1. Introduction
The debate on group beliefs plays a prominent role in social epistemology. It also has an important role because it concerns a foundational topic in the field. This debate concerns the conditions groups should satisfy to be attributed beliefs. If we can make sense of doxastic or epistemic group attributions, we can make social epistemology something more than a debate on socially gathered evidence or the influence of social institutions on people’s beliefs. To use Goldman’s (Reference Goldman, Goldman and Whitcomb2011: 16) classical terminology, an affirmative answer makes a collective doxastic agent variety of social epistemology possible.Footnote 1 Thus, this debate plays a crucial role in shaping social epistemology as a research program.
The debate on group beliefs can be divided along different dimensions. One such dimension divides the debate into two camps: non-agential and agential accounts (Miller, Reference Miller2015: 420; Miller, Reference Miller2024: 13–14). Roughly, non-agential accounts are concerned with groups that do not behave as an individual agent, whereas agential accounts are concerned with groups that mimic the behavior of an individual agent (Pettit, Reference Pettit and Lackey2014b; Lackey, Reference Lackey2021; Bird, Reference Bird2024). Another dimension divides groups into deliberative and non-deliberative groups. These dimensions can intersect. One such intersection is agential deliberative groups. In this paper, I focus on cases of agential deliberative groups.
Jennifer Lackey (Reference Lackey2021) has introduced the Group Agent Account. This account aims to provide an account of group beliefs for agential groups, including also agential deliberative groups. Although the Group Agent Account is a recent contribution to the literature on group belief, it has attracted substantial interest and is already one of the most influential, especially for its role in Lackey’s (Reference Lackey2021) account of justified group beliefs (Ball Reference Ball2021; Brown Reference Brown2022; Biebel Reference Biebel2023).
In this article, I do two things. First, I present a problem for the Group Agent Account that is based on cases of pluralistic ignorance. I argue that the Group Agent Account does not attribute the correct beliefs in cases of pluralistic ignorance. Moreover, I argue that pluralistic ignorance presents a systematic problem for the Group Agent Account. This problem shows that, in these cases, the Group Agent Account ascribes states that do not play the role that Lackey takes beliefs to have in agential groups. Thus, in a kind of group that is highly relevant to Lackey’s project (i.e., agential deliberative groups), pluralistic ignorance presents a problem for the Group Agent Account. Second, I discuss two seemingly natural ways to fix the Group Agent Account to make it immune to the pluralistic ignorance problem. However, I argue that these solutions are unsatisfactory. Along the way, we see how addressing cases of pluralistic ignorance raises important issues concerning a group’s self-knowledge, its failures, critical reasoning, and group evidence responsiveness.
Lackey’s Group Agent Account is currently a very influential account of group beliefs for agential groups. It is also perhaps the most influential summative account of group belief currently in the literature. Since cases of pluralistic ignorance show that the Group Agent Account yields wrong belief ascriptions and it cannot be easily fixed to address the pluralistic ignorance problem, this article gives evidence in support of the claim that the Group Agent Account cannot be correct. Moreover, since problems come from an assumption generally held by summative accounts, pluralistic ignorance is a problem for summative accounts. However, pluralistic ignorance is not a problem for nonsummative accounts. Therefore, this paper provides evidence in support of a nonsummative approach to accounts of group beliefs.
The remainder of this article is structured as follows. In Section 2, I introduce the Group Agent Account and explain its main features. In Section 3, I introduce pluralistic ignorance, present an example of pluralistic ignorance that is problematic for the Group Agent Account, and argue that pluralistic ignorance poses a systematic problem for the Group Agent Account. In Section 4, I discuss two ways to improve the Group Agent Account, which I call the Group Agent Account with Open Expression and the Group Agent Account with Sincere Open Expression. I argue that both face problems of their own. Therefore, neither is a satisfactory solution. Section 5 provides the conclusions.
2. The group agent account
In this section, I introduce the target of my paper: the Group Agent Account. Before I do so, however, I place the account within a broader perspective and set out some important concepts and distinctions. First, I distinguish different types of groups. The cases I consider concern agential deliberative groups. Second, I distinguish summative from nonsummative accounts of group beliefs. The Group Agent Account falls within the summative camp.
Let me give you some intuitive examples.Footnote 2 Left-handed people are a mere collection.Footnote 3 The whole physics and mathematics community is a non-agential group. The Board of Directors of a big company is an agential group. In this paper, I am concerned with agential groups. Very coarsely, the differences are as follows. Mere collections share a common feature but are not organized in any way. Non-agential groups are loosely organized. Agential groups are groups whose members aim to act as a single agent. According to Lackey (Reference Lackey2021: 12), agential groups can be evaluated normatively along epistemic and moral dimensions.
Among groups that can be the subject of normative evaluation, there are agential deliberative groups. These are groups that can engage in collective deliberations that motivate their collective actions and collective reasoning. The relationship between these groups’ doxastic or epistemic states and their actions is such that their states can motivate or rationalize actions.Footnote 4 I restrict the focus of this article to agential deliberative groups that find the reasons for their collective actions in the outputs of their collective reasoning and deliberations. I do not claim that these are the only groups that are relevant in a social epistemology project.Footnote 5 However, these are agential groups of the kind to which Lackey ascribes beliefs.
The claim that the group is the primary bearer of responsibility may seem puzzling because, within groups, it is individual members who carry out actions. However, the idea that groups can be the primary bearer of responsibilities has a long history and imbues many well-established practices.Footnote 6 For instance, it imbues legal practices. A case Lackey uses to motivate her approach concerns National Semiconductor’s scandal (Velasquez, Reference Velasquez2003). National Semiconductor was charged by the United States Department of Defense on March 5, 1984. The charge was that they sold badly tested computer chips to the Department of Defense and falsified their record to conceal the fraud. National Semiconductor’s strategy was to accept responsibility but only as a company. That is, no individual employee was individually responsible for the fraud. National Semiconductor’s defense prevailed, and only the company was held liable for the crime.
Agential groups occupy a prominent position in the debate on group beliefs (Lackey, Reference Lackey2021; Brown, Reference Brown2024). Groups of this kind are organized to mimic the behavior of an individual (Pettit, Reference Pettit2014a). We can understand and predict the behavior of agential groups, and we commonly ascribe beliefs to them (Brown, Reference Brown2024). Moreover, ascriptions of beliefs to agential groups are important for assessing them epistemically and morally (Lackey, Reference Lackey2021; Brown, Reference Brown2024). Examples of agential groups include government and companies. For instance, the Board of Directors of a company.
Very generally, accounts of group beliefs for agential groups can be categorized into summative and nonsummative accounts. According to summative accounts, for a group G to believe that p, it is necessary that some (most or all) of its members believe that p.Footnote 7 A currently influential summative account is Lackey’s Group Agent Account. By contrast, according to nonsummative accounts, for a group G to believe that p, it is neither necessary nor sufficient that there is a member who believes that p. Gilbert’s (Reference Gilbert1987) joint-acceptance account, Tuomel’s (Reference Tuomela1992) positional account, and Brown (Reference Brown2024) are all prominent examples of nonsummative accounts.
Lackey (Reference Lackey2021: 30–47) motivates the introduction of the Group Agent Account with counterexamples to standard nonsummative accounts of group beliefs. These counterexamples are based on cases of group lie and group bullshit.Footnote 8 These standard accounts include Gilbert’s joint-acceptance account, Tuomela’s positional account, and List and Pettit’s (Reference List and Pettit2011) judgment-aggregation account. Lackey’s argument is that, since all these accounts mistakenly ascribe group beliefs in case of lies (and bullshit), they are not correct. This conclusion provides motivations to introduce a new account of group belief.Footnote 9
Def. 1 Group Agent Account: “A group, G, believes that p if and only if (1) there is a significant percentage of G’s operative membersFootnote 10 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 Reference Lackey2021: 48–49)
The Group Agent Account defines group beliefs in terms of the individual beliefs of at least some of the members of the group, Condition (1), and in terms of a collective condition, Condition (2), that applies to the bases (i.e., the reasons, evidence, and motives) that the individual members of the group have for believing that p. Condition (1) is meant to ensure that group beliefs aim at truth (Lackey Reference Lackey2021:49–50). Condition (2) is meant to provide the collective dimension of group beliefs (Lackey Reference Lackey2021:12–13).
The feature of the Group Agent Account that we should keep in mind in Section 3 is that this account does not require that the beliefs of the operative members (Condition 1) be openly available and shared within the group. It is consistent with the Group Agent Account that beliefs and bases are never openly discussed. This does not mean that group agents do not engage in discussions and deliberations. However, the Group Agent Account ascribes beliefs even though these beliefs are not discussed.
In the next section, I argue that this feature leads to the pluralistic ignorance problem. The pluralistic ignorance problem shows that the Group Agent Account ascribes beliefs that do not fulfill some relevant functional roles that Lackey assigns to group beliefs.
3. The pluralistic ignorance problem
Recently, de Ridden (Reference de Ridder2023: 863–864) and Bird (Reference Bird2024: 16) mentioned, without developing an extended argument, that Group Agent Account does not seem to take into account cases in which group members do not reveal their true beliefs. Neither de Ridden nor Bird addressed how problematic these cases might really be for Lackey’s group epistemology. In this section, I argue that cases of pluralistic ignorance provide counterexamples that are relevant to the kind of epistemological project that Lackey carries out. Along the way, we will also see that these cases raise important epistemological issues about group critical reasoning and evidence-responsiveness.
Pluralistic ignorance refers to a social setting in which there is a systematic mismatch between the private beliefs of the members of the group and their public behavior and expressions of beliefs (Bicchieri and Fukui Reference Bicchieri, Fukui, Galavotti and Pagnani1999: 91–95). Pluralistic ignorance is a well-known phenomenon in the social sciences (Miller and McFarland Reference Miller and McFarland1987, Reference Miller, McFarland, Suls and Wills1991). When a group is in a state of pluralistic ignorance, (i) each member of the group privately believes that p while (ii) believing that all the other members believe that not-p. In these cases, because of this misrepresentation of beliefs and the interplay of this misrepresentation with peer pressure and social norms, (iii) each member behaves and speaks as if believing that not-p.Footnote 11
Pluralistic ignorance has been studied by social scientists in a variety of situations. For instance, it has been studied in field experiments to explain attitudes towards alcohol consumption and gender stereotypes on campuses or schools (Prentice & Miller, Reference Prentice and Miller1996), but it has also been studied in experimental settings (Miller & McFarland, Reference Miller and McFarland1987). Pluralistic ignorance is important in the field of managerial studies because it can impact the quality of boards’ decisions (Halbesleben et al., Reference Halbesleben, Wheeler and Buckley2007). This last example is particularly pressing for the Group Agent Account, because it shows that pluralistic ignorance can occur in agential groups such as a company’s board. That is, the kind of group that is the Group Agent Account’s main concern.
Groups in which there is a systematic mismatch between privately held and openly expressed beliefs are not new in the literature on group beliefs. Gilbert’s (Reference Gilbert1987: 187) tribe case is an early example. In this case, Gilbert describes an anthropologist who attributes to the Zuni tribe the belief that the north is the region of force and destruction. However, the anthropologist’s grounds for this attribution are that each member of the tribe believes this, even though each is afraid to tell everyone else what she believes. Although she does not provide a full argument, Gilbert goes on to claim that these grounds cast doubt on the original attribution. Furthermore, Bird (2010: 29–30; Reference Bird, Fricker, Graham, Henderson and Pedersen2019: 276) lists as a desideratum for an account of group belief that it respects cases in which agents misrepresent their beliefs. In such cases, according to Bird (Reference Bird, Fricker, Graham, Henderson and Pedersen2019: 276), a good account of group beliefs should attribute the openly expressed belief rather than the privately held one. Bird’s (2010: 29–30) short argument can be reconstructed as follows. He considers Hans Christian Andersen’s The Emperor’s New Clothes.Footnote 12 Bird argues that in this story, everyone sees the birthmark on the Emperor but behaves as if believing that she does not know that the Emperor has a birthmark on his shoulder; thus, it would be wrong to attribute to the group the belief that the Emperor has a birthmark. The reason that makes such attribution wrong is that, according to Bird (Reference Bird2010: 30), it has no social effects. It is, as it were, an idle wheel because everyone acts as if believing that she does not know that the Emperor has a birthmark on his shoulder. Although Bird does not establish the connection between the cases he discusses and pluralistic ignorance, there are strong similarities that support the claim that this desideratum bears on cases of pluralistic ignorance. However, neither Gilbert nor Bird gives a full argument in support of the claim that we should attribute the openly expressed belief. Moreover, Lackey does not address these cases or provide reasons to attribute the privately held belief. Thus, there is an argumentative gap to be filled. Let me argue that the Group Agent Account’s ascription is incorrect.
The following example aims to show that a group agent’s belief cannot be determined by individual beliefs because it is possible that members misrepresent their beliefs to others. I believe that Lackey is correct when she writes that group beliefs are the sort of things that groups can use in their deliberations as premises of their arguments or justifications for their actions. These are important functions that group beliefs play and that can help us determine whether an ascription of belief to a group is correct.Footnote 13 My example shows that there are relevant cases in which individuals’ beliefs cannot always fulfill this role. Thus, the Group Agent Account faces a problematic case.
To show that cases of pluralistic ignorance are a source of problems for the Group Agent Account, let me elaborate on Lackey’s Tobacco Company case (Lackey Reference Lackey2021: 30) in a way that introduces pluralistic ignorance.Footnote 14
Tobacco Company: The members of the Board of Directors of Philip Morris assess the scientific evidence on smoking-related diseases and tell each other that this body of evidence shows smoke’s dangers for health. Admitting that to the public would bring serious economic consequences. They then decide to issue statements denying that smoking cigarettes is dangerous for one’s health. However, unbeknownst to each other, each member of the board is secretly a science denialist who hides his/her true beliefs for fear of being sacked. When acting as a member of the board, they assess evidence in light of the best epistemic practices. Each of them, unbeknown to each other, disregards scientific findings as untrustworthy and believes that smoking is safe because it has been widely practiced for years.Footnote 15
In the Tobacco Company, each member of the Board of Directors privately believes that smoking cigarettes is safe.Footnote 16 Thus, Condition (1) of the Group Agent Account is satisfied. Moreover, the bases of their beliefs are not substantially incoherent because they all believe that smoking cigarettes is safe because people have been smoking for years. Thus, Condition (2) of the Group Agent Account is also satisfied. Hence, the proposition “smoking cigarettes is not dangerous for one’s health” (call it p) satisfies the two conditions of the Group Agent Account. Therefore, the Group Agent Account ascribes to the Board of Directors the belief that p.
However, the members of the Board of Directors are in a state of pluralistic ignorance. The members of the board all privately believe that what they did was wrong. Thus, the point (i) of pluralistic ignorance is satisfied. Each of them believes that the other directors believe that what they did was right. Thus, point (ii) of pluralistic ignorance is satisfied. Moreover, they all act and speak as if believing that what they did was right. Thus, point (iii) of pluralistic ignorance is satisfied.
Since, according to the Group Agent Account, the board believes that p, the board does not lie to the public when it issues public statements denying smoking-related health risks. Thus, we cannot negatively evaluate the group’s actions on the ground that they are lying to the public. We might be tempted to go with Bright (Reference Bright2020: 212–218) and argue that pluralistic ignorance shows that evaluating groups for their lies is not important. Instead, social epistemology should pursue a different project. I believe Bright’s point is an important one. However, his argument does not succeed in supporting the claim that power structures are more important than holding groups responsible for their lies. It is perfectly reasonable to think that both projects are equally worth taking.
To place my example in a broader context within the literature, pluralistic ignorance is not a concern for nonsummativists. For a nonsummativist, cases of pluralistic ignorance are just another case in which the beliefs of a group do not depend on its members’ beliefs. However, a nonsummativist intuitive analysis of the Tobacco Company is not a sufficient reason for dismissing Lackey’s account. To prove her wrong, we need to rely on claims she would seemingly endorse and show that they support the claim that the Group Agent Account’s attribution is not correct.
Indeed, Lackey (Reference Lackey2021: 46) seems to provide a reason for regarding the Group Agent Account’s ascription as problematic: “Group beliefs have to be the sorts of things that can coherently figure into collective deliberation about the future actions of the group.” The belief that the Group Agent Account attributes to the Board of Directors (i.e., that smoking cigarettes is not dangerous) does not and cannot coherently figure into the collective deliberations about the future actions of the Board of Directors.Footnote 17 It cannot coherently figure into the group’s deliberations, at least in relevantly similar scenarios, because there is a strong social pressure against science denialist claims. Members of the board are not likely to reveal their true beliefs. Instead, the belief that smoking cigarettes is dangerous coherently figures into the collective deliberations of the Board of Directors and motivates the decision to issue a misleading statement to the public.
Lackey’s remarks about the place group beliefs should have in collective deliberations seem grounded in a specific understanding of the relationship between belief and action. Indeed, Lackey (Reference Lackey2014: 290) argued that a classic role of beliefs is, together with desire, to rationalize action. Imagine, for instance, that the members of the Board in the Tobacco Company say, “We need to lie to the public and deny that we believe that smoking is dangerous,” after which the board approves issuing public statements saying that they do not believe that scientific research has proved that smoking is dangerous. They then summon senior managers to brief them about what they are about to do. When senior managers ask the members of the board to explain their actions, these members say that, since they believe scientific research to be correct, but their desire is to preserve the economic stability of their company, they decided to lie to the public. The members of the board then take extra measures to guarantee that these conversations are not leaked to the public. The belief that smoking is dangerous rationalizes the board actions. For instance, it rationalizes what they say to their senior managers. It also rationalizes why to take extra measures to keep conversations behind closed doors. In contrast, the belief that smoking is not dangerous does not rationalize what the board says to their managers. If the board believed that smoking was not dangerous, they would not need to tell the managers that they are about to lie to the public. Moreover, they would not need to take extra measures to prevent leaks.
We are now in a position to better assess the Group Agent Account’s pluralistic ignorance attributions in light of claims Lackey commits to. First, Lackey seemingly endorses the claim that beliefs are characterized by their role in rationalizing actions. Second, she (2014: 290) appealed to this claim to argue that Bird’s (2010) account of social knowing cannot be correct because it makes the relations between beliefs and actions wrong. Thus, Lackey seems committed to regard the Group Agent Account’s attribution in the Tobacco Company case as problematic.
I said that the board members tell senior managers that they believe that the scientific evidence is correct. They believe that smoking cigarettes is dangerous. This discussion presupposes that the group self-attributes the belief that smoking is dangerous. These sorts of self-attributions aimed at rationalizing an agential group’s behavior to others are the sort of things that are carried out in complex groups. They are used to justify or rationalize the group’s standing on a matter. If the members of the board took themselves to believe that scientific evidence is not strong, its members would not rationalize their decision to issue a statement denying scientific evidence as an instance of lying to senior managers. They could simply say that the board did not take scientific evidence to be sound.
Appealing to self-attribution of belief to argue that the correct attribution is the one the board attributes to itself leads to a natural objection. That is, cases of pluralistic ignorance are an instance of failure of self-knowledge. In such a case, the group believes that smoking is safe but mistakenly attributes to itself the belief that smoking is not safe.
In the next subsection, I argue that considerations on critical reasoning and self-knowledge suggest that the Group Agent Account faces a dilemma: either agential groups cannot engage in critical reasoning or the Group Agent Account is false. In subsection 3.2, I further argue that cases of pluralistic ignorance show that the Group Agent Account has difficulties in accounting for direction of fit and evidence responsiveness of at least some relevant cases of group beliefs.
3.1 Pluralistic ignorance, failure of self-knowledge, and critical reasoning
Many in the debate on group beliefs maintain that we need to develop accounts of group agents’ doxastic states by presenting instances of everyday belief ascriptions.Footnote 18 Lackey (Reference Lackey2021: 1–5), along with Gilbert (Reference Gilbert1987: 186), Tuomela (Reference Tuomela1992: 285–286), and Brown (Reference Brown2024: 1–2), draw on this pre-theoretical understanding of how doxastic states are attributed to groups to motivate their own accounts. For example, it can be natural to say that a committee believes that A is the best candidate. On the face of it, the kind of reasons that motivate group belief ascriptions motivate the claim that agential groups can engage in critical reasoning. According to Burge (Reference Burge1996), subjects engage in critical reasoning when they take a standpoint in which they assess their own attitudes.Footnote 19 It is natural to say that groups engage in critical reasoning. Groups can assess whether their reasons justify their beliefs. They can assess whether their beliefs satisfy their epistemic standards. They can assess whether their beliefs or actions are morally right. These are all cases that seem naturally executed by agential groups. For example, a Board of Directors may assess whether their beliefs are in line with their company values. If their beliefs are not in line with the company’s values, they may change their group beliefs. They can also assess whether their decision to lie to the public was motivated by their belief that smoking cigarettes is dangerous. Thus, the Board of Directors can critically assess their own beliefs. For instance, the members of the group might say, “Since we believe in scientific evidence but admitting that to the public would have been economically catastrophic, we lied to the public.” Thus, it seems that agential groups can be critical reasoners, and Lackey’s already mentioned remark that group beliefs are the kind of things that can be used in deliberations about the group’s actions suggests that she is committed to group critical reasoning.
For the sake of my argument, I restrict this discussion to belief-attitude. Specifically, critical reasoners should be able to assess their beliefs and how they are connected to the reasons we have. If there are mismatches between reasons and beliefs, critical reasoners should be able to revise their states in the appropriate way. For example, if I assess my belief that the cat is on the table but realize that my reason for this belief is that I dreamt of it a few minutes ago, then I should change my attitude towards this belief. That is, I should withdraw my belief.
A point that has been argued for in the literature on critical reasoning and content externalism is that, even if the content of our belief is externally determined, critical reasoners cannot easily mistake the content of their belief. Thus, even though we can surely fail to know some of our attitudes, we are nonetheless knowledgeable with respect to occurrent and salient states that are critically assessed. When we are in the process of assessing whether I believe that p, it seems constitutive of our conception of critical reasoning that we are in a position to know our state (Burge, Reference Burge1996, Reference Burge, Wright, Smith and Macdonald1998; Brown, Reference Brown2000, Reference Brown, Beckermann, McLaughlin and Walter2009).
The debate on critical reasoning and self-knowledge concerns individual subjects.Footnote 20 For this reason, some specifications are in order. When we assess agential groups, we regard them as mimicking sufficiently well a single individual. In our assessment of their states, the subject is the group. Moreover, when group members engage in the activities of their group, they assess their own group as members of the group. When they evaluate their own group, they critically reason on their bases of actions and on the kind of reasons given to justify a statement. Thus, if they are able to do so as a single collective agent, on the face of it, they must be able to correctly identify their salient and current group beliefs. When the board are pressed by senior managers to evaluate their actions and deliberations, they engage in critical reasoning. For instance, they assess whether their beliefs motivate their actions. If we deny that groups are able to identify their own salient and occurrent group belief-states, then advocates of the epistemology of groups should present a way to make sense of group critical reasoning when groups are so easily at risk of failure of self-knowledge.
We may attempt to bite the bullet and rescue the Group Agent Account by arguing that this is in fact a case of failure of self-knowledge and the board cannot engage in critical reasoning. However, this strategy has some important consequences. To claim that the group lacks self-knowledge when in a state of pluralistic ignorance is to maintain that, even when the board appears (both to us and to its own members) to be engaged in critical reasoning, it is in fact not doing so. Moreover, since neither outside observers nor the group’s own members can assume that members’ beliefs are transparent, denying the group’s self-knowledge amounts to saying that we have serious difficulties distinguishing genuine instances of critical reasoning from those that are merely apparent. Thus, according to this strategy, we naturally attribute beliefs to groups, and we are right in doing so. However, even though we naturally attribute critical reasoning to groups, we are wrong to think that, when in a state of pluralistic ignorance, the group can engage in critical reasoning. At this point, the Group Agent Account’s advocates must explain why it seems so natural and correct to ascribe beliefs to groups while attributing critical reasoning to a group, which likewise feels natural, is in fact mistaken at least in some cases.
Cases of pluralistic ignorance present a problem to the Group Agent Account. If we want to maintain groups’ ability for critical reasoning, then the Board of Directors cannot be mistaken when it attributes to itself the belief that smoking is dangerous. Thus, the Group Agent Account gives the wrong belief attribution, and the Tobacco Company case presents a counterexample to it. Moreover, we are not inclined to claim that the Board of Directors believes that smoking is dangerous merely by reflecting on the group’s capacity for critical reasoning. Since we are concerned with group agents rather than mere group subjects, we want our accounts of group beliefs to identify the correct beliefs that figure in the processes that lead to action and motivate it. As noted in Section 3, the belief attributed to the Group Agent Account does not play the appropriate role, as it cannot coherently figure in the group’s deliberations.
3.2 Pluralistic ignorance, direction of fit, and evidence responsiveness
Lackey (Reference Lackey2021: 44) argued that prominent nonsummative accounts do not take the direction of fit between (alleged) group beliefs and the world right. The direction of fit that characterizes beliefs goes from them to the world. It is a belief that tends to adapt to the world. By contrast, the direction of fit that characterizes desires goes from the world to them. It is the world that adapts to desires. An account that identifies states that do not aim to adapt to the world as beliefs, Lackey’s reasoning goes, cannot be a good account of group belief. Thus, Lackey takes being in a good direction of fit with the world and being responsive to how the world is as a defining feature of group beliefs.
In this subsection, I extend an argument presented by Brown (Reference Brown2023). She argued that summative accounts, including the Group Agent Account, face difficulties accommodating the correct direction of fit. I argue that cases in which there is a mismatch between privately held and openly expressed beliefs show that the Group Agent Account makes the assessment of evidence-responsiveness and direction of fit wrong. Since the Group Agent Account is sensitive to the possibly concealed beliefs of the members, it assesses the group’s evidence-responsiveness and fitness to the world as a function of the members’ beliefs, irrespective of how the members of the group openly behave. However, if the project is to epistemically assess groups such as a corporation’s Board of Directors, university committees, political parties, and so on, it does not seem relevant for this project what they privately, unbeknownst to everyone else, believe. Rather, it is what they openly say or do that is relevant. To clarify this step of my argument, consider the following example.
Consider a jury whose members all happen to believe that the defendant is guilty. They are very scrupulous citizens and assess the evidence available rigorously in light of the judicial rules of their country. They in no way reveal their own beliefs to the others because those beliefs are based on racial bias, which each is ashamed to admit to the others. The defendant is in fact innocent. The admissible evidence overwhelmingly supports the innocence verdict, and they unanimously vote for the defendant’s acquittal.
In this case, the jury seems to be responsive to the evidence. Moreover, it seems a clear case of an agential group and one that yielded a group belief. That is, the jury believes that the defendant is innocent. We should assess their epistemic state as a good one. Yet, the Group Agent Account makes the group believe that the defendant is guilty, and, moreover, it makes the group unresponsive to the evidence. Thus, it makes the jury epistemically blameworthy even though it intuitively seems epistemically praiseworthy. Nonsummativists have an easy job arguing that the intuition that the jury is not epistemically blameworthy is correct. They have the resources to distinguish the epistemic blameworthiness of the individual members of the jury, who do not change their beliefs in light of the evidence they gather, from the epistemic praiseworthiness of the jury. Therefore, pluralistic ignorance shows that the Group Agent Account yields counterintuitive assessments of the jury, whereas nonsummative accounts provide an account of the intuition that the jury is epistemically praiseworthy, but its individual members are not.
In this Section 3, I have argued that pluralistic ignorance presents a counterexample to the Group Agent Account. Moreover, the problem introduced in this section is not specific to the Tobacco Company example but generalizes. If we take a group agent whatsoever and assume that it is in a state of pluralistic ignorance, we know that its members believe that p (point (i) of pluralistic ignorance), but all act and speak as if believing that not-p (points (ii) and (iii) of pluralistic ignorance). Provided that the members of the group believe that p for reasons that are not substantially incoherent, Lackey’s Group Agent Account ascribes to the group the belief that p irrespective of the details of the example. Thus, the Group Agent Account is generally unable to deal with cases of pluralistic ignorance.
I have motivated this conclusion using considerations Lackey endorses on the relationship between belief and action, considerations about critical reasoning and self-knowledge, and finally, extending Brown’s (Reference Brown2023) argument, considerations about direction of fit and evidence responsiveness. In all these cases, nonsummative accounts have an easy job in addressing the problems the Group Agent Account faces. In the next section, I argue that some plausible ways to fix the Group Agent Account lead to new problems.
4. Fixing the group agent account?
My diagnosis of the problem of pluralistic ignorance is that the Group Agent Account makes group beliefs dependent on the beliefs of the group’s members. However, beliefs can be kept private (i.e., they are not communicated to the other members) or can be misrepresented (i.e., some members mislead the other members as to what they truly believe). In this section, I address a seemingly straightforward solution to the problem. I argue that, notwithstanding its plausibility, it fixes cases of pluralistic ignorance but at the expense of new problems.
One way to make the Group Agent Account immune to the pluralistic ignorance objection is to change Condition (1) and make it range over openly shared expressions of beliefs.
Def. 2 Group Agent Account with Open Expression: A group, G, believes that p if and only if (1) there is a significant percentage of G’s operative members who openly express to 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.
The Group Agent Account with Open Expression changes Condition (1) to restrict its scope to open expressions of beliefs. The way I use “open expressions of beliefs” comprises both expressions of one’s beliefs that match the belief truly held and expressions of one’s beliefs that misrepresent the belief truly held. The way I use “misrepresent” is meant to refer to a situation where agent a believes that p, but tells the other members to believe that not-p, and the other members of the group take it to be a sincere expression of belief. For instance, the members of the Board of Directors in the Tobacco Company misrepresent their beliefs because, even though they all privately believe that what they did is wrong, they tell each other that what they did is right.
Lackey used cases of group lies to argue that existing accounts of group beliefs are not correct and argued that the Group Agent Account deals correctly with cases of group lies. Thus, it is important to show that also the Group Agent Account with Open Expression deals correctly with cases of group lies. It is also important to consider that cases of group lying do not necessarily involve pluralistic ignorance. Thus, the next paragraph addresses a point distinct from pluralistic ignorance.
Let me briefly explain why the Group Agent Account with Open Expression ascribes the correct beliefs in the case of group lies Lackey discusses. A group G lies to an agent B (Lackey Reference Lackey2021: 135) when “(1) G states that p to B, (2) G believes that p is false, and (3) G intends to be deceptive to B with respect to whether p in stating that p.” If we borrow Lackey’s (Reference Lackey2021: 28) Philip Morris example, we can see that the Group Agent Account with Open Expression ascribes the correct belief.
In Lackey’s example, after deliberations within the board, the members of the board jointly accept that smoking cigarettes is neither addictive nor detrimental to one’s health to mislead the public. They do so while being aware that each of them believes that smoking is addictive and detrimental. The joint acceptance accounts (Gilbert Reference Gilbert1987; Tuomela Reference Tuomela1992) do not constrain joint acceptance to truth norms. Thus, they regard Philip Morris as believing that smoking is neither addictive nor detrimental to one’s health. Instead, as Lackey (Reference Lackey2021: 30–31) argues, the Group Agent Account correctly ascribes to the group the belief that smoking is addictive and detrimental.
Since Lackey’s example is one of a group that discusses whether smoking is addictive and detrimental, and where the members take stances on that and express their belief that smoking is addictive and detrimental, Condition (1) of the Group Agent Account with Open Expression is satisfied. Condition (2) is also satisfied because the members of the group base their beliefs on the same body of evidence. Thus, the Group Agent Account with Open Expression ascribes to the board the belief that smoking is addictive and detrimental. Therefore, the Group Agent Account with Open Expression is in accord with Lackey’s (Reference Lackey2021: 31) desideratum.
However, the Group Agent Account with Open Expression is not generally successful. For instance, it does not solve all cases of group bullshit. Lackey used cases of group bullshit as counterexamples to standard accounts of group beliefs. Moreover, she argued that accounts of group beliefs should satisfy her Group Bullshit Desideratum. That is, they should not mistake a group bullshit as a case of group belief. Thus, it is uncontroversial that Lackey would regard a revision of her account that does not satisfy the Group Bullshit Desideratum as wrong. What I want to argue now is that the Group Agent Account with Open Expression, seemingly a simple fix for cases of pluralistic ignorance, fails to deliver correct ascriptions in cases of group bullshit.Footnote 21
Following Frankfurt (Reference Frankfurt2005), Lackey regards bullshit as speech acts performed with complete disregard for the truth. Their only purpose is to win the argument. She considers the case of the Oil Company. This company is facing pressure from environmental activists for cleaning activities carried out with polluting dispersants. In the original case, the Board of Oil Company jointly accepts to believe that these dispersants are safe but does so with complete disregard for the truth. For this reason, Lackey argues that this is a case of group bullshitting rather than a group belief. Let me adapt the original case to make it relevant to the Group Agent Account with Open Expression. Thus, let us do without joint acceptance, a concept used by Gilbert’s account, which was Lackey’s target, and use in its place members’ open expressions of beliefs. Therefore, the members of the board have now openly claimed that they believe that the dispersants are safe. The Board of Oil Company makes condition 1 of the Group Agent Account with Open Expression true. We can also assume that condition 2 is true too.
In highly professional groups, it can be well-known when some claims completely disregard the truth. A board of an Oil Company seems the kind of group whose members know very well if their dispersants are safe and that can guess the implicatures of other members’ utterances. For example, they can reliably guess that some members express the belief that these dispersants are safe just to win the case against the environmental groups. It is not necessary that the members openly agree to bullshit the public. Thus, the group should be counted as bullshitting the public rather than as expressing their belief.
We may try to further refine the Group Agent Account with the Group Agent Account with Sincere Open Expression and make Condition 1: a significant percentage of members of G sincerely openly express to believe that p. This solution may seem convincing. It addresses the pluralistic ignorance problem because now we do not consider insincere expressions of belief. Thus, we cannot have the kind of mismatch that induced cases of pluralistic ignorance. Moreover, we do not ascribe beliefs in cases of bullshit, because in such cases we do not have a sincere expression of belief. However, there is a drawback. The Group Agent Account with Sincere Open Expression does not make us able to ascribe to groups trivial implicit beliefs. For instance, according to the Group Agent Account with Sincere Open Expression, a group of mathematicians cannot be claimed to believe that 2 + 2 = 4 unless they openly claim so. However, this ascription seems desirable because it can be used to explain the behavior and the deliberations of the group. In agential groups, members may leave implicit many background beliefs. A group of mathematicians or engineers need not openly express beliefs on their background mathematical knowledge.Footnote 22 This may very well be taken for granted. Furthermore, Lackey’s original Group Agent Account is able to address cases of implicit beliefs. Thus, in this regard, the Group Agent Account with Sincere Open Expression fares worse than the account it attempts to fix.
It might be proposed that one way to address these points is to take the summation over acceptances and introduce an Acceptance-Based Group Agent Account in which a significant percentage of members accept that p.Footnote 23 Although this solution would address the problem of pluralistic ignorance, it is not without shortcomings. As I noted in Section 2, Lackey’s own motivations for introducing the Group Agent Account included her counterexamples to prominent nonsummative accounts of group belief. Among the accounts she targeted are those of Gilbert (Reference Gilbert1987) and Tuomela (Reference Tuomela1992). Both accounts construe group belief as a form of joint acceptance to behave as if believing that p. As I mentioned in Section 2, Lackey used her counterexamples to argue that these accounts attributed group beliefs when they should have attributed group lies. Lackey’s counterexamples highlight what seems a significant weakness of acceptance-based accounts; that is, that they make group beliefs largely voluntaristic and not aimed at the right direction of fit.Footnote 24 Moreover, acceptance-based accounts face problems in attributing group background beliefs (Lauffer, Reference Lauffer2022). Thus, we have reason to believe that Lackey would not favor taking the summation over acceptances, and we also have reason to believe that taking the summation over acceptances leads to its own problems.Footnote 25 Moreover, even if an acceptance-based account of group beliefs that addresses Lackey’s counterexamples can be proposed,Footnote 26 endorsing a revision of the Group Agent Account in terms of a summation of acceptances would, by the taxonomy proposed in Section 2, make the Acceptance-Based Group Agent Account a nonsummative account. Recall that the distinction between summative and nonsummative accounts turns on whether an account of group belief takes the summation over individual beliefs.Footnote 27 When an account takes the summation over individual beliefs, it is a summative account; when it does not, it is nonsummative. Thus, an acceptance-based group agent account would support my conclusion; that is, cases of pluralistic ignorance present a strong case in favor of nonsummative accounts of group belief.
Finally, pluralistic ignorance is a problem for any summative account. It is constitutive for summative accounts that at least some individual members believe the belief is attributed to the group. Individual members’ beliefs are a necessary condition of group beliefs. It is constitutive for nonsummative accounts that individual members’ beliefs are neither necessary nor sufficient for group beliefs. Thus, if my argument succeeds, it presents a strong case against summative approaches to group beliefs and in favor of nonsummative ones.
5. Conclusions
To conclude this article, let me sum up its main points. The Group Agent Account is concerned with agential deliberative groups. I have argued that it yields problematic attributions of beliefs when there is a systematic mismatch between privately held and openly expressed beliefs. To that end, I employed an example based on cases of pluralistic ignorance. Pluralistic ignorance is a well-studied phenomenon in the social sciences that has been observed in a variety of experimental settings. I have argued that the Group Agent Account yields wrong belief ascriptions in cases of pluralistic ignorance and that this result does not depend on the specific assumptions of the example considered. I concluded that pluralistic ignorance presents a serious problem for the Group Agent Account. Rather, it is a source of systematic counterexamples.
I have discussed two seemingly natural improvements of the Group Agent Account that purportedly solve the pluralistic ignorance problem. I called these revised accounts the Group Agent Account with Open Expression and the Group Agent Account with Sincere Open Expression. However, I argued that these solutions are far from satisfactory. Instead, they face their own problems.
Pluralistic ignorance might seem a straightforward and superficial objection, worth a question, perhaps, but not a full discussion. I disagree. The structure of the example is simple, but pluralistic ignorance connects a foundational debate in social epistemology to a host of important discussions concerning the relationship between group belief and action, group critical reasoning, and group evidence responsiveness.
Moreover, cases of pluralistic ignorance are easily addressed by nonsummative accounts. For these accounts, pluralistic ignorance is not a source of problems because they deny that individuals’ beliefs are either necessary or sufficient for group beliefs. Thus, cases of pluralistic ignorance are just another example in which group beliefs diverge from the beliefs of the members of the group. Therefore, the discussion carried out in this paper provides evidence in support of nonsummative accounts.
Finally, as I said in the introduction, the debate of group belief is not some kind of niche debate whose results do not have widespread repercussions. The debate on group beliefs is a cornerstone of social epistemology because it concerns one of the foundational topics of the field. A result in the debate on group beliefs can have important consequences because it affects other subfields. Moreover, since the debate on group beliefs provides a foundation for our epistemic evaluation of agential groups, it can also influence debates outside social epistemology and philosophical circles.
Acknowledgments
This research has been supported by generous funding from the Deutsche Forschungsgemeinschaft (DFG, GRK 2503) and the Italian Ministry of Research and Education through a PRIN–PNRR-funded program (grant number: P2022A8F82_001, CUP J53D23017180001). The Department of the Arts at the University of Bologna supported a research visit, during which this paper was completed, at the Arché Philosophy Research Centre, University of St Andrews, through the Marco Polo Research Scholarship. This paper has benefited from much feedback. I thank the audience and the organizers of the British Society for the Theory of Knowledge 2024 Conference at the University of Glasgow, as well as people attending the Knowledge and Science Colloquium at the University of Hamburg and the online Bologna–Bonn–Padova (BoBoPa) seminar for precious feedback on earlier versions of this paper. I particularly thank Alexander Bird, Jessica Brown, Filippo Ferrari, Sanford Goldberg, Thomas Kroedel, Sebastiano Moruzzi, Daniele Sgaravatti, and Giorgio Volpe for feedback that greatly improved this paper.
Open access
