1. Introduction
We often refer to the beliefs of groups. We might say we think it’s a lovely day out, or that Americans believe democracy to be the best form of government. To what sorts of phenomena do we refer when speaking this way? In the context of belief, should words such as Americans and we be viewed as shorthand for summing up the individually held beliefs of a group’s members? Or, rather, do such words sometimes correctly refer to a group engaged, as a unified entity, in the collective activity of believing? If the latter is true, then it’s possible for a group to believe that p, while the individuals who constitute that group each believes that not-p. Call this sort of belief, in which a group literally has belief of its own, collective belief (while, for clarity’s sake, I’ll use group belief as a neutral term).
On my view, groups are not capable of collective belief. This distinguishes belief from many other sorts of activities. When Mary sings a C and Larry sings an E, for example, Mary and Larry collectively sing a major third, though neither individually sings a major third. Likewise, they might hug, cook dinner, or make the bed. Belief doesn’t seem to work this way: to correctly say that Mary and Larry believe their house is haunted, both must believe that proposition. Ascriptions to larger groups may afford some leeway in this regard, but still require that most members hold the group belief.
In this paper, I argue that the phenomena invoked by such examples do not amount to collective belief. First, I introduce the influential opposing view of Margaret Gilbert, whose joint commitment account would allow for Mary and Larry to individually disbelieve their house haunted while collectively believing otherwise.[1] I respond by observing three basic categories of group belief, pertaining to: a casually formed group (e.g., two strangers passing on the street); a formally organized collective (e.g., a committee); a large mass of people (e.g., “the French”). Our evaluation of the joint commitment account may differ depending on a group’s nature. As I go along, I attempt to show that the results of that evaluation also give us reason to doubt the existence of collective belief, though I grant that it might turn out that joint commitment is false while collective belief is true. This in mind, I conclude by sketching out what I take to be a more difficult variety of joint commitment to challenge, involving a robust group epistemic agent whose belief is formed within a framework designed to maximize rational coherence, and consists in conjoined propositions individually contributed to in roughly the same way that Mary and Larry produce a major third. Though a compelling scenario, I claim that, even here, the special interaction of cognitive states that collective belief requires remains out of reach due to the internal and isolated nature of individual human minds.
Before getting started, a brief note to motivate the inquiry. This paper is meant to make a small contribution to broader discussions regarding the relationship between individual and group epistemic states. Such discussions address, for example: the extent to which individuals should be held accountable for group decisions; the nature of group knowledge; whether groups are smarter than their individual members; and what the social, jurisprudential, and political implications of our answers to these questions might be.[2] These discussions often involve liberal talk about group minds without developed consideration of important fundamental philosophical questions having to do with the ontological nature of group cognitive states—e.g., regarding individuals’ contributions to the formation and expression of group belief. I presume our understanding of this relationship will inform how we make sense of the broader discourse.
2. Joint Commitment
One of the most cited and systematic thinkers invested in such questions is Gilbert, who argues that ascriptions often do correctly refer to collective belief, in particular when group members are jointly committed to maintaining that belief. She has been developing this joint commitment account since the late 1980s. Consider a recent illustration:[3]
Joe meets Karen and, wanting to say something pleasant, comes out with “Lovely day!” Karen, wanting to be agreeable says “Yes, indeed!” Joe and Karen then come across Fred, who grumbles about the day’s weather. Karen confidently responds, on behalf of Joe and herself “We think it’s a lovely day!” Karen’s statement seems to be on target, as a statement of collective belief, irrespective of any personal beliefs of the parties regarding the weather.[4]
In the example, Karen and Joe jointly commit to uphold the collective belief that it’s a lovely day out. In honor of this commitment, Karen intends we to function collectively rather than distributively—i.e., to refer to the group rather than to herself and Joe as individuals. I find it difficult to accept that Karen would mean her polite statement as a literal assertion of belief, though if she did, it seems it would simply amount to a false assertion about what each of them personally believes. I’ll revisit this concern shortly.
Joint commitment comes with rights and obligations—namely, it cannot be rightfully rescinded unless all parties agree to do so. Gilbert nicely illustrates this point in the example of a reading group whose members have agreed that a particular poem is admirable:
Were the contrary personal views of the members to come out into the open after the formation of the collective belief it might cause some degree of shock or at least surprise. It is unlikely, though, to be seen as a rebuttal of the claim about what the group believes or to show that the group is ‘of two minds.’ [5]
Here, my above-noted concern is reignited, as this example seems to show that members referring to the group’s beliefs would likely do so distributively—thus the surprise upon learning about the contrary views of one or more fellow members. It’s also easy to imagine members expecting contrary views to surface, given the differing reasons why members might express agreement—e.g., giving in to loudmouthed members, politeness, apathy, having not yet read the poem. Nevertheless, according to Gilbert, any one member may hold the rest to their commitment, so long as all members have expressed agreement:
A joint commitment… is not something composite, a conjunction of the personal commitment of one party with the personal commitment(s) of the other(s). Rather, it is simple. A joint commitment is the creation of all the parties to it, rescindable only with the concurrence of all. Insofar as it involves a type of self-directed order, it involves an order issued jointly by all the parties to all the parties.[6]
I would find these examples more convincing had members collaborated on their group’s belief in an explicit effort to maximize rational coherence, an idea I’ll develop as I examine the three categories of group belief. First, some comments about methodology.
There seem to be two broad questions in play here, the first semantic, the second ontological: (1) What phenomenon do we aim to pick out when referring to group belief? (2) When we intend our group belief ascriptions to pick out essentially the same phenomenon we target when referring to individual belief, are we ever correct to do so? In short: Does collective belief exist?
I view (1) and related questions—e.g., do we intuit joint commitments?—to be empirical, in which case our appeals to intuition about them should be viewed as untested speculations, even though these appeals claim to have a handle one what we, as everyday speakers, intuit about our semantic intentions. Indeed, it’s possible that studies might conclude that our interactions often do amount to joint commitment scenarios, whether or not we report experiencing such intuitions (and whether or not collective belief exists).[7]
That said, while I’m particularly interested in (2), it’s difficult, if not impossible, to separate these questions.[8] Indeed, Gilbert often characterizes the semantic and ontological perspectives as different routes to the same question: “To what phenomena are ascriptions of collective cognitive states intended to refer? In other terms, what are collective knowledge, collective belief, and so on?”[9] Accordingly, while she grants that “it is possible that all or some ascriptions of cognitive states to [non-human entities] are metaphorical,” she maintains that “if such-and-such a phenomenon is referred to, seriously and literally, as belief, it is hard to argue that it is not, after all, belief as this is ordinarily understood.”[10] Gilbert largely leaves it to the reader’s intuition to make sense of “belief as this is ordinarily understood,” though she makes clear that this involves a cognitive state, and does give some idea of how she means to characterize that state. For individuals, this generally means the cognitive state experienced while having the thought that p is true, or at least while behaving consistently with such a thought.[11] This is the state that individuals agree to emulate when they engage in joint commitment—i.e., when they behave as though they are a single believing entity.[12]
I agree that groups can collectively behave as though they believe something. This is not enough, however, as there may be multitudes of arbitrary groups of two or more entities whose cumulative activity consistently affirms p—e.g., two strangers who never cross paths, yet whose day-to-day combined behavior accidentally forms a p-affirming regularity. I doubt, though, that anyone would ascribe collective belief to arbitrary sets of disparate individuals. This marks another critical difference between belief and collectivize-able activities: Mary and Larry could accidentally produce a major third while crossing paths on the street.[13]
Joint commitment eliminates arbitrary groups by requiring intentional collaboration. Still, it seems to me that for anything like collective belief to occur, individual belief must be collectivized according to conditions more stringent than what Gilbert’s account requires, more about which below, when I consider a weaker form of joint commitment. Few, if any, groups will satisfy such conditions, though they may fail to do so in different ways, depending on the nature of the group. Likewise, ascriptions to groups seem to function differently depending on the group in question. To parse this out, I’ll now examine three basic categories of group belief: casual-informal, formal commitment, and summative.
3. Three Categories of Group Belief
(a) Casual-Informal, in which group belief is formed informally, such as in the examples of Karen & Joe and the reading group. Characteristically, such groups have an at least group-agent-like constitution, though speakers generally don’t make explicit whether ascriptions to the group are meant collectively or distributively. I aim to show that, when intended collectively, such ascriptions are readily undermined by members’ personal beliefs.
Casual-informal belief is collaborative and thus non-arbitrary. Three unacquainted subway commuters who casually exchange shrugs confirming to one another that the announcement they just heard was unintelligible would count as such a collaboration. Had they shared that belief without exchanging belief-reinforcing confirmation, one might still ascribe the belief to the group, but only arbitrarily—as arbitrarily as one might ascribe belief to, say, everyone who’s ever believed themselves to sneeze on a Tuesday.
Collaborative examples are compelling due to the resemblance they bear to more obvious instances of irreducible activities, such as when three people carry a sofa. In fact, some aspects of group belief function just in this way: Karen and Joe may collectively author and accept p. For either to correctly say we believe p, however, each must believe it.[14] Gilbert disagrees, of course, and has relied substantially on casual-informal scenarios over the course of developing her account, which I shall now consider more closely.
Suppose Karen and Joe establish a joint commitment as described above. It seems rather easy a commitment to exploit in order to produce counterintuitive results. A simple example would have Joe tell Fred that he actually believes it to be awful out. It would be odd for Karen to then tell Fred, “Joe may believe it’s awful out, but we—Joe and I—think it’s lovely out!” It seems clear to me the group’s belief would be undermined by Joe’s assertion whether or not Karen exerts her right to maintain their commitment. One might expect her not to exert that right; I would agree, but I think it is precisely due to an understood lack of collective belief that we wouldn’t expect Karen to exert that right. This is the same reason, I think, that we would find it odd were Karen to say that she and Joe have the distributive group belief that it’s awful out, and the collective belief that it’s lovely out.[15] [16]
In a more complicated example, Karen commits to a second collective belief by agreeing with Fred when he says, “Well, I think it’s awful out.” Now, when Edna walks up, Karen may correctly assert, “Fred and I think it’s awful out, though Joe and I disagree, at least collectively; distributively speaking, Joe and I agree with me and Fred.” While these observations don’t disprove collective belief, they at least seem to threaten key intuitions relied on by the joint commitment account, particularly in casual contexts.
The reading group provides a more difficult case given its quasi-formal structure, an essential feature of which is to form a casual consensus about what the group believes. As noted above, dissenting beliefs appear to easily undermine the group belief. And, in fact, it’s not a stretch to imagine a controversial book leading one half of the group to “agree to disagree” with the other half, resulting in two group beliefs. Were the reading group’s deliberations to be properly formalized, however, no member would need to believe the group’s official position for that position to stand. This brings us to formal commitment.
(b) Formal Commitment, in which a formal framework—as with committees, tribunals, etc.—is in place for establishing group belief. Formal membership implies commitment to the group’s goals, intentions, and judgments (over time). As such, this is the category for which I find Gilbert’s account most illuminating, particularly when it comes to groups organized so as to maximize rational coherence. It seems, though, that when we say the judges believe the defendant to be liable, it would in most cases be more accurate to replace the word believe with judge, find, or accept. Once established, the tribunal’s judgment endures on record, even as its members change. Unless, that is, we mean that each judge personally believes the defendant to be liable, in which case those same judges may still find the defendant innocent (perhaps due to a technicality or bribery). Gilbert provides an example for contemplating this distinction:
Assume that there are two committees—say, the Library Committee and the Food Committee of a residential college—with the same members. It seems quite possible to say, without contradiction, that (a) most members of the Library Committee personally believe that college members have to consume too much starch, and this is common knowledge within the Library Committee; (b) the same goes, mutatis mutandis, for the members of the Food Committee; (c) the Food Committee believes that college members have to consume too much starch, whereas the Library Committee has no opinion on the matter.[17]
I take it that this example conflates differing connotations of belief: (a) and (b) refer to the personal beliefs of the members, while (c) refers to a judgment—or perhaps a series of judgments that establish a coherent position—made by the group in its capacity as Food Committee.[18] Committee members need not personally agree with the judgment, but must accept it in order to continue on as productive committee members; terms of acceptance presumably must conform to some codified procedure. Procedures may vary, of course—e.g., via a unanimous vote or even a coin toss—provided the group commits to the procedure.[19] [20] In order to elevate group judgment to group belief (much less collective belief), it seems that, among other things (more about which below), the group should strive for rational coherence of the sort expected of individual believers.[21] The challenge of achieving this is nicely represented by what List & Pettit call the discursive dilemma.[22] A typical example runs as follows. Three judges must rule on whether a defendant has committed a contractually forbidden act. To be considered liable, the defendant must have been contractually obligated not to commit the act, and must have indeed committed the act. The judges vote:[23]
| Obligation? | Act? | Liable? | |
| Judge 1 | True | True | True |
| Judge 2 | True | False | False |
| Judge 3 | False | True | False |
| Majority | True | True | ? |
The dilemma is whether to aggregate individual judgments according to a premise-based or conclusion-based model. Here, the former would find the defendant liable, the latter not liable. A common response is that groups should somehow determine which model to follow, and then stick to that model. Hélène Landemore, for example, argues that “deliberation about the best way to proceed” would allow decision-makers to commit to either approach as a matter of policy.[24] Kornhauser & Sager prescribe a case-by-case “metavote” in which members select “a particular method after discussing such factors as whether the outcome or rationales for it are more important, whether the issues to be decided are independent, the seriousness of the consequences of the outcome, hierarchical management concerns, and internal management considerations.”[25] This solution yields new difficulties. For one, a case-by-case policy seems to invite ad hoc strategizing for influencing outcomes. Moreover, there is the problem of interrelated premises arising over time, a possibility not addressed by Landemore or Kornhauser & Sager. This would seem to demand a premise-based model, which, largely due to this diachronic dimension, is what List & Pettit recommend.
Unsurprisingly, premise-based aggregation comes with challenges; for example, working out how to carve up complex sets of interrelated and overlapping premises, and determining in what order to address those premises—which, again, can lead to strategic manipulation. Another concern is that the deductive path from premises to conclusion is not determined by logic alone—indeed, even resolving which propositions count as premises and conclusion may be prone to bias. “One person’s conclusion may be another person’s premise,” as List & Pettit put it.[26] Such difficulties may urge us to organize propositions in less easily exploited arrangements, such as by adopting a modus tollens rather than modus ponens procedure (e.g., determining the status of one premise solely based on how the group votes on the second premise and the conclusion). Pettit explores this possibility[27], but he (along with List) ultimately endorses the premise-based model as the best solution as it leads the group to “‘collectivise reason,’ in the sense of guaranteeing that the attitudes of the group are consistent as well as derived from a rational process.”[28] [29] Perhaps this is correct, as it seems that honoring the judgments made on premises does greater justice to the group as a collective of rational individuals. At the very least, being cognizant of these issues, as opposed to merely relying on blind aggregation, upholds the idea that “cases are supposed to be decided on their merits rather than by an unconsidered choice of voting protocol.”[30]
Landemore dismisses theoretical examples of the discursive dilemma as “not particularly realistic takes on what happens in actual elections, which generally proceed on a conclusion basis.”[31] Significantly, though, her concern—and I presume that of Kornhauser & Sager—is about what is more reliable for getting to the right answer, while my concern is with what achieves rational coherence as a condition for robust group belief. Indeed, it may turn out that acceptance is more reliable than belief! List & Pettit, however, persuasively cite rational coherence and consistency as reasons for adopting a premise-based model, though, interestingly, they reject the most rationally coherent option—appointing a dictator—for its failure to maximize the epistemic advantages of group deliberation (in addition to being undemocratic).[32]
I belabor these difficulties in order to emphasize the challenge of determining what should count as a group’s attitude, even in formal contexts—such difficulties would be even harder to track in informal contexts. I’m also attempting to shift focus towards troubles with the epistemic (rather than semantic) grounds for collective belief.
A further threat to group coherence involves justification: a group would need only ignore or fabricate evidence to maintain whatever belief it likes. Jennifer Lackey has produced the example of a large tobacco company exploiting what Lackey characterizes as joint acceptance in order to maintain the group belief that smoking poses no health risks.[33] The upshot of this is that joint acceptance—which, as far as I can tell, functions essentially in the same way as Gilbert’s joint commitment model—cannot ground justification of group belief. Furthermore, group justification poses discursive-dilemma-like difficulties, such as when a group believes that p due to subgroups being aware of different evidence sets that, despite being mutually incompatible, imply that p. For instance, when all players on a team justifiedly believe a teammate is planning to sabotage a game, but subgroups within the team each justifiedly believes the saboteur to be a different person.[34] Lackey’s solution to these problems is compelling: for a group belief to be justified, the individual beliefs that contribute to that belief must be justified and jointly coherent, and the belief must stand up to a process of rational deliberation in which all p-relevant beliefs are disclosed.
This is in contrast to Gilbert’s account, in which the group emulates an individual believer presumably after formation of the belief, so that justification for belief doesn’t figure in—collective belief may be arrived at for the sake being polite, via coin-flip, or through rational deliberation. It seems to me, however, that to characterize a group’s position as belief (as it’s ordinarily understood), rational coherence should be a minimal goal; thus, such belief should involve something like a group epistemic agent engaged not just in emulating an individual believer in belief expression, but also in belief formation through a deliberative process designed to reconcile, for example, my belief that p and yours that not-p.
Group agency is a complicated topic that I don’t have space explore here. But I will note that I acknowledge the existence of group agents, more or less according to the criteria commonly cited in recent decades—namely, when a group of individuals intend to unify, as coherently as possible, their individual cognitive perspectives in the pursuit of a shared goal, thus making the group as a whole accountable for the group’s activities.[35] In the two categories considered so far, belief and judgment are ascribed to groups that have at least some claim to group agency (though, as has been amply demonstrated, collective activity does not require group agency; indeed, Mary and/or Larry need not even be human—a cat and a car alarm might accidentally produce a major third). Many ascriptions of group belief, however, refer to large groups that have no such claim. I consider these next.
(c) Summative, in which ascriptions of cognitive states to large groups amount to a shorthand for indirectly talking about the cognitive states of individuals. So, when I say Chicagoans prefer deep-dish to thin-crust pizza, I assume that many individual Chicagoans have that preference (though I myself am a Chicagoan who prefers thin-crust). The term summative is after Anthony Quinton:
Groups are said to have beliefs, emotions and attitudes and to take decisions and make promises. But these ways of speaking are plainly metaphorical. To ascribe mental predicates to a group is always an indirect way of ascribing such predicates to its members. With such mental states as beliefs and attitudes the ascriptions are of what I have called a summative kind. To say that the industrial working class is determined to resist anti-trade-union laws is to say that all or most industrial workers are so minded.[36]
This passage is often cited by proponents of collective belief as exemplifying the opposing view. Gilbert & Pilchman, for instance, offer Karen & Joe as an explicit counterexample to the above passage and thus as proof that there are good reasons “to reject all summative accounts of group belief,” concluding, “If, indeed, a group G can believe that p without any of its members believing that p, no form of summativism, however complex, can be right.”[37]
It seems, however, that Gilbert’s small-scale examples have no such implications for large groups.[38] I don’t see how Quinton fails to account for joint commitment, for example, when he notes, “to say that the French middle class is thrifty is to say that most French middle class people are.”[39] [40] Likewise, when we say Americans believe in God, we’re not imagining that Americans are jointly committed to emulating an individual believer. Rather, we mean that were we to do a reliable survey, we would find that many Americans believe in God. Were the survey to reveal the contrary, we would be pressed to stop ascribing that belief to Americans.
That said, I suspect that large-scale ascriptions often have less to do with belief, and more to do with the observable consequences of accumulated individual actions. To say that 60% of a population voted a leader into office is not to say that this group believed the candidate to be the best person for the job, though it’s convenient talk that way. It would be instructive, then, to consider the summative account in light of our differing intuitions about belief and behavior in a group context. Gilbert borrows a nice example from Emile Durkheim for doing so.[41]
Suppose an anthropologist writes, “the Zuni tribe believes that the north is the region of force and destruction.”[42] This seems fine until we learn that the basis for the statement is that each member of the tribe believes this proposition secretly out of fear of being mocked. Gilbert produces the case in order to demonstrate, by way of the anthropologist’s mischaracterization, that our ascriptions of group belief can’t be merely summative accounts of individual belief. I agree that the anthropologist has made an interesting mistake, but disagree that it works against summativism.
Namely, belief applies here in the distributive sense, which the anthropologist should have made clear. That is, there’d be no problem with saying that all or most Zuni tribe members believe that p, yet do not behave as though they believe that p. The original wording, by using the singular form Zuni tribe, suggests that all or most of the tribe members both hold the belief in question and, most significantly, act in accordance with that belief. Thus, confusion arises due the original statement’s implication of group behavior on the one hand, and the secrecy’s implication of a lack of that same group behavior on the other.
Notice that this distinction between belief and action would obtain even if tribe members did openly express their belief, but is made salient by the members’ lack of belief-expressing activity. This exposes a telling disconnect between individual and group states, such that individuals can hold a belief internally, without acting on it, while a group cannot; put another way, internally held individual beliefs, if not acted upon, do not result in group belief of any sort, figurative or literal. So, what we are often really referring to in ascriptions of group belief is an accumulation of individuals outwardly expressing some belief (or desire or other internal state), rather than the accumulation of the (internal) activity of believing.
These results support Gilbert’s assertion that large-scale ascriptions aren’t merely about aggregating individual belief. I propose, then, that we should refine the summative view with the thought in mind that, while some large-scale ascriptions are indeed indirect ascriptions of individual beliefs, others are more concerned with the behavioral disposition of a group composed of a large number of (more or less) mutual strangers.
If that disposition is referred to as belief, it’s only done so as a way of saying that the group’s accumulative activity makes it appear as though the group possesses a coherent attitude. Should it be revealed that many of the aforementioned voters in the 60%-group had voted against their actual preferences for strategic purposes, it would become difficult to talk even figuratively about what that group “believes.” But we can talk about the group’s accumulative activities and the outcomes thereof. It seems, then, that our everyday large-scale ascriptions are often aimed at talking about a highly complex network of social interactions that cannot be detailed in casual conversation (if at all). So we summarize.
Sometimes our summative ascriptions are reducible to individual beliefs, as when we say Zuni tribe members believe that the north is the region of destruction or Americans believe in God, and other times they are reducible to individual actions, as when we say Zuni tribe members hide their beliefs or the French spend thriftily. When large-scale ascriptions aren’t reducible, as in the case of a candidate being elected, the group’s activity may supervene on individual behaviors, but on my (and seemingly Gilbert’s) view, not on individual beliefs.[43] Simply put, belief ascriptions that refer to genuine collective activity are best understood as shorthand for talking about a group’s behavioral disposition, surmised from the observable results of accumulative action.
To stress this point: What makes an activity collective is not intentional collaboration, but rather its multi-authored outcome. This outcome must be something observable—i.e., that appeals to the mental structure of the human mind—as a singular thing. We ascribe the major third to the group comprised of Mary and Larry because we know who composes the group. But what marks the behavior as collective is that the major third is observable as a phenomenon not produced by an individual; we make no such ascriptions when Mary blinks while Larry sneezes. Given that we cannot observe very large groups, we’re left only with outcomes (e.g., rising milk sales) as evidence according to which we make inferences about collective behavior and the attitudes that motivate that behavior.
So far, I have attempted to cast doubt on the existence of collective belief by examining how our belief ascriptions function and fare in a range of group contexts. I’ve used Gilbert’s joint commitment account as a framework, or starting point, for doing so.[44] However, while this may encourage skepticism about collective belief in many contexts, this is not enough to show that collective belief doesn’t arise under the strictest possible formal conditions—namely, when a group agent intends to emulate an individual believer by taking into account rational coherence, justification, and so on, while engaged in the formation of a belief that functions as much as possible like other sorts of collective phenomena. In other words, I have not yet shown that collective belief would be prohibited were steps taken to correct the points of criticism I’ve so far noted. Next, then, I consider a form of joint commitment strictly organized with that goal in mind.
4. A Tougher Case
One of the challenges here is in being clear about what sorts of phenomena we aim to pick out when referring to belief. Right now, I believe I’m sitting in a room in Brooklyn, writing these words. Outsiders may reasonably infer that I have this belief insomuch as my behavior gives them good evidence for doing so; but, according to our ordinary understanding, the inference isn’t correct unless I either have, or am disposed to have, one or more mental representations that treat the proposition as true.[45] In other words, there is something that it is like—a kind of felt confidence—for me to believe I’m writing these words; and I gather that this experience is essential to what we aim to pick out when referring to individual belief, and is what grounds our (figurative) belief ascriptions to mindless entities (whether for convenience, aesthetic effect, or as a predictive tactic).
This isn’t to say that behavior doesn’t play an important role in belief. Tom might sincerely claim to believe that p until noticing his behavior contradicts that claim.[46] It’s tempting, then, to view the phenomenological dimensions of belief as incidental, and thus to define belief as something like: An agent believes that p when exhibiting a pattern of behavior that consistently affirms p.[47] Notice, however, the priority granted to the internal when we view Tom’s belief to have been properly formed once he’s aligned his thoughts about p with his behavioral evidence. We often demonstrate this priority. We recognize that someone’s “true” belief could be exposed after years of secrecy, and that people often act against their genuine beliefs, such as the heavy smoker who believes smoking to be a terrible idea, or the student who regrets repeatedly failing to speak up in class.[48] It also strikes me that Karen and Joe’s behavior would be better viewed as the collective performance of a lie about the group’s distributive belief, precisely because it is at odds with their genuinely held internal beliefs. I also take as a signal of experience’s priority that it is uncontroversial to say that a thoroughly paralyzed human with an active mind can form beliefs, while it is controversial to ascribe literal belief to a mindless zombie.
Groups, being mindless, cannot experience belief. This fact strikes me as posing an insuperable obstacle to collective belief, though perhaps I would need to show more than I have so far in order to hinge belief on experience, particularly if I wish to persuade proponents of collective belief for whom group emulation of individual behavior is enough to produce essentially the same phenomenon we aim to pick out when referring to individual belief: as Gilbert puts it, “if collective beliefs lack the allegedly essential feature of belief, [one may argue that] that throws doubt on the claim that this feature is indeed essential to belief.”[49] I might, then, need to do a better job of showing, for example, that Tom’s belief that p cannot be sustained by his behavior alone in case no thoughts about p ever occur to him, or should he turn out to be a zombie. This would be a lengthier undertaking than I have space for, so I will go forward presuming a more modest picture: when thinking persons believe, their belief is most robustly formed when they have introspected as a kind of “self-deliberation,” taking all relevant evidence (their patterns of thought, expressed attitudes, etc.) into consideration. I presume that a belief thus formed would provide a paradigmatic, or at least utterly uncontroversial, case of what we aim to pick out when referring to individual belief.
I claim that, while a group may collectively emulate the introspective process, the separation of human minds prohibits the formation of uncontroversial collective belief. [50] To see this, consider what I take to be a weaker form of joint commitment, one that might more plausibly result in collective belief. Namely, when a group agent collectively authors and expresses a belief whose ontology functions like other sorts of collective phenomena (e.g., a major third); which is to say that it would require at least two distinct conjoined propositions, say p&q, where roughly half the group believes that p while merely accepting that q (or vice versa); and the rest of the group believes that q while merely accepting that p (or vice versa); and any given member accepts that p&q, though does not believe that p&q.[51]
Further conditions would be required; at least: collaboration of a group agent; rational exchange for the sake of ironing out inconsistencies; a solid justification norm such as that described by Lackey; passivity so that the group’s belief is involuntarily sensitive to new information. There would also need to be some qualifications about what sorts of propositions may be conjoined in this context. Namely, some way of demonstrating the capacity of p and q, by way of some potential for mutual compatibility or incompatibility, to compose a unified phenomenon or “composite belief.” P¬-p would not constitute a valid collective belief, for obvious reasons. Nor would, say, the disparate propositions humans need air to survive–&–this table is blue; for one thing, before collective authoring begins these would likely already be believed by most or all members of a group seated at a blue table; more to the point, one of these propositions being believable or acceptable has no bearing on the other’s being believable or acceptable. Humans need air to survive–&–this clean-air policy is important might be fine, given that if humans did not need air to survive, the clean-air policy might no longer be seen as important. This may seem strict, but I think it no stricter than the natural distinction between Mary and Larry sing a major third (results in a meaningful collective phenomenon) and Mary emits an E while Larry reads a book (does not so result). Indeed, I expect that collective belief would demand more—including group agency—than do other sorts of collective activities, in order to bridge the phenomenological (or at least cognitive) gap between minds that prohibits the collectivization of belief-as-experience.[52]
I find this weaker account compelling, though it yields what may appear to be a strange result: collective belief requires that all (or most) group members explicitly disbelieve the composite proposition. Why not count it as collective belief when all members share the belief? To do so wouldn’t say anything interesting about the group’s belief, and in fact would be robustly summative, as there would be no discernable difference between collective and distributive belief. It would also make it wholly unlike other sorts of collective activities, in which the group’s ascription floats above those of its members. For belief to work in the same way, collective beliefs must reside in the intersection of things that are beliefs and things that are collective phenomena. And so, even if this result seems strange, it poses no threat to collective belief, except insomuch as it is not a goal towards which a group can strive, as the members’ attitudes towards the composite belief will be involuntary. If they all happen to believe the composite, there is no collective belief. Perhaps, though, there may be a subgroup among them that collectively believes. I won’t pursue this here, but will instead move on to certain differences between belief and collective phenomena that stoke my skepticism about anything residing in their intersection.
In general, collective events—a sofa being carried, rising gas prices, a major third—involve some observable phenomenon resulting from collective activity (one sees the sofa being carried, graphs the rising gas prices, hears the major third).[53] Individual belief cannot be examined with the external senses, which suggests that an inability to observe collective belief should not be enough to impede its formation or cast serious doubt on its existence, though this leaves us with the problem of how to verify collective belief’s presence.[54] Perhaps the group members’ participation in the belief’s authoring and its dynamical constitution over time (i.e., when it changes in response to new information) is the thing that is experienced by group members—is the collective belief. This seems plausible enough, though it’s not clear to me how the group could track, much less intuit, such a thing given the strictures of the belief conditions (namely, knowing that members collectively satisfy the belief conditions in relation to p&q would be an unattainable analog to an individual’s felt confidence about p).
There is a further, related, worry. Members’ behavior will only count as belief if they jointly intend their behaviors to be counted as belief. But that same behavior could occur with several members acting insincerely. Only a given member knows if that particular member acts sincerely. Given this indiscernibility between genuine and false collective belief (i.e., the performance of a lie) to the group itself—something I presume, crucially, not to be true of individual belief—I remain skeptical about collective belief. However, I do find this weaker account of joint commitment to pose a greater challenge to my skepticism.
5. Implications and Conclusion
I hope the present inquiry is not only of philosophical interest, but also contributes to discussions about group decision making more broadly. Lackey’s manipulative tobacco company is a clear example of the topic’s relevance to member accountability. Furthermore, the importance of deliberation in fortifying group belief may have significance for whether representative deliberative democracy is a smarter—i.e., more knowledge-conducive—form of government than dictatorship or direct democracy. If (p&q)àr, and I have p in my head and you have q in yours, it’s through deliberation that the group realizes r. Such revelations are unlikely to emerge from an arbitrarily connected mass of millions of people. Similar observations lead Landemore to suggest that political decisions are best made by cognitively diverse representative assemblies, no greater than 400 to 500 people in size.
Deliberation also has significance for notions such as the “miracle of aggregation” and the “wisdom of crowds.” These are often offered as the smartest way to make decisions, according to models in which large numbers of individual judgments are aggregated, and in which deliberation need not play a role. Indeed, deliberation may be discouraged for the sake of voter independence.[55] This approach is often supported by proofs such as Condorcet’s Jury Theorem, which is meant to show that the larger the group, the greater likelihood it has of getting the right answer, provided certain conditions are met, among them being that each member has a greater than random chance of getting the right answer. This is problematic for several reasons. Relevant to the discussion here is that we might be wary of epistemic models that subsume large masses of people into a “group mind” that blindly makes decisions whose consequences are ultimately experienced—i.e., suffered or enjoyed—by individuals, not by mindless aggregates. Also of significance is that this model presumes a majoritarian standard for problem solving that may not be appropriate for highly technical or ethical questions. This is why Landemore, who relies somewhat on Condorcet’s theorem, only commits her theoretical models to addressing purely political questions. I would think, though, that the most important and challenging political questions are those that have significant technical or moral dimensions: Should there be a minimum wage and, if so, how much should it be? Should the U.S. open its borders to the rest of the world?[56] How should we go about regulating bankers’ compensation levels?[57] Finally, while it’s claimed that aggregation cancels out individual mistakes, it also cancel out successes and any other ideas that deviate too far from the mean, thus silencing minorities purely by force of greater numbers.[58]
To conclude, I hope to have at least cast doubt on the possibility of informal groups of any size constituting collective believers. Perhaps this gives us reason to take seriously the possibility that genuine belief either always comes down to individuals or, at most, to robust group epistemic agents. However, even if the activities of such an agent never amount to genuine collective belief, its example provides an instructive standard for evaluating claims about group believers and for setting limits on their activities (e.g., in proportion to their rational capacities). Finally, I would like to emphasize that these considerations are important to keep in mind in an era of increased interest in aggregation, big data, crowdsourcing, corporate agency (to what extent should individuals be held liable for a group’s belief?), and in which “collective mind” language is so freely used in the promotion of those ideas.
REFERENCES
Audi, Robert. (1994). “Dispositional Beliefs and Dispositions to Believe.” Noûs 28.04: 419-434
Bird, Alexander. (2014). “When Is There a Group that Knows? Distributed Cognition, Scientific Knowledge, and the Social Epistemic Subject.” In Lackey, J. (ed), Essays in Collective Epistemology (42-63). Oxford: Oxford University Press.
Churchland, Paul. (1981). “Eliminative Materialism and the Propositional Attitudes.” Journal of Philosophy 78: 67–90.
Churchland, Patricia. (1986). Neurophilosophy: Toward a Unified Science of the Mind-Brain. Cambridge, MA: MIT Press.
Clark, Andy & Chalmers, David (1998). Analysis 58:10-23.
Durkheim, Emile & Mauss, Marecel. (1963). Primitive Classification. Chicago: The University of Chicago Press.
Estlund, David. (2008). Democratic Authority: A Philosophical Framework. Princeton, NJ: Princeton University Press.
Gilbert, Margaret. (1987). “Modelling Collective Belief.” Synthese 73: 185-204.
Gilbert, Margaret. (1989). On Social Facts. New Jersey: Princeton University Press.
Gilbert, Margaret. (2004). “Collective Epistemology.” Episteme 1: 95-107.
Gilbert, Margaret. (2014). Joint Commitment: How We Make the Social World. See “Belief and Acceptance as Features of Groups” (131-167). New York: Oxford University Press.
Gilbert, Margaret & Pilchman, Daniel. (2014). “Belief, Acceptance, and What Happens in Groups: Some Methodological Considerations.” In Lackey, J. (ed), Essays in Collective Epistemology (167-188). Oxford: Oxford University Press.
Gilbert, Margaret. & Priest, Maura. (2013). “Conversation and Collective Belief.” Perspectives on Perspectives on Pragmatics and Philosophy 1: 1-36.
Goldman, Alvin. (2014). “Social Process Reliabilism: Solving Justification Problems in Collective Epistemology.” In Lackey, J. (ed), Essays in Collective Epistemology (167-188). Oxford: Oxford University Press.
Kornhauser, Lewis & Sager, Lawrence (1993). “The One and the Many: Adjudication in Collegial Courts,” California Law Review 81.01: 1-59.
Lackey, Jennifer. (2016). “What Is Justified Group Belief?” Philosophical Review Volume 125, Number 3: 341-396
Landemore, Hélène (2012). Democratic Reason: Politics, Collective Intelligence, and the Rule of the Many. New Jersey: Princeton University Press.
List, Christian & Pettit, Philip. (2011). Group Agency: The Possibility, Design, and Status of Corporate Agents. Oxford: Oxford University Press.
Pettit, Philip. (2011). “Groups with Minds of Their Own.” In Whitcomb, D. & Goldman, A. (eds), Social Epistemology: Essential Readings (242-268). New York: Oxford University Press.
Quinton, Anthony. (1975/76). “Social Objects,” Proceedings of the Aristotelian Society 75: 1-27.
Rousseau, Jean-Jacques ([1762]/1997). Gourevitch, V. (ed), The Social Contract and Other Later Political Writings. Cambridge: Cambridge University Press.
Sunstein, Cass (1999). “The Law of Group Polarization.” The Law School of The University of Chicago, The Public Law and Legal Theory Working Papers series: John M. Olin Law & Economics Working Paper No. 91 (2nd Series)
Surowiecki, James (2004, 2005). The Wisdom of Crowds. New York: Random House, Inc.
FOOTNOTES
[1] There are more radical views than those considered here. E.g., Bird’s (2104) distributed model counts within the group-epistemic-agent any entity essential to the cognitive activities required of a given task. So, the agent that knows how to steer a ship includes the crew, mapmakers, the ship itself, etc. I take Bird’s account to also be more radical than others that extend cognition beyond the body—e.g., that of Clark & Chalmers. On their view, Otto’s notebook is an extension of his mind, and perhaps of his self, not an entity with which Otto forms a group agent (Clark & Chalmers 1998).
[2] A popular example is (Surowiecki 2004). For a political science oriented discussion see (Landemore 2012).
[3] I refer to “Gilbert’s” account with sensitivity to the fact that some of her work is co-authored.
[4] (Gilbert & Pilchman 2014: 195)
[5] (Gilbert 2004: 98)
[6] Ibid, 100
[7] Discerning what we mean in these contexts would likely require more than asking. On-the-spot attempts at self-explanation might run counter to semantic and other behavioral practices.
[8] I take (2) to be open to philosophical reflection, including when it takes the reductive form: What is belief, if it exists at all? Belief isn’t something you can point to like a stone, distant cloud, or doughnut hole.
[9] Ibid, 96
[10] Ibid, 103
[11] See especially (Gilbert 2004: 95, 103) and (Gilbert 2013: 3).
[12] “The parties will fulfill their joint commitment if—at least in the context of the present interaction, and to the extent that this is possible, they emulate, by virtue of the several actions and utterances of all combined, a single body that believes that p. In more familiar terms, they are to act as if they are ‘of one mind’ with respect to p” (Gilbert & Priest 2013: 11).
[13] Proximity isn’t required, as when Mary and Larry are strangers located in separate rooms simultaneously bugged by the CIA, and they happen to produce a major third (heard by a surveilling CIA agent).
[14] In larger groups, some significant portion of members, though not all, must believe p. I won’t attempt to resolve this vagueness here.
[15] One might also think that Karen’s behavior here is not that of someone committed to emulating a single believer. It seems to me, though, that so long as she makes clear when she is expressing her personal belief and that of the group, she honors the parameters of joint commitment, given that she and the group are taken to be distinct epistemic entities.
[16] This brings to mind Rousseau’s distinction between the will of all, which is the sum of particular individual interests, and the general will, which is what is left over after “the pluses and the minuses” within the will of all have cancelled one another out (Rousseau [1762]/1997: 60). (Quinton 1975/1976) touches on this briefly. For a more developed interpretation, see (Landemore 2012: 69), who discusses it in the context of “the emergent phenomenon of collective intelligence.”
[17] (Gilbert 1987: 189)
[18] Gilbert sometimes refers to collective belief*. The asterisk denotes agnosticism about the genuineness of collective belief. The most viable alternative, on her view, is that collective belief talk is really about acceptance. She rejects this, however, on the grounds that it characterizes joint commitment as willing a belief into existence, something groups don’t do. That is, it’s the act of creating the belief that forms the group—pre-belief, there was no group to do any willing (Gilbert & Pilchman 2014: 202).
[19] David Estlund (2008) has pointed out that a coin toss may be the fairest way to make political decisions, which suggests that fairness in itself is not sufficient justification for democratic procedures. Instead, we need an epistemic justification; namely, we need to show that democratic bodies yield better-than-random results.
[20] Deciding on procedure invites a problem faced by voting bodies in general: How do you establish a decision-making structure in the absence of an established decision-making structure? I bring this up to emphasize that if it’s difficult to establish group attitudes in formal situations, it must be more so in casual-informal cases.
[21] Individual belief may contain inconsistencies; e.g., believing that p&q when (pàr) and (qà~r) may be tough to avoid. But not occurrently believing that p&~p is a minimal expectation of individual rationality.
[22] The discursive dilemma generalizes a jurisprudential concern identified by Kornhauser & Sager (1993) as the doctrinal paradox, which applies specifically within a legal context, while the discursive dilemma applies to any group decision-making setting.
[23] Table is from (List & Pettit 2011: 44).
[24] (Landemore 2012: 189)
[25] (Kornhauser & Sager 1993: 1)
[26] (List & Pettit 2011: 57)
[27] (Pettit 2011: 249)
[28] (List & Pettit 2011: 58)
[29] For further elaboration, see (List 2011).
[30] (Kornhauser & Sager 1993: 1)
[31] (Landemore 2012: 189)
[32] (List & Pettit 2011: 53)
[33] (Lackey 2016)
[34] I intend the example to be a simplified variation on Alvin Goldman’s British Museum case (Goldman 2014: 15), which Lackey utilizes in order to demonstrate the problem of justification within aggregative models.
[35] See for example (Rovane 1997: Chapter 4 Section 3) and (List & Pettit 2011: 33, 176).
[36] (Quinton 1975/1976: 17)
[37] (Gilbert & Pilchman 2014: 195)
[38] To be clear, although I generally find the term summative most useful for referring to large groups, I also think it appropriate to say that we speak summatively when ascribing belief to small groups.
[39] (Quinton 1975/1976: 9)
[40] This observation is also detectable in (List & Pettit 2011), who support a “realism under which group-agency talk is non-redundant, while remaining faithful to methodological individualism” (10). That is, they posit group agents (composed of relatively small numbers of people) as literal entities, while maintaining that large-scale social phenomena are attributable to individuals acting independently. Thus, referring to “what the market expects” is “metaphorical shorthand” for referring to “expectations behind individual investments” (2). They do seem to agree with Gilbert’s account, however, insofar is it involves a group agent in at least a quasi-formal setting (the Karen & Joe case would, I presume, count as an “over-ascription of agency” (216)). I wonder, though, if they’re hedging in their support of collective belief given their tendency to substitute the words belief and desire with judgment and preference, due to a group agent’s lack of mental properties—i.e., “representational and motivational attitudes” (42).
[41] (Gilbert 1987: 187) and (Gilbert 1989: 257)
[42] (Durkheim & Mauss 1963: 44)
[43] This will be true whether belief turns out to come down to mental representation, external behavior, or some combination of the two.
[44] In summary: casual groups are ill-equipped to account for the constraints of rationality, and collective ascriptions to such groups are easily undermined by the views of dissenting members; formal group belief is, in general, better seen as judgment or acceptance; large-scale groups seem to have little to do with any form of joint commitment, are arbitrary given that most group members are mutual strangers, and are undermined when it turns out that most group members don’t hold the belief in question.
[45] Robert Audi (1994) convincingly claims that beliefs are either occurrent (i.e., in working memory), or are dispositional in the sense that they’re stored in one’s long-term memory awaiting retrieval; a third category consists in the occurrent beliefs one is disposed to form, should the appropriate situations arise.
[46] To determine what we believe, we might first need to contemplate our salient behavior, ongoing attitudes and such; which of those are made salient would likely depend on context—e.g., on who’s asking. Perhaps, though, we simply don’t form belief until prompted. If so, then belief really does come down to the sincere thought that p is true, and we may believe that p or not-p depending on when asked (and thus be disposed to believe that p or not, depending on context). I won’t press this, but I find it appealing.
[47] By phenomenological, I mean that which involves mental representation, or experience.
[48] Perhaps these involve competing beliefs and it’s not clear what, if anything, the “true” belief is.
[49] (Gilbert, 2004: 103)
[50] Though I won’t pursue it here, I also reckon experience to be indirectly crucial for belief, as belief does not exist in isolation, but rather comes with desires and other emotions: to believe something is saddening is, at least to some degree, to feel saddened.
[51] Acceptance of q here means that you have come to accept that q after deliberation has demonstrated both that (1) q is compatible with your belief that p and (2) a significant portion of your collaborators believe that q.
[52] It may be that conditions would need to be even stricter than what I’ve described. For example, perhaps every member would need to uniquely contribute a proposition only believed by that member, but accepted by all as a constitutive element of the final composite belief.
[53] I cannot think of a clear counterexample to this.
[54] The implications of this lead some to propose that belief doesn’t exist, not even for individuals. For compelling views in this direction see (Churchland 1981) and (Churchland 1986).
[55] Sunstein (2002), for example, argues that deliberation polarizes rather than resolves.
[56] I find this one particularly interesting given that, in principle, posing this question to the global population would maximize the benefits of large numbers.
[57] This example is offered as a purely political question in (Landemore 2013: 97).
[58] Condorcet’s theorem may be incompatible with groups small enough to deliberate. Recent research suggests that group deliberation spreads bias, leading subjects to share, rather than cancel out, one another’s mistakes—see (Surowiecki 2004: 255) on groups estimating the number of jelly beans in a jar.
Enjoy or find this post useful? Please consider pitching in a dollar or three to help me do a better job of populating this website with worthwhile words and music. Let me know what you'd like to see more of while you're at it. Transaction handled by PayPal.Or click the banner to shop at Amazon (at no extra cost: it just gives me some of what would have gone to Amazon).