Group Belief: The Cognitive Non-Summative Account

Abstract

This article concerns the criteria for when a group can collectively hold a belief. By proposing a cognitive non-summative account of group belief (GB), I highlight three necessary features at the individual level: commonality, mutuality, and group-based considerations. My account asserts that group G believes proposition p if and only if a sufficient majority of its members believe (1) p_G, where p_G is “Given some G-based considerations, p,” and (2) “the majority of G also believe that p_G.” The article critiques three significant accounts of GB: Anthony Quinton's simple summative account, Margaret Gilbert's and Raimo Tuomela's conative non-summative views, and Jennifer Lackey's modified summativism.

Résumé

Cet article examine les critères qui déterminent quand un groupe peut collectivement détenir une croyance. En proposant une approche cognitive non-sommative de la croyance de groupe (CG), je souligne trois caractéristiques nécessaires de cette croyance : le fait que la croyance est partagée par tous les membres du groupe (« commonality » en anglais), la mutualité et les considérations basées sur le groupe. Ainsi, le groupe G croit en la proposition p si et seulement si une majorité suffisante de ses membres croit que (1) p_G, où p_G signifie : « étant donné certaines considérations basées sur G, p », et que (2) « la majorité de G croit également que p_G ». L'article examine trois théories de la CG, soit la conception sommative simple d'Anthony Quinton, les approches conatives non-sommatives de Margaret Gilbert et Raimo Tuomela, et enfin la conception sommative modifiée de Jennifer Lackey, avant de montrer leurs limites.

1. Introduction

Groups are typically considered as epistemic agents capable of holding beliefs. We attribute beliefs to groups and address them to describe or anticipate their behaviours or to praise/blame them for holding those beliefs. But what does it mean for a group to believe? On the one hand, group beliefs are, unsurprisingly, beliefs and must share properties with other beliefs. On the other hand, group beliefs are distinctive in that they are attributed to more than one epistemic agent. Consequently, any proper account of GB must encompass both the general features of any belief and the particular criteria of GB that distinguish it from individual beliefs.

Regarding these primary desiderata, Section 2 of this article will emphasize taking p as true as the basic cognitive dimension of any belief. Then, it will propose two essential conditions of GB, derived from summative and non-summative accounts, which represent opposing camps in the literature. Finally, I will suggest a third condition that is necessary for every GB and enables us to integrate the first two conditions into a unified account.

Equipped with the foregoing criteria, Section 3 will outline a new account for GB, namely the “Cognitive Non-Summative Account,” and will respond to some potential concerns about it.

2. Toward the Cognitive Non-Summative Account of GB

Beliefs, in philosophy, refer to particular mental states or attitudes. Attitudes that are expressible by propositions are called “propositional attitudes” (PAs). Beliefs, however, are usually considered as distinct types of PAs. This distinction is typically formulated by the concept of truth-directedness. Bernard Williams, for instance, famously posits that “beliefs aim at truth” and means that (a) we ascribe true and false to beliefs, and (b) “to believe that p is to believe that p is true” (Williams, 1973, pp. 136–137). While (a) leads many epistemologists toward normative interpretations of truth-directedness (i.e., you should believe what is true), (b) allows for more phenomenological interpretations, which are more relevant to this article. When I believe that p, I am dealing with p as true. But what exactly does it mean to deal with something as true?

To believe that p is something between merely seeing p as true and taking p as true in non-cognitive senses. It is not reducible to seeing p as true because I may experience the Müller-Lyer illusion, for example, even after measuring both arrows and then believing that they are equal. It is not extendable to conative senses of taking p as true either,¹ since one may take p as if true without believing that p. Suppose I am playing chess with a friend while my queen is captured. Her pawn has reached the promotion square, and she wants to replace it with a queen. We both take it that, from now on, my black queen is her second white queen. We accept that agreement and commit ourselves to that throughout the game. Nevertheless, it does not make us believe that the black queen is a white queen. Many manifestations of taking p as true, such as accepting, assuming, or considering that p, can be associated with either belief or non-beliefs — e.g., pretense, obedience, lies, deception, and bullshit. This shows why we should consider belief as a subset of acceptance but not the converse (Bratman, 1999; Wray, 2001). We accept propositions either out of belief or for the sake of action. Consequently, epistemologists typically offer a second-order, reflective sense of taking or regarding p as true as a definition of believing that p (Leitgeb, 2017; Schwitzgebel, 2019; Velleman, 2000). In this sense, what a rational agent's action implies is his acceptance, and does not necessarily represent his belief.

Besides and based on the preceding general feature of any beliefs, GBs have some exclusive features as well. For example, if we attribute a belief to a group in a non-metaphorical sense, we must explain what it is for a group to believe that p, more than the aggregation of people who each individually believes that p. Let us first consider the two following scenarios as our central cases of GBs throughout this article:

1. (a) Bahar and her nine other friends believe that wearing jeans is cool. They used to wear dresses, but since they have made friends and are hanging out with each other they believe that jeans are cooler. Should you ask them “Why do you wear jeans all the time?,” they will answer “Because we believe they're cool!”

2. (b) Bahar and her nine other friends want to camp somewhere this weekend. After talking together, they come to the decision of going to Banff. Should you ask them “Why Banff?,” they will answer, “Because we believe it is the best option.”

Accounts of GB usually deal with type-b examples in which the members have active agencies in forming and holding the GB in question. However, a comprehensive account should be able to capture a-type examples as well.² Let me claim that we are faced with GBs in both foregoing cases. In what follows, I will spell out the main conditions of GBs based on these cases.

2.1. Commonality of GB

Accounts of GB vary based on constituency and structure, among other factors. Constituency refers to the question, “Who contributes to forming a GB that p?”³ while structure examines the question, “How do the constituents form the GB that p?” By “structure,” I address two further questions: whether they should hold the same attitude (i.e., belief) and whether they should hold the same content (i.e., p).

Regarding the structure, many social ontologists believe that a group cannot hold an attitude φ entirely independent of the group members’ having φ-based attitudes (List & Petit, 2011; Ludwig, 2016; Tollefsen, 2015; Tuomela, 2013). It sounds strange, for example, to say that group G is angry that p while no member of G is angry or the majority are indifferent or happy that p. If this is the case for all kinds of group PAs, why make an exception for GBs? Bahar and her friends, for example, cannot hold that “Jeans are cool” or “Banff is the best place to camp,” as their GB, without having a certain number of constituents holding the same content as true, given their group-based considerations. Therefore, I suggest the first condition of GBs as follows:

Condition 1: For group G to believe that p, a sufficient majority of its members (M) must share a belief that takes p as true.

The most straightforward reading of this condition is to say that a group's belief supervenes upon the summation of its members’ beliefs with the same content. In other words, group G believes that p if and only if most members of G believe that p. Following Anthony Quinton's articulation, collective epistemologists usually refer to this account as the “summative account.” In “Social Objects,” Quinton begins with an ontological claim that individuals are “no more, and arguably less, dependent on social objects than the latter are on them” (Quinton, 1976, p. 15) and proceeds to an epistemological conclusion that,

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. (Quinton, 1976, p. 17)

As of today, one of the most influential criticisms of the summative accounts is formulated by Margaret Gilbert (1987, 2004). Gilbert argues that the summation condition is neither necessary nor sufficient for having a GB (Gilbert, 2004, pp. 97–98). It is not sufficient for two reasons. First, she provokes us to imagine a situation in which all the members of group G individually believe that p but there is still no collective belief since no one is aware of the others’ beliefs. Suppose, for example, that Bahar and all of her friends personally believe that “Banff is the best place to camp,” unbeknownst to each other. Should you ask Bahar “Where do you want to go?,” she might plausibly answer, “I don't know. We have not yet come to any decision.” It is worth noting that this is not a case of individual ignorance since every other member of their group would answer the same. There is no we-mode in Bahar's answer because she does not believe that the others also have the same belief. This shows that the members of the group must be mutually aware that p is their collective belief. In other words, there must be some sort of we-mode or mutuality at work among the group's members about p.

The second reason Gilbert advances against the sufficiency of the summative condition relies on “an example in which two (or more) distinct groups have the same members,” where one group may believe that p, while the other withholds (or even disbelieves) that p (Gilbert, 2004, p. 98). Imagine, for example, an art competition being held in a small town, where the same panel of experts serves on two distinct juries: the Modern Art Jury and the Classical Art Jury. A piece of artwork is submitted to both the modern and classical sections. Gilbert believes the fact that one jury can find it worthy of an award while the other rejects it, despite both being consisted of the same members, shows that GB is not determined by the members’ individual beliefs.

Gilbert persuasively demonstrates that the summation of beliefs is not sufficient to have GB. Therefore, for group G to believe that p, an additional condition must be met as well. Next, we formulate and interpret the second condition.

2.2. Mutuality of GB

Taking Gilbert's criticism of summativism into account, we need another necessary condition of GB granting the property of mutual or collective awareness in each case of GB:

Condition 2: For group G to believe that p, the members of M must be mutually aware that the sufficient majority of G shares a belief that takes p as true.

However, this condition is also prone to misinterpretations. One common misreading of this condition is to put it in contrast to Condition 1, deeming it unnecessary. The accounts that suggest such misreading must substitute “belief” with another attitude shared by the group's members. Many of these accounts replace “belief” with a conative attitude required for group decision-making. I call these accounts the “conative non-summative accounts” of GB. The main figures of this camp are Gilbert and Raimo Tuomela, who deny the necessity of Condition 1 and substitute sharing the attitude of “belief” with that of “commitment” or “acceptance,” at the individual level, respectively.

Along with calling Condition 1 insufficient, Gilbert maintains that it is also unnecessary because one can find cases of group compromise in which most of the group G's members, or even all of them, disbelieve that p, yet they simultaneously take p as G's belief. Recall case (b) where Bahar and her friends believe that p, where p is “Banff is the best place to camp.” According to Gilbert, it is conceivable that p is no one's individual belief, but that they collectively come into believing that p because everyone's first option is totally unacceptable to others.

Rejecting the summative account of GB, Gilbert puts her “plural subject” account forward, according to which — for a group G to believe that p — “it is both necessary and sufficient for members” to be “jointly committed to believing as a body that p” (Gilbert, 2004, p. 100). Simply put, in this view, whether the members of G individually share the same belief does not matter. What matters is whether the members “have openly expressed their readiness to let the belief in question be established” as G's belief (Gilbert, 2004, p. 100, emphasis added).

Like many other non-summative accounts, Gilbert substitutes Condition 1 with another condition of sharing a conative attitude that is necessary for group decision-making and is common in type-b cases. According to the “plural subject” account, this attitude is a “joint commitment” to let p be the group's belief. What makes a joint commitment possible is a two-layer procedure: First, “common knowledge” must be there between the members to show that they are ready to let p be their group's belief. This knowledge is to be produced and distributed by their open expressions. Second, from this common knowledge, a “joint commitment” emerges that makes members “answerable to one another,” preventing them from conducting actions that violate the GB in question (Gilbert, 2004, pp. 100–101). In a nutshell, Gilbert claims that the members of G can create a GB by sharing a non-belief attitude of commitment. More precisely, for group G to believe that p, the majority of all members must be committed to p as their group's belief.

In a similar vein, Tuomela (1992, 2004) substitutes Condition 1 with sharing another conative attitude — that is, “acceptance.” Tuomela's “positional account” of GB is heavily based on his social ontology. Groups, he believes, cannot enjoy group attitudes unless their members achieve some sort of we-mode, according to which “the members are supposed to function as group members almost as if they were intentionally functioning as parts of an organism” (Tuomela, 2013, p. 34). This description covers both autonomous groups “that are internally governed” and non-autonomous groups “that are under the governance (possibly under threat of coercion) of an external authoritative power” (Tuomela, 2013, p. 56). In this way, a group can have an attitude that φ only if its members have a φ-related attitude in a we-mode, even if they accept that attitude due to external power. More precisely, Tuomela defines a “we-attitude” of the members as follows:

The person a) has [an attitude] ATT(p) and b) believes that also the others in the group have ATT(p) and also c) believes (or at least is disposed to believe) that it is mutually believed (or, in a weaker case, plainly believed) that the members have ATT(p). (Tuomela & Bonnevier-Tuomela, 2020, p. 30)

In his exclusive analysis of GB (Tuomela, 1992, 2020), Tuomela uses the same view to define GB. According to him,

[group] G believes that p in the social and normative circumstances C if and only if in C there are operative members A_a, …, A_m of G in respective positions P₁, …, P_m such that:

• (1') the agents A₁, …, A_m, when they are performing their social tasks in their positions P₁, …, P_m and due to exercising the relevant authority system of G, (intentionally) jointly accept that p, and because of this exercise of authority system, they ought to continue to accept and positionally believe it;

• (2') there is a mutual belief among the operative members A₁ ~, …, A_m to the effect that (1');

• (3') because of (1'), the (full-fledged and adequately informed) nonoperative members of G tend tacitly to accept - or at least ought to accept - p, as members of G; and

• (4') there is a mutual belief in G to the effect that (3'). (Tuomela, 1992, pp. 295–296, emphasis added)

In this way, Tuomela diverges from the summative accounts not only in terms of the attitude content sharing among the members (similar to Gilbert) but also in terms of its constituency (contrary to her). The main constituents of GB, in Tuomela's account, are the operative members, and the shared belief-maker attitude is acceptance.

The conative non-summative accounts of GB, though promising prima facie, encounter a couple of serious difficulties. First, as mentioned, it is not clear why we are permitted to claim arbitrarily that GB is an exceptional case of group attitude that does not require the group's members to share the same kind of attitude.

Second, and more importantly, replacing the condition of sharing a belief with that of sharing a commitment or acceptance may lead to replacing group beliefs with group actions. Recall when a friend of mine and I pretended that my captured black queen was her second white queen to continue playing chess, without believing that “it is white.” Conative non-summative accounts may over-generate cases of GB by taking cases of group pretense as GB. Tuomela, for example, invites us to imagine a communist party in which “no one personally believes” that “capitalistic countries will soon perish” while they collectively accept to act in their positions that they believe so (Tuomela, 1992, p. 302). Tuomela explicates that this party holds “the belief in question” but labels it as “a spurious group belief” (Tuomela, 1992, p. 302, emphasis added). Tuomela's verdict, however, faces two problems. First, the term “spurious belief” seems oxymoronic. As mentioned, to believe that p is to take p as true in a cognitive sense. In this sense, one cannot take p as true while one believes that what one takes is false. In other words, to take a spurious stand, one must pretend that p; and when one pretends that p, one does not believe that p.

Second, as we have observed, Tuomela's account in particular, and the conative non-summative accounts in general, cannot properly identify the difference between the cases where the majority of group G believes that p and where they only pretend that p. In many cases of group pretense, such as group obedience, game, deception, lies, and bullshit, the members of a collective only act as if p is the case due to collective norms, mutual expectations, external threats, power relations, or social sanctions, without holding the belief that p.⁴ To the extent that the relevant constituents accept or commit to p, the conative non-summative accounts cannot distinguish these cases from GB.

To get a better epistemological grip on the distinction between conative and cognitive meanings of “taking p as true,” let's appeal to G. E. M. Anscombe's idea of the “direction of fit.” Consider Anscombe's example of the shopping list.⁵ The shopper should buy butter, but there is no butter in the store. He sees a packaged margarine and grabs it. Now, imagine two scenarios: (1) he buys margarine because he thinks it is butter; (2) he knows that it is margarine, but since there is no butter, he takes margarine as butter in his list and crosses over “Butter.” In both cases, the shopper takes or regards margarine as butter. In both cases, there is a sort of commitment and acceptance, but in two different directions. Following John R. Searle (1983), one can make a distinction between the world-to-mind direction of acceptance in (2), and the mind-to-world one in (1). In other words, while the shopper as a believer in (1) has a backward-looking commitment to what he takes as true, in (2) he holds a forward-looking commitment to the action he intends to do. To take p as true has a world-to-mind direction or prospective attitude in the conative cases and must not be confused with when a believer takes p as true retrospectively due to fitting her mind to the world. When a rational agent believes that p, she holds both commitments, while when she accepts that p, she can commit herself to p only prospectively. Hence, the cognitive conception of taking p as true can also accurately show why belief is not completely a voluntary attitude.

However, when Gilbert speaks of commitment in her account, she construes it in a forward-looking manner. In group decision-making, you and other members commit yourself to do A. As Gilbert mentions, if someone rescinds her commitment and does or asserts bluntly against A, she deserves to be blamed (Gilbert, 2004, p. 100). However, this reduces GB to a group decision. When Bahar's group believes that p (“wearing jeans is cool”), they only hold a backward-looking commitment toward p by taking p as true. They do not commit themselves to acting as if wearing jeans is cool in the future, even though it may affect their future behaviours. Nor do they deserve rebuke or blame if they wear, say, dresses tomorrow, unless they have decided together to do so beforehand. Replacing commitment with belief at the individual level reduces the collective cognitive attitude of believing to a collective conative attitude of behaving.⁶

Now we are faced with the main question: provided that our criticisms of Gilbert's and Tuomela's accounts are correct, how can we respond to their objections to Condition 1 and associate it with Condition 2? To answer this question, we need to proceed to the last condition, which is surprisingly absent in the literature on GB.

2.3. Group-Based Considerations

Condition 1 establishes a correlation between every group G's belief that p and the individual beliefs among a sufficient majority of G's members (that is, M's members). However, what is the relationship between the content of G's belief and what M's members hold? Apparently, this crucial question has not been explicitly discussed in the literature on GB.

The contents of beliefs, in some cases, are not solely what is expressed. More precisely, sometimes when I say, “I believe that p,” the precise content of my belief may not be merely p. Rather, it can be “p, given the context.”⁷ In such cases, what is expressed implies considerations and contexts. For example, I may believe that “buying a new CR-V is the best option” if I intend to buy a car, but since it is not affordable for me, I buy an old, used Civic. Have I acted against my belief? The answer is negative only if I hold two different beliefs, simultaneously: one is “buying a new CR-V is the best option if it is affordable for me” and another is “buying this old, used Civic is the best option, given my current circumstances.”

The constituents of GBs contribute to forming and/or holding GBs by incorporating group-based considerations into the content of their individual beliefs. When group G believes that p, the members of M can state that we believe that p. However, what they hold at the individual level is “p, given some considerations.” GB has specific contextual considerations on its own that I call “group-based considerations.” Therefore, group G believes that p only if M's members believe that p_G where p_G is “p, given G-based considerations.” In Conditions 1 and 2, I mentioned that M must share a belief that takes p as true. p_G is, in fact, that belief.

What group-based considerations are and who forms them in each situation is contingent on the metaphysical structures of and psychological relations within each group. But, in general, by “group-based considerations,” I mean what social psychologists refer to as intergroup “cognitive processes” — i.e., a wide range of evidential and non-evidential factors, such as group identity, group rules and etiquette, group epistemic privilege, group discussions and information flow, internal hierarchy and power relations, and mutual expectations, while they affect not only the individuals’ behaviour but also their cognition and perception (Haslam et al., 1996; Ickes & Gonzalez, 1996; Levine & Hogg, 2010).⁸

Simultaneously, I use the term “group-based considerations” rather than “group considerations” to highlight that the factors being considered by M's members are derived from or pertain to the group in question, whether or not they collectively take them into account. Moreover, M's members may or may not contribute to making their group's considerations, but as constituents of G's belief, they must be conceived by some G-based considerations that p is the case. In this sense, GBs are partly involuntary; and like individual beliefs, group believers are more belief-holders than belief-formers. In other words, these are G-based considerations that make M's members believe that p_G, not the converse.⁹

Put simply, being members of G must provide good reasons for M to take p as true, even though G-based considerations are neither comprehensively nor exclusively available to the members of M or even G. Thus, the last condition of any proper account of GB, proposed in this article, must be as below:

Condition 3: For group G to believe that p, M must believe that p is true given G-based considerations. In other words, M's members must believe that p_G, where p_G is “p, given G-based considerations.”

It is worth noting that Condition 3 is not sufficient in and of itself. Without Condition 1, a group G2 that holds “p, given G1-based considerations” does not contribute to G1's belief that p.¹⁰ Without Condition 2, a Gilbertian counter-example is conceivable where all the members of G believe that p_G while we cannot ascribe believing that p to G. In case (b), for example, even if all of the members believe that “Banff is the best place to camp, given my group-based considerations,” it seems odd to call it a GB when no one is aware of the others’ beliefs.

Regarding the second point, I should consider all three conditions together to provide my account for GB. Next, I will take this step.

3. The Cognitive Non-Summative Account of GB

There might be other necessary conditions for GBs, overlooked in this article. However, for now, I can draw on the three foregoing conditions, each of which is necessary for GB, and I consider them collectively sufficient to propose the following account of GB, named the “Cognitive Non-Summative Account” of GB:

Group G believes that p if and only if,

• (1) a sufficient majority of its members (M) believe that p_G, where p_G is “Given G-based considerations, p,” and

• (2) M's members are mutually aware of (1).¹¹

It is worth recalling that G-based considerations do not only make M's members accept that p; rather, they make them believe that p, given some new, socially situated considerations into account. This may immediately raise concerns that must be addressed. Below, we will assess a couple of potential concerns about the Cognitive Non-Summative Account.

3.1. Group Compromise

Recall when we accepted Gilbert's criticisms of the summative condition of GB (as it is neither necessary nor sufficient) while maintaining that it does not undermine the necessity of Condition 1 in our non-summative account. One might wonder how Gilbert's compromise case is not a counter-example of the necessity of Condition 1 along with the summative condition. To illustrate, consider our case (b), for example, when no one individually believes that “Banff is the best place to camp,” while after a long discussion, they all come to the consensus that “Banff is the best place to camp.” One may state that the fact that they all hold ~p at the individual level while their group's belief is p demonstrates that Condition 1 is not necessary either.

To address this objection, one must consider the distinction between p and p_G. Condition 1, in our non-summative account, stipulates that M shares a belief that takes p as true, rather than p itself. In this framework, the individual-level counterpart of p is p_G, expressed as “Given my group-based considerations, p.” Just like the example of buying a car (Section 2.3), in the modified version of case (b), where everyone compromises on Banff, each individual, in fact, holds two associable beliefs, simultaneously: one is ~p that is held based on personal considerations: “Banff is not the best place”; and the other is p_G that “Given my group-based considerations, Banff is the best place to camp.” With that in mind, even the cases of group compromise make no problem to Condition 1 that we cannot have a group belief without having some belief in common among the members of the group. However, the precise content of their individual belief is p_G, or “Given my group-based consideration, p.” Consequently, similar to the summative account, the Cognitive Non-Summative Account of GB requires sharing some belief among the members, whereas, contrary to the former, the latter does not mandate to share exactly the same belief, nor does it understand the direction of considerations from members to groups.

3.2. Consideration-Free GBs?

Laying emphasis on the “group-based considerations” puts another concern forward, whether these considerations are really at work in any case of GB. Consider case (b) again, for illustration. Suppose Bahar and each of her nine friends individually believe that “Banff is the best place to camp.” Then they gather, find out that they all believe this, and so decide to go to Banff together. When asked why, they say, “Because we believe that ‘Banff is the best place to camp’.” Given that they already believed that p, how can we ascribe a group-to-members direction of considerations to their GB that p?

Once again, distinguishing between p and p_G is the key answer to the preceding question. Bahar and her friends will not have a GB until they all believe that “Given our group-based considerations, Banff is the best place to camp,” even if they all individually believe that “Banff is the best place to camp.” To believe that p_G, they need to be a member of G and mutually endorse p, given their group's considerations, whether or not they individually believe that p.

The same point also applies when we deal with a-type cases and other examples of GB. Consider the following scenario:

1. (c) A climbing group C is ascending a mountain when one of them shouts: “Look! A falcon!” They all turn and look at the bird flying far away. Should you explore their belief, they will answer: “We saw a falcon.”

One could argue that Condition 3 is not fulfilled in this scenario, because this is merely witnessing the bird that makes them take “we saw a falcon” as true, rather than being a member of C or regarding C-based considerations. To my mind, agreeing with this verdict does not go too far. The pronoun “we” in “We saw a falcon” could either refer to any aggregation of people who saw the bird, including some members of C, or to the group as a whole. In the first case, we are not faced with a GB, confirming the necessity of Conditions 2 and 3. In the second case, C-based considerations must still play their role, though the less restrictive, the further they are from C's goal and expertise. To illustrate, consider this modified example:

(d) A climbing group D is ascending a mountain when one of them shouts: “Look! A falcon!” They all turn and look at the bird flying far away. However, this time D is a special climbing group — they are all ornithologists. A few members cast doubt on the first assertion by pointing out that, say, “Its cry sounds harsher than a falcon's” or “It seems bigger than a falcon,” and it is enough to make them suspend the belief that “we are seeing a falcon” or even makes them take it as a hawk.

Provided that both cases (c) and (d) involve GBs, we see how different group-based considerations lead them to take p differently, despite witnessing the same bird. It is worth noting that group-based considerations in (d) extend beyond the last two assertions, encompassing various factors, from counter-evidence that those assertions provide for the other members to the power relations among the members. Regardless, group-based considerations consistently influence the background of a GB that p and make M believe that p_G. In some situations, such as group compromises, it creates a parallel belief for each member of M; in other situations, such as case (d), it can radically change the members’ beliefs.

3.3. Modified Summativism

As the last concern, one might wonder if there is a more straightforward account of GB sharing all of the strengths of the Cognitive Non-Summative Account but with fewer conditions or simpler articulation. Recall that the classic summative accounts, according to the literature, consider the condition of the “summation of individual beliefs” as both necessary and sufficient for GB. However, one might call them “simple” summative accounts, juxtaposing them with modified summative accounts according to which the summation condition is necessary, but not sufficient. According to this definition, I refer to Jennifer Lackey's (2021) approach as one of the most important instances of modified summative accounts, even though she introduces her view as “neither strictly summative nor non-summative” (Lackey, 2021, p. 20).¹²

In her recent work on GB, Lackey confirms a modified summative condition that we cannot attribute belief p to group G unless a sufficient majority of G's members believe that p. The subtitle of her chapter, “Lessons From Lies and Bullshit,” aims at showing the fact that the conative non-summative accounts of GB (including Gilbert's and Tuomela's accounts) cannot capture the cases in which a group decides to lie or bullshit, while we know “both a lie and bullshit undeniably involve the absence of belief” (Lackey, 2021, p. 34). To avoid this weakness, Lackey retreats to a modified version of summativism, according to which “group, G, believes that p if and only if: (1) there is a significant percentage of G's operative members who believe that p, and (2) are such that adding together the bases of their beliefs that p yields a belief set that is not substantively incoherent” (Lackey, 2021, pp. 48–49).

While Lackey's account effectively addresses cases of group lies, or group pretense as we discussed in a broader sense, it encounters a pitfall warned by Gilbert, like other summative accounts. Lackey's approach to coping with the cases of group compromise involves refusing to classify them as genuine cases of GB. Instead, she offers to call such cases as the group's “official position,” “verdict,” or decision (Lackey, 2021, pp. 22, 51–52).

I do not dispute that such labels may sometimes clarify the motivations behind groups’ assertions or actions. However, replacing all cases thought to represent GBs with these terms presents two significant challenges. First, it creates semantic difficulties by contradicting our common intuition that perceives such cases as GBs. Second, it makes GB a phenomenon that is too demanding and too rare. As mentioned in Section 3.2, it seems quite plausible that the group members in the cases of compromise ascribe p as their GB.

The Cognitive Non-Summative Account not only proposes a solution to the group compromise predicament but also satisfies Lackey's concern without succumbing to her second condition, which requires coherent bases for shared beliefs at the individual level.¹³ According to the Cognitive Non-Summative Account of GB, each member of M believes that p_G whose basis, at least partially, is “G-based considerations,” even though with different interpretations of “G-based considerations.” This allows each member of M to maintain different beliefs at the individual level, while simultaneously sharing a collective belief that p_G based on “G-based considerations.”

Condition 3 also enables the Cognitive Non-Summative Account to capture the cases of GB that Lackey rejects under the title of “judgment fragility.” According to Lackey, if members of a group who believe that ~p “deliberate about the same body of evidence at T1 and T2 with no relevant difference in the information that emerges via the deliberation” and change their mind to p only to avoid having to devote a great deal of time to such matters, while not a single member of the group actually believes that p, we cannot call this a case of GB (Lackey, 2021). Lackey's plausible concern here is to grasp stability as a basic feature of any belief.¹⁴ However, stability is compatible with the contextuality of beliefs. Beliefs are stable in a way that a rational believer has a retrospective commitment to it and does not change it without any reason. However, practical reasons are also reasons that stabilize a belief, and time restriction may serve as a practical reason.

Finally, unlike non-summative accounts, particularly those proposed by Gilbert and Tuomela, Lackey's account of GB does not emphasize mutuality as a crucial criterion. Recall Gilbert's main criticism of the sufficiency of the summative condition that shows all the members of group G may believe that p, yet p is not a GB. Lackey, inaccurately, reduces this objection to an “irrelevance argument” and undermines its significance. To illustrate Gilbert's point, Lackey takes the example of an academic group whose members believe that “the best red pepper hummus in Chicago can be found at Whole Foods.” Lackey believes that Gilbert rejects this as a case of GB “because assessment of red pepper hummus is entirely irrelevant to the goals and purposes of the group” (Lackey, 2021, p. 23). However, as discussed in Section 3, Gilbert's argument is deeper. She asserts that while all members may individually hold the same belief, the absence of “common knowledge” regarding others’ beliefs results in a lack of “joint commitment,” which is, in turn, a necessary condition of GB. Put simply, Gilbert's critique against the sufficiency of the summative condition is based on the mutuality or we-mode of attitudes without which group agency is not available. While Lackey fails to explicate this condition in her account, the Cognitive Non-Summative Account incorporates it as a necessary condition.

In brief, any attempt to undermine Condition 3, whether by resorting to conative non-summative accounts or adopting modified summative accounts of GB, could result in either an excessive proliferation or an insufficient generation of GBs.

4. Conclusion

Current accounts of GB are usually caught in a dilemma between summative accounts and conative non-summative accounts. Summativism, even though plausible for social or epidemic beliefs, over-generates cases of GB. However, the non-summative alternatives typically fall short of properly attributing a belief to a group. This is because they substitute beliefs with conative attitudes without sufficient justification. In this article, I have argued that the only way out of this predicament, while adopting the insights of both camps, is the Cognitive Non-Summative Account of GB. By taking beliefs’ contextual considerations into account, the Cognitive Non-Summative Account maintains that for group G to believe that p, a sufficient majority of G's members (M) must mutually believe “p, given G-based considerations.” It is not the summation of individual beliefs that makes M's members believe that p; rather, being G's members contributes to making them take p as true. This view aligns with group realism and taking the group epistemic agency seriously.