “problem with no name,” feminists writers have often turned to conspiratorial rhetoric. […] Metaphors of conspiracy, I want to suggest, have played an important role within a certain trajectory of popular American feminist writing over the last thirty years in its struggle to come to terms with—and come up with terms for—the ‘problem with no name’” (Knight, 2000, p. 117).

Knight further discusses the figuration of conspiracy in creating a coherent women’s movement and also how it has become a source of division between feminists. Hill and Allen emphasize the conspiracy framework as a two-edged sword for feminist when identifying the patriarchy. On the one hand “feminist—and others—are utilizing the concept” (Hill & Allen, 2021, p. 170), and they suggest that it has an important community-forming aspect. But simultaneously, they warn that it can be a “risky strategy for feminists” (p. 165). The reason it is a risky strategy, is because the notion of patriarchy as a conspiracy is, as they say, also a key theme in anti-feminist strategy to undermine feminist critiques of power. Further, the authors argue that the discourse that the patriarchy “is ‘nothing more than a conspiracy ‘—created in order to victimize men—has important implications for feminist political claim-making [and] attempts to discredit feminism.” (p. 184–185) I believe this neatly illustrates how one can utilize conspiracy theories both to expose powerful organizations or people, and also to undermine critique.

Table 3 considers alternatives with the perspective of the conspiracy of women being oppressed by the patriarchy. The patriarchy is conspiring to keep women oppressed, and to maintain their power over institutions and other social constructs.

Table 3 Utility preferences of a feminist, to undermine the patriarchy and expose the conspiracy against women in order to stop the oppression
Full size table

A reasonable preference relation (again with no information available of probabilities) for the social power cases can be U1 > U2 > U4 > U3, as follows:

To undermine the patriarchy with little to no risk

is better than

undermining the patriarchy, with a great risk of backlash and being exposed, is better than

continued oppression of women, but at least with no lost opportunity of undermining the conspiring powers

is better than

the continued oppression of women with the missed opportunity to expose the conspiring patriarchy.

In the reversed scenario, it is the patriarchy undermining the criticism, by utilizing the conspiracy theory, CT* which says that women are conspiring by falsely accusing the patriarchy of conspiring, labeling it a conspiracy theory. Table 4 represent just such a case. The preference order is the same as in Table 3.

Table 4 Matrix of the patriarchy utilizing CT* to retain social power, and undermine criticism directed towards it
Full size table

Applying some standard decision rules, maximax and maximin, to the decision-matrix in Table 3, starting with the maximax rule in the social power case, the optimal choice under the maximax rule is to believe the conspiracy theory (BCT). On the other hand, considering the maximin rule, based on the decision-maker being pessimistic about the future, the recommendation is choosing the best of the worst alternative, the choice to undermine the patriarchy (alternatively to undermine feminism), with the risk of backlash, U2.

5 Discussion

Arguably, there are other standard decision rules that would be insightful to consider in the cases considered above, such as the regret criterion and the Hurwicz criterion. However, in order to apply these rules, we must deviate from the account above and assign numerical utilities to the alternatives (see Tables 5 and 6 for the utilities ascribed to the puzzle hunt and social power criterion respectively). To produce the regret criterion (minimax regret), we assign to each outcome the difference between the utility of the maximal outcome in its column and the utility of the outcome itself (Hansson, 2005). Then the rule recommends to choose the alternative with the minimum of all maximum regrets (see Tables 7 and 8 for regret matrices for both cases).

Table 5 Assigning utility values and applying the Hurwitz criterion to the Puzzle hunt condition
Full size table
Table 6 Assigning utility values and applying the Hurwitz criterion to the Social power condition
Full size table
Table 7 Regret matrices for the Puzzle hunt condition
Full size table
Table 8 Regret matrices for the Social power condition
Full size table

With social power as non-epistemic utility the minimax regret criterion recommends believing the conspiracy theory (BCT). The same is true for the puzzle hunt case.

A middle way between the pessimism (maximin) criterion and the maximax optimism criterion is the Hurwitz criterion (sometimes called the optimism-pessimism index). According to the Hurwitz rule, we choose the best weighted average payoffs; a compromise between the best and worst payoffs. We assign the utility values as per Tables 5 and 6 above and we choose a index, i, between 0 and 1, that reflects the decision-makers level of optimism or pessimism. In the present case I have assigned index .7. A regret matrix may be derived from the above utilities, starting with social power: (50 × 0.7) + (30 × 0.3) = 44 and (− 10 × 0.7) + (− 2 × 0.3) = −7.6, in which case the best choice would be BCT. In the puzzle hunt condition, we get the following values: (50 × 0.7) + (− 10 × 0.3) = 32 and (− 2 × 0.7) + (10 × 0.3) = 1.6, and again the Hurwitz criterion recommends BCT.

One concern commonly raised (as old as decision theory itself!Footnote 15) is to what degree the decision maker or the agent can choose what to believe. The discussion so far has been framed in terms of believing or not believing in a conspiracy theory being the main options among which the decision maker has to choose. However, nothing hinges on any particular account of belief. Rather, what is required is some form of minimal commitment on the part of the decision maker. Unlike full-blown belief, committing oneself may be under the decision maker’s voluntary control. The distinction between acceptance and belief can be useful to consider. For instance, Maher (1993) proposes a decision theoretic account of scientific acceptance of a hypothesis, where accepting a hypothesis involves a voluntary decision to cease inquiring in the matter.15 Dentith (2021) also discusses the possibility of people sometimes only having a “weak commitment” to a conspiracy theory belief.

They argue that some people may only be entertaining the idea, but don’t necessarily believe it. For example, some conspiracy theorists will claim to be suspicious about politicians or business leaders, leading them to suspect something is not right in the politics. However, as Dentith argues, “a suspicion that something is wrong does not necessarily commit them to any particular conspiracy theory; they may well entertain the idea of a variety of conspiracy theories without necessarily being strongly committed to even one of them. […] but such suspicions do not entail any strong commitment in a resultant conspiracy theory” (Dentith, 2021, 9902–3).

Further, a decision can be rational without being right and right without being rational. Decision theory as I have engaged with it in this account is concerned with rational decisions, rather than right ones. Instrumental rationality presupposes that the decision maker has some aim, such as to become rich and famous, hel** homeless cats, or as in our case to gain social power or be the first to reveal a conspiracy. The aim is external to decision theory, and the recommended choice will be relative to that aim. Relatedly, a reasonable objection to this account might be to question the notion that belief in conspiracy theories can come out as rational on essentially non-epistemic grounds. In response, one may interpret decision theory, as I have indeed intended above, not as a normative theory, but as a descriptive theory of the rational decision maker’s choices, characterizing and explaining regularities in the choices that people are disposed to make (Chandler, 2017). Of course, the choices people should make according to rational standard are not completely independent from the choices they in fact make, which is why the normative and the descriptive are not always easy to separate.

However, the subjectivist uncertainty account, as I have argued for here, is an important addition to understanding belief in conspiracy theories. While empirical, and epistemic features of conspiracy theories are one important aspect, I believe a more comprehensive account also acknowledges the social dimension of beliefs in conspiracy theories; a dimension which is difficult to ignore when the phenomenon is so often political, and a moving target, to which real-world data are hard to come by. I believe this account has highlighted conspiracy theories as a social phenomenon; and it provides a framework that looks at the inherently social cognitive dimension of the decision-making process in the beliefs we hold.

6 Conclusion

The philosophical discussion on conspiracy theory has to a large extent centered around the question of whether or not conspiracy theories are justified to believe in, and whether or not we may prima facie dismiss such beliefs as irrational. Empirical research has suggested that most people believe at least one conspiracy theory. As I have argued, such results beg for an explanation as to what separates rational from non-rational conspiracy belief. I have considered three strategies to this end: the objectivist, the subjectivist under risk and the subjectivist under uncertainty. I have argued that the objectivist framework in its most developed form faces a dilemma: it either grossly overestimate the probabilities of intrinsic failure or provides no guidance as to the viability of a given conspiracy theory. The subjectivist under risk may not be able to provide the necessary probability estimation and quantitative utility assessments. In which case an alternative—or as a compliment to these accounts—I have developed an account of subjectivism under uncertainty about probabilities given non-epistemic utilities. In this model I consider two sources of non-epistemic value, puzzle hunt and social power, showing that whether conspiracy belief is rational depends not only on the type of non-epistemic value, but also on the given decision rule used. Importantly, there are several standard decision rules that do not require any quantitative utility assessments.

Of course, there may exist many more non-epistemic utilities. But my account as it stands, I believe, is an important addition to the conspiracy theory literature in that it, without excluding the epistemic importance of these beliefs, highlights the social aspects of conspiracy theories. The epistemic discussion is one important part of understanding conspiracy theories. But conspiracy theories are infamously moving targets, which makes it difficult to exclusively understand belief in them from an objectivist point of view. This account has argued that much research identifies conspiracy theories as a social phenomenon, which calls for an account that looks at the inherently social cognitive dimension of the decision-making process in the beliefs we hold, especially when facts about the world are not readily available. Utilizing the well-established philosophical tools of decision theory to conspiracy theory theory (the academic study of conspiracy theory), I believe, may provide helpful nuance to the research and frameworks going forward. Finally, while the account developed here has a distinctive descriptive flavor, it is not without normative import of which the details require a lengthy discussion, which is best left for another paper.