When is a group justified in believing something? Jennifer Lackey articulates the importance of this question in the following way.

If we do not understand the justification of group beliefs, then we cannot make sense of our widespread epistemic attributions to collective entities—of evidence they have, or should have, and of propositions that they know, or should have known. Moreover, the justificatory status of such beliefs matters a great deal to whether groups are morally and legally responsible for certain actions and, accordingly, the extent to which they ought to be held accountable. (2021, pg. 55)

Despite its importance, surprisingly few have addressed the question. Those who have addressed it typically take one of two approaches. The first approach is the more traditional deflationary approach wherein group justified belief is to be understood purely in aggregative terms. To figure out what the group is justified in believing just add together the justification of the beliefs of some critical mass of the members (List, 2005; Goldman, 2014; Dunn, 2021). The second is the inflationary approach, which argues that the justificatory status of a group’s belief is not necessarily reducible to the justified beliefs of the individual members (Mathiesen, 2011; Schmitt, 1994; Hakli, 2011; Simion et al., 2022; Silva, 2019). Inflationary views typically result from so-called divergence arguments which rest on cases in which it seems natural to ascribe a justified belief to the group agent but not the members or vice versa.

Recently, however, Lackey (2016, 2021) has developed a third approach that is critical of the previous two, but makes use of elements of both, arguing that groups are indeed “epistemic agents in their own right, though ones whose justified beliefs are constrained by the epistemic status and normative obligations of their individual members,” which she calls the Group Epistemic Agent Account (pg. 82). Importantly, these constraints commit Lackey to denying that there are any genuine cases of divergence. It is this commitment that the present paper calls into question.

Much of the dialectic has attempted to understand group justified belief along similar lines as individual justified belief. For instance, Goldman’s (2014) deflationary account is an extension of his process reliablism to group beliefs, and Lackey’s account, along with Silva (2019) attempts to incorporate both evidentialist criteria and responsibilist diachronic normative requirements common in accounts of individual justification. In the same spirit, this paper aims to investigate how another increasingly popular notion relevant to epistemic justification might apply to groups. That notion is often termed ‘pragmatic encroachment’ and it is most famously developed by Fantl and McGrath (2002, 2007, 2009, 2012), Hawthorne (2004), and Stanley (2005) among others.Footnote 1

Briefly, the theory of pragmatic encroachment claims that certain pragmatic factors play a role in determining whether you are justified in believing (and hence know) a proposition. Such factors are typically understood in terms of what is at stake if you are wrong. For instance, the fact that you have a career defining meeting makes a difference to whether you ought to double or triple check the address of the building where the meeting is being held. Similarly, the possibility that a child may get hurt matters a great deal to whether a single cursory check of the playground equipment is sufficient to justify a belief in its safety. In other words, whether the amount of evidence you have justifies you in believing that p depends in part on how badly things could turn out if p is false. If this is true of individual beliefs, then it is worth exploring whether it is true of group beliefs. Interestingly, when we apply the theory of pragmatic encroachment to group justification, we reveal a novel form of divergence argument because the practical stakes for the group can come apart from the stakes of the members, or so I shall argue.

Put slightly more systematically, in this paper I will defend the following argument:

Pragmatic Encroachment Divergence (PED)

  1. i.

    Practical stakes make a difference to what an agent (group or individual) is justified in believing.

  2. ii.

    The practical stakes of a group can come apart from the practical stakes of the (operative) members.

  3. iii.

    Therefore, it is possible for the justified beliefs of a group to diverge from the justified beliefs of its members.

So, then, the claim of the paper is conditional: if pragmatic encroachment is true of group agents, then we should be inflationists about justified group belief. Here is how the paper will proceed. Section I is dedicated to explaining my starting assumptions regarding the nature of group belief, the type of group I’m interested in, what it means to claim a belief is justified, and finally to explain the particulars of Lackey’s influential account and the parts of it with which the theory of pragmatic encroachment is compatible and with which it might conflict. In section II I defend premise (i) by briefly explaining the theory of pragmatic encroachment, and then I show how it also applies to group beliefs. In section III I defend premise (ii) by offering cases of two different types. Those where the stakes for the group are significantly higher than they are for the members, and those where the stakes for the group are lower than they are for the members. Importantly, both kinds of case represent a separate type of divergence argument.

1 Background and Lackey’s view

Since my aim in this paper is to discuss justified group belief it is necessary before I begin to explain the kind of group that I’m interested in, what I mean when I say that such groups ‘believe’ something, and what it means to say a belief is justified. Moreover, since my target is Lackey’s deflationary view, it is necessary to explain her view in detail and which part of it my arguments are meant to target.

The epistemology of groups as I understand the dialectic has concerned itself primarily with groups that constitute an organization, institution, or that might otherwise be thought of as a group agent capable of intentional action, being subjects of duties, and deliberating (Tuomela, 2013; Gilbert, 2004; List & Pettit, 2011; Collins, 2019; Schwenkenbecher, 2020; Lackey, 2021). As Lackey explains, the motivation for concentrating on groups as agents is that they seem to possess the specific feature of being subject to normative evaluation. “Put succinctly, if we can properly hold a group, G, responsible for φ-ing, then this is sufficient for regarding G as a group in the sense relevant for this project” (Lackey, 2021, pg. 7). Henceforth, following Lackey, when I speak of a ‘group’ I mean a group agent that is the proper subject of normative evaluation.Footnote 2

Having said that, I aim to be as neutral as possible regarding the ontology of group beliefs. I am arguing for an inflationary view, so I accept that group agents can be appropriately ascribed justified beliefs that are not reducible to the justified beliefs of the members, but I make no claims about the nature of such beliefs. Perhaps group beliefs should be understood in functionalist terms like those endorsed by List and Pettit (2011). Or perhaps reader’s prefer Gilbert (2002)’s and Hakli (2011)’s use of Cohen’s (1992) distinction between belief and acceptance, labeling it “group acceptance” which Mathiesen (2011) also invokes to sidestep the debate about the ontological status of group beliefs. The point is, I will proceed as though groups can believe things and leave it to readers to employ their favored understanding of ‘group belief’. Whatever notion we employ, I take it that such beliefs will be subject to norms of justification, so what I have to say will apply.

Finally, when I say a belief is justified, I mean it in the binary sense that the belief is based on evidence sufficient to reach a threshold beyond which the belief is no longer open to normative epistemic criticism, and one has satisfied one’s normative epistemic requirements (e.g., one has gone to sufficient lengths to gather more evidence). Conversely, a belief is unjustified if it is based on evidence that is insufficient to meet that threshold, or else one has not satisfied one’s normative epistemic requirements, thereby leaving one open to epistemic rebuke.Footnote 3 Put another way, if I am justified in believing that p, then no one can appropriately claim that I ought not believe that p. Conversely, if I am not justified in believing that p, then it is appropriate to claim that I ought not believe that p. As we shall see in the next section, whether it is appropriate to say that I ought not believe that p will partly depend on how I am going to put p to work in action. In other words, whether my belief is open to criticism will depend a great deal on whether acting on that belief is itself open to criticism and vice versa (Fantl & McGrath, 2009; Biebel, 2018; Hawthorne & Stanley, 2007).

Before I explain pragmatic encroachment, however, I should explain what part of Lackey’s view is incompatible with the application of pragmatic encroachment to group beliefs. Here is her view:

The group epistemic agent account

A group G justifiedly believes that p iff:

(1) A significant percentage of the operative members of G (a) justifiedly believe that p, and (b) are such that adding together the bases of their justified beliefs that p yields a belief set that is coherent, and

(2). (a) Full disclosure of the evidence relevant to the proposition that p, (b) accompanied by rational deliberation about that evidence among the members of G (c) in accordance with their individual and group epistemic normative requirements, would not (d) result in further evidence that when added to the bases of G’s members’ beliefs that p, yields a total belief set that fails to make [it] sufficiently probably that p. (pg. 83)

For Lackey, (1a) commits her to a denial of divergence arguments and (1b) ensures that the belief is properly sensitive to epistemic features that track the truth. The justification for the group belief comes from the justification possessed by the members, but different members can be justified for different reasons—the basis for their beliefs can be different, but for the group belief to be justified those bases cannot conflict. Moreover (2) ensures both that the group’s evidence is reducible to the evidence possessed by the individual members, and that the group, as Silva (2019) puts it, believes responsibly by performing all their individual and group epistemic duties. This is thanks to (2c).

Others have challenged Lackey’s account by offering divergence cases. For instance, Silva argues that (1a)Footnote 4 is false because a group can come to a justified belief that is based on very complex and hard to follow evidence. Later each member struggles to recall the evidence because it is so complex (say, it contains a lot of charts that are hard to recall if one is not looking at them). Because each member has this struggle, imagine that they all become much less confident in their personal beliefs and withhold judgement. Yet no one reveals this change of opinion to anyone else so that each goes on acting in their role in the group under the assumption that the others have not changed their judgement, and so neither has the group. In such a case, Silva claims that it seems natural to think of the group as having a justified belief even though all the members now withhold judgement. This is an instance of a divergence argument. Others have offered similar divergence arguments (Schmitt, 1994; Mathiesen, 2011; Koscholke, 2020), but I think none so far are quite convincing. Silva’s case, for instance, is just very difficult to imagine. After all, plausibly, I can have a justified belief that is based on good evidence, even if I can no longer remember precisely what that evidence is. As such, it is difficult to imagine the group members withholding judgement rather than maintaining their belief in full strength because they are highly confident in what the evidence points to, even if they cannot recall the specifics of that evidence.Footnote 5

Even so, I think (1a) is false. This is because the stakes play a large role in determining the threshold of sufficient probability in (2d) which itself plays a large role in determining what one’s normative epistemic requirements are in (2c). So then, what I shall attempt to show in what follows is that the theory of pragmatic encroachment affects (2c) and (2d) in such a way that makes (1a) false.

2 Practical stakes make a difference to what any agent is justified in believing

The theory of pragmatic encroachment is largely compatible with Lackey’s view in that it is consistent both with treating groups as epistemic agents in their own right, and with the claim that group justified beliefs are importantly constrained by the epistemic status of its members. However, contrary to Lackey’s Group Epistemic Agent Account, pragmatic encroachment will allow the justified beliefs of the group agent to diverge from those of its members. But, to see why, we first need to understand both the theory of pragmatic encroachment and how it applies to group agents.

2.1 Pragmatic encroachment and individual belief

To begin to understand the thesis of pragmatic encroachment, consider the following story.

Sarah wants to go to a quaint local café for lunch on her day off. Getting to the café requires taking the underground train, which has two lines that board from the same platform. The express train which goes directly to her destination without making any additional stops, and the scenic train, which takes much longer because it makes several stops along the way. Sarah is in no rush to get to lunch, but still, it would be nice to know whether the next train is the express. She asks the person waiting beside her if the next train will be the express train. The friendly stranger replies, “Yep, it’s the express train, I heard the clerk say so when I got my ticket.” Sarah, having no good reason to doubt the stranger, believes what he says.Footnote 6

If we are told nothing further, it is natural to describe Sarah as justified in her belief that the next train will be the express. Her belief is based on testimony from someone she has no reason to doubt and who has revealed their source of information as likely reliable. It seems like nothing more is epistemically required from Sarah; it would therefore be inappropriate to criticize her belief. Now consider a slightly different case.

Nathan has a meeting at the immigration office to turn in the forms necessary to get his work permit. He knows that if he is late by more than five minutes, the meeting will be canceled, and he will have to reschedule. Unfortunately, the next available date is too late, so he will be forced to leave the country and lose his work contract. Nathan needs to take the underground train to get to the meeting, but it must be the express train, or he will be late, and his goose will be cooked! Nathan, worried that he might get on the wrong train, nervously asks the stranger waiting next to him if the next train is the express. The helpful stranger replies, “Yep, it’ll be the express, I heard the clerk say so when I got my ticket.” Nathan, understandably cautious, thinks to himself of all the reasons this stranger might be wrong. He could have bad information. Perhaps the guy at the ticket counter was talking about a different train. The stranger probably doesn’t really care if it’s the express train anyway, what’s it to him? And besides, the schedule might have changed! Nathan concludes that he just can’t risk it. Under ordinary circumstances he’d take the stranger’s word for it, but this meeting is too important. He needs to be sure. So, he goes to find out for himself.

Nathan, it seems, is not justified in believing that the train will be the express, which explains why it is necessary to investigate further. So, we have competing intuitions: Sarah is justified in believing based on the stranger’s testimony, but Nathan is not. The theory of pragmatic encroachment embraces both intuitions. The reason for the discrepancy in justificatory status, according to pragmatic encroachment, has nothing to do with their epistemic situation, for their epistemic situations are the same. We may even posit that the stranger’s testimony makes both of their credence rise the same amount: they both have a credence of 0.8 that the next train will be the express. The difference between Sarah and Nathan is just this: for Sarah, 0.8 credence is good enough, but for Nathan, for whom the truth is very important, 0.8 credence just isn’t good enough. In other words, the practical stakes make a difference to whether one’s belief is justified.

This represents the main claim of pragmatic encroachment, and its significance is that it constitutes a denial of what is typically considered the orthodox view, which Fantl and McGrath label ‘Purism’:

Purism

For any subjects S1 and S2, if S1 and S2 are just alike in their strength of epistemic position with respect to p, then S1 and S2 are just alike in whether they are in a position to know that p. (2009, pg. 28)

For Fantl and McGrath, being in a position to know that p comes to the same thing as being justified in believing that p (a claim they argue for at length, see Chap. 4), and, as such, the claim that factors unrelated to one’s epistemic position can nevertheless affect whether one is epistemically justified constitutes a denial of Purism.Footnote 7

Pragmatic encroachment also makes salient another important relationship: the relationship between justified belief and action. On the pragmatic encroachment picture, justified beliefs are tied very closely to action such that the degree of evidence required for a belief to be justified is the same as the degree of evidence required for one to be justified in acting as if that proposition is true. Sarah can justifiedly act as if the next train will be the express because she’s reasonably sure that it will be. Nathan cannot justifiedly act as if the next train will be the express because, if it turns out to be the scenic train, then his entire life is turned upside down. If his evidence isn’t good enough to support action, then neither is it good enough to support belief. Fantl and McGrath describe this relationship using the following biconditional, which I endorse:

You are justified in believing that p iff p is warranted enough to justify you in φ-ing, for any φ. (2009, pg. 123)

In other words, only once Nathan’s evidence reaches a sufficient threshold, is it then permissible for him to get on the train. But being sufficiently certain that it is the express train is just to claim that he is justified in believing it. Moreover, as several theorists have pointed out, it is not just the stakes of the believer that matter. The same arguments apply to the moral consequences if one’s beliefs turn out false.Footnote 8 That is, if acting on a belief could lead to foreseeable morally bad outcomes, then that too makes a difference to how much evidence is required (and the lengths one must go to gather more) before a belief is justified. This is one form of what is referred to as ‘moral encroachment’, but since the claim is also about the consequences of action, for ease of exposition I shall include this type of moral encroachment under the heading ‘pragmatic encroachment.’Footnote 9 Still, it is important to point out that the stakes are not necessarily limited to the personal interests of the believer, what matters is that one’s evidence is good enough to offset the risk of things going wrong in general should one’s belief turn out false – in other words considerations like “someone might get hurt” matter as much or even more than considerations like “I might get fired” or “it could cost me a fortune.”

Of course, someone who endorses Purism will point out that pragmatic encroachment has some very strange implications. How can the same evidence justify a doxastic state for one person, but not another? Surely, if Sarah is justified in believing, then so is Nathan, or conversely if Nathan is not justified in believing, then neither is Sarah. Indeed, the staunchest of purists like Feldman (2000) and Feldman and Conee (1985) insist that whether Sarah and Nathan’s beliefs are epistemically justified depends only on the strength of their evidence. Whether it is permissible to act on that belief is a separate question, and not relevant to that belief’s epistemic status.

While I haven’t the space to provide a full argument, I should at least provide some motivation for favoring pragmatic encroachment over Purism. I grant that it sounds a bit strange to claim that Sarah and Nathan share the same evidence and yet only one is justified in believing, but the theory of pragmatic encroachment provides a very cogent explanation for why this is so. Moreover, Purism has its own challenging puzzles to solve.

Consider that, if Purism is true then either both Sarah and Nathan are justified, or neither is. The first puzzle for Purism is to find some non-arbitrary way to determine which claim is the true one. As Owens (2000, pg. 26) asks by way of challenging Purism, “how are you going to tell us, in purely evidential terms, what level of evidence is needed to justify belief?” In other words, Purism must explain which threshold of evidence is the right one, Sarah’s or Nathan’s, without appealing to pragmatic factors, and that is no small task. Even if that can be done (and I doubt that it can), in my view Purism struggles to maintain the obvious connection between justified belief and action.

In other words, if what we epistemically ought to believe is divorced from practical or moral norms in the way Purism demands, then it makes the study of epistemology hopelessly myopic. There would be no reason for anyone other than epistemologists to care about epistemically justified beliefs (or knowledge) if they are not intimately connected to our reasons for action. Indeed, Lackey should welcome this. Recall from the quote in the introduction that one of the main motivations for inquiry in the first place is that whether a belief is justified “matters a great deal to whether groups are morally and legally responsible for certain actions and, accordingly, the extent to which they ought to be held accountable”. To my mind the biggest upshot of pragmatic encroachment is that it provides a very straightforward and plausible explanation of the connection between justified beliefs and these other normative realms (Biebel, 2018). If the importance of discussing the justification of group beliefs comes from the connection between justified belief and moral and legal responsibility, then, given that such normative standards are the very considerations that constitute pragmatic encroachment, it is worth exploring how it applies to group agents.

Of course, these remarks are hardly an argument. Many who hold more orthodox, Purist views, including Lackey (2010), are critical of pragmatic encroachment, and offer their own explanations for the connection between justified belief, knowledge, and action.Footnote 10 The discussion above is meant only to motivate the application of pragmatic encroachment to the epistemology of groups. The first step in this application is to establish (i) that pragmatic encroachment applies to group agents just as it does to individuals.

2.2 Pragmatic encroachment and group agents

Consider the following example.

Flatbed Incorporated (Inc.) are a company dedicated to providing environmentally safe fertilizer to rural farms by transporting it in large crates that weigh around 2,000 pounds on the backs of flatbed trucks. They are in the market for new cables to keep the crates secured to the trucks. The company has formed a committee which must decide unanimously on which cable supplier to choose. There are no specific regulatory standards for the strength of the cable, although they have a slight preference for a stronger cable in case of unforeseeable circumstances if it isn’t too expensive (better safe than sorry as long as it doesn’t break the bank). Bill, a cable sales rep, offers a cable that is flexible and weather resistant, and Bill informs them that internal testing indicates that the cable’s tensile strength exceeds 5,000 pounds of force (lbf), although those tests have not yet been independently verified or certified by any regulatory boards. Given that there are no specific regulatory standards, company policy dictates that they need not wait for independent verification or board certification. As such, the committee is satisfied with the internal tests. They agree to buy the cable, and, in their report, they include in their reasoning their endorsement of the claim that the cable’s tensile strength exceeds 5,000 lbf. As a result, the company publishes those claims as ‘facts about our equipment’ on its website.

This case roughly mirrors the case of Sarah above. Flatbed Inc. seems justified in believing that the rope will hold 5,000 lbf. After all, their evidence makes it probably true. Moreover, not much turns on the actual tensile strength of the cable. If it turns out to be weaker than 5,000 lbf, then it will still hold the crates under normal working conditions, and there is no enforced industry standard that they will be violating if it turns out the cable has a tensile strength of only, say, 4,500 lbf. The worst that will happen is that they’ll have to adjust their website for accuracy, and perhaps they will encounter an extremely unusual circumstance where a cable snaps. Moreover, even if a cable did snap, the risk of human injury is very remote. Let us stipulate that it would not matter much if it turned out that the cable’s tensile strength was not 5,000 lbf. Given this stipulation and the strength of their evidence, it seems natural to describe Flatbed Inc. as justified in believing that the tensile strength of the cable is 5,000 lbf and thereby justified in asserting it on their website.

Now consider this slightly different case.

Bill the cable sales rep is making another pitch to a company, Skywashers Corporation (Co.), that has also formed a committee that must unanimously decide whether to buy his cable. Skywashers Co. is a company that specializes in window washing for high rises and skyscrapers. They will be using the cable to hold metal platforms many hundreds of feet above the ground on which their employees stand to wash windows. There are strictly enforced regulations requiring the cables they use to have a minimum tensile strength of 5,000 lbf. Violating the standard incurs a hefty fine, and for very good reason. Even a remote possibility of a cable snapping means an unacceptable risk of serious injury or death of a company employee. However, Skywashers Co. takes the safety of their employees very seriously, and they would take these precautions even if there were no sanctions for violating industry standards. Bill informs them that internal testing has shown with a high degree of confidence that his cable’s tensile strength exceeds 5,000 lbf, although those tests have not yet been independently verified or certified by any regulatory agency. Unfortunately, the committee votes to reject a deal with Bill. In their reasoning they state that there is insufficient evidence to show that the cable meets the required tensile strength of 5,000 lbf.

Despite sharing the exact same evidence as Flatbed Inc., it seems that, intuitively, Skywashers Co. would not be justified in believing the cable’s tensile strength exceeds 5,000 lbf. This is because the stakes for Skywashers Co. could not be higher. Not only will they incur a large fine if those internal tests are inaccurate but using unproven cable creates an unacceptably high risk to their employees – something Skywashers Co. would consider decisive even if sanctions were not an issue. Given the high stakes, they must take extra steps to be certain that the cable’s tensile strength exceeds 5,000 lbf, which means requiring better evidence such as independent verification and board certification. So then, Skywashers Co. is not justified in believing the cable has a tensile strength exceeding 5,000 lbf, despite sharing the same evidence as Flatbed Inc.

If this is the right diagnosis, then pragmatic encroachment applies to groups in roughly the same way it does to individuals and it establishes premise (i) of PED. In the next section, to establish premise (ii) of PED, I will argue that the practical stakes for the group can come apart from those of the individual members, and hence the threshold for when evidence makes a proposition sufficiently probable to justify belief will be different for the group than it is for its members. Hence (iii) will follow: group justified beliefs can diverge from the justified beliefs of the members.

3 The practical stakes of a group come apart from the practical stakes of the members

There are two ways in which the practical stakes for a group can come apart from the practical stakes of the members: (a) the stakes for the group can be higher than they are for the individual members, and (b) the stakes for the members can be higher than they are for the group. These both represent a separate form of divergence. As such, I will address them in separate sub-sections, taking care to explain how each type of divergence is different from, and thereby avoids objections Lackey has lodged at other divergence arguments in the same vicinity.

3.1 Group stakes are higher

It is clearly possible for the practical stakes of a group to be higher than those of the individual members. To see this, consider that the standards for a jury in a US criminal trial are “beyond a reasonable doubt,” which is characterized as an exceptionally high standard of evidence. But it is clearly true that the individual members of a jury (qua individuals) can apply less restrictive standards of evidence and thereby come to a different conclusion than the jury (qua group). For example, it is possible that the members of a jury in a criminal case are all convinced by the available (and admissible) evidence that the defendant is guilty, but also recognize that the evidence does not establish the defendant’s guilt beyond a reasonable doubt. In which case, the members of the jury themselves can hold a justified belief that the defendant is guilty because the evidence is sufficient for them to each personally draw a conclusion of guilt, but still rationally vote ‘not guilty’ because they each recognize that the evidence isn’t good enough for a criminal jury to draw a conclusion of guilt.Footnote 11

This brings up an important question: how is it possible that an individual juror can both have a justified belief that the defendant is guilty and yet not be justified in voting ‘guilty’ in their capacity as a jury member? After all, if the biconditional offered by Fantl and McGrath that I endorsed above is true, then on the face of it the individual jurors would not be justified in believing the defendant is guilty unless they were also justified in voting ‘guilty’, so some explanation is needed.

As others have pointed out, it is possible for the individual members to set aside their own rational point of view, sometimes called the “I-mode”, and take up the rational point of view (including interests and implications) of the group agent when acting in their capacity as a group member, sometimes called the “we-mode” (Hess, 2014; Schwenkenbecher, 2022; Tollefson, 2002; Tuomela, 1995, 2013; Bacharach, 1999). When doing so, an individual jury member can, in principle, recognize that the stakes from I-mode (e.g., when they act separately from the group) are lower than we-mode and vote in their capacity as a jury member, considering only we-mode stakes. For instance, if an individual member of the jury were to believe that the defendant is guilty and then put that belief to work in action outside their role as a member of the jury, from the I-mode, very little of consequence is likely to occur. Perhaps she will avoid the defendant in the future, or even engage in interpersonal blaming practices on her own. But interpersonal blame is typically a relatively low stakes affair (Coates, 2016), and this is especially true for jurors since they can have no personal connection to the defendant.Footnote 12 Conversely, if the jury believes that the defendant is guilty and then puts that belief to work in action, then the defendant will be given a criminal punishment and clearly her life will be significantly negatively affected. So, then, the case of a jury seems to represent one in which the group agent’s stakes, and therefore standards of evidence (we-mode), can come apart from the stakes, and therefore standards of evidence of the individual members (I-mode).

Schmitt (1994) offers a divergence argument using the case of a jury, and Lackey makes a convincing argument for why Schmitt’s version won’t work, so it is important to highlight where my version is different and does not succumb to the same criticisms. Schmitt’s version of the case involves the jury being instructed to ignore some very compelling hearsay evidence because it is not admissible to the court. In such a case, Schmitt claims the individuals have different evidence than the jury, and hence the jury ought to come to a different conclusion than the members. But, as Lackey rightly points out, being asked to ignore certain evidence because it is inadmissible is a legal difference not an epistemic one. As such the jury is legally required to ignore inadmissible evidence when coming to a legal verdict, but that has nothing to do with their epistemic justification, nor the total evidence the jury has. As such, a verdict that ignores perfectly good hearsay evidence is legally justified, but it does not represent what Lackey takes to be the actual epistemically justified belief of the jury: which is that the defendant is guilty.

My version does not have this problem because there is no difference in evidence. My claim is that the divergence occurs because the threshold of evidence required for the jury to be justified is different than it is for any of the individual members. This is because the standards of evidence are partly determined by the stakes; by how bad it would be if they got things wrong. It is prima facie, far more important that the jury gets things right than it is for any of the individual members when they are not acting in the context of the jury itself (we-mode). The difference in standards of evidence is a legal requirement to be sure, but one very good reason we have such a legal requirement in the first place is just that we recognize that the collective judgement of the jury has such high stakes. We had better be sure beyond a reasonable doubt that a person is guilty before we subject her to state directed criminal sanctions. So then, this legal difference is based on good epistemic practices; in other words, this legal difference is also an epistemic difference.

Lackey criticizes Schmitt’s case in a second way, however. In Schmitt’s version of the case, the jury excluding certain evidence is just based on the law’s requirement that hearsay is inadmissible in legal contexts. But hearsay might be very good evidence epistemically speaking. Indeed, it might be better evidence than that which the jury is allowed to consider. As such, the law’s exclusion of hearsay may get in the way of the jury coming to a true belief. Hence, legal justification and epistemic justification come apart for the simple fact that the law’s requirements are sometimes “radically disconnected from truth-conduciveness” (pg. 63). The same complaint might be lodged at my version of the jury case. After all, the practical relevance of one’s belief has no bearing on whether that belief is true. So, Lackey might insist that the practical relevance – the stakes – cannot have any bearing on whether the belief is epistemically justified because they are not truth-conducive and hence are not part of one’s epistemic position.

But again, the legal difference in my version is based on intuitively good epistemic practices. It is true that the pragmatic stakes are not properly part of one’s epistemic position. However, the practical stakes are epistemically relevant. The stakes provide a threshold that the truth-conducive features of an agent’s epistemic situation must meet—that is, the stakes help determine what counts as ‘sufficient’ in the claim that the evidence must make it sufficiently probable that p is true. So, they are not radically disconnected from truth-conduciveness. Indeed, the point of requiring such high standards of evidence is precisely because they make a belief far more likely to be true.

Aside from Lackey’s criticisms of Schmitt’s case, jury examples in general might be called into question. It might be wondered, for example, whether the jury’s verdict actually represents what the jury believes. After all, when the jury enters deliberation, they are not attempting to answer the question “do we believe that p?” They are attempting to answer the question “does the evidence support p to the degree of being beyond a reasonable doubt?” A critic will point out that these are different questions because a jury (qua group) can answer ‘yes’ to the first while answering ‘no’ to the second. That is, the jury might be convinced that the defendant is guilty but also acknowledge that the evidence is not strong enough to draw a criminal verdict of ‘guilty’.Footnote 13

I have two responses to this worry. First, a jury’s criteria for drawing a conclusion are part of their established decision procedure, and it is common for group agents to include specific evidential standards in that decision procedure. Such criteria represent the rational point of view of the group agent mentioned above (Hess, 2014). In the case of a jury this matches up well with the thesis of pragmatic encroachment because the criteria are set at “the evidence must be strong enough for a guilty verdict to be beyond a reasonable doubt.” That is, the question they are considering is whether the evidence is strong enough for the group to act by declaring a verdict of “guilty.” Hence, if the group believes the defendant is guilty, but they also think the evidence is not strong enough to declare a criminal verdict of “guilty,” then they are implicitly admitting that the group’s belief is not sufficiently justified according to the pre-determined rational point of view of the group (the we-mode), a rational point of view that is sensitive to the stakes for the group.

Having said that, my second response is that I don’t see any good reason to doubt that a jury’s verdict represents what the group agent justifiably believes. To say otherwise is to claim that a group agent can form a belief that runs counter to its pre-established decision procedures, its rational point of view. That would be very odd, given that such decision procedures are meant to be the way a group agent deliberates, draws conclusions, makes decisions, and acts. When a group agent’s decision procedure includes a pre-established criteria for evidence, then, as I see it that is the established boundary below which the group agent’s decision procedure dictates that the group agent withhold belief (or below which a belief would not be justified), even if the members themselves are justifiably convinced. Indeed, jury cases are often employed in the group agency and group epistemology literature as paradigms of group deliberation and group belief precisely because their purpose is to evaluate evidence and determine whether it is strong enough to warrant action by the group (Brown, 2022; Schmitt, 1994; Lackey, 2021).

To be clear, I am not claiming that the proposition “we all think he’s guilty, but the evidence isn’t strong enough to convict” cannot be true, but I think the word ‘we’ in that proposition does not refer to the group agent ‘the jury’ as an entity distinct from its members. Instead, it refers to a group that is not an agent, but also happens to be constituted by the same individual people, namely the collective ‘the members of the jury’. And, of course, I agree that all the members can individually, and even as a collective, conclude that the defendant is guilty while the group agent ‘the jury’ comes to a verdict of not guilty. That is just to say the beliefs of the members can diverge from those of the group agent.

Even if readers are skeptical about jury cases, many group agents have a pre-determined decision procedure that includes specific evidentiary standards. Hence, we can always look to other cases where the stakes for the organization are higher than for the individual members. Such cases are common. Consider this anecdotal case. I was recently in the process of signing a lease for a new apartment through a property management company. The agent for the company, let’s call her Jane, informed me that the company has a strict policy that requires definitive evidence of sufficient income in the form of an official document from a potential tenant’s employer, such as an employment contract or a letter from human resources (HR) stating a salary and start date. Unfortunately, the reason I needed this apartment was to begin a new position at a local university, and I had not yet signed a contract. Typically, HR provides the required letter, but I had not yet received one because, as is often the case in university administration, they were waiting for various approvals from different department heads and the process was very slow. With the deadline approaching, I showed Jane the email from the university notifying me that I got the job, and a spreadsheet HR sent me showing my salary calculation and start dates hoping that they would be enough to satisfy the company. Jane very clearly believed me because she continued to prepare additional paperwork for handing over the apartment, but she also made it very clear that the company would not accept any of those things as proof of income. They would only accept an official letter, which thankfully arrived just in the nick of time.

On the one hand, Jane believed that I met the income standards required by the company, and she was also clearly well justified. On the other hand, the evidence I provided was not good enough to satisfy the company itself, and hence the company withholds belief until I can provide evidence that they will accept. This is easily explained by the difference in stakes. Nothing of consequence would happen if Jane turned out to be wrong and I did not really have sufficient income. This is because, regardless of what Jane believes, the company will not issue a lease until they have the required evidence. At worst, Jane loses a bit of time preparing additional paperwork that turned out not to be needed. Indeed, I think that any of the individuals at the company would be equally justified in believing me. However, the fact that they would be justified in believing me does not mean that the company would be, and this is reflected in the company’s policy – its rational point of view – of requiring very specific evidence. The company policy is reasonable because the company has quite a lot more at stake than any of the individual members alone. The time, hassle, and costs involved in evicting a tenant for lack of payment are very significant, so it behooves them to demand that the evidence be impeccable before believing that a potential tenant has sufficient income.

If either the jury case or the case of Jane and the real estate management company is convincing, then we have a case of divergence where a group is not justified in believing a proposition that the members are justified in believing. This is because the stakes for the group are higher than for the individual members. However, there can also be cases in which the group’s stakes are relatively low, but the stakes for the members are high. In such a case the group will be justified in believing, but not (enough of) the individual members, creating a second type of divergence argument.

3.2 Group stakes are lower

Mathiesen (2011) has offered a case that is close to what I am looking for where the stakes of the group are lower than for the members, which is also discussed in detail by Lackey (2021). The case is one where the group and the individual members are said to have what Mathiesen describes as different risk settings. These risk settings, much like my claims about practical stakes, determine how much evidence is necessary before a belief is justified. So, we may find that “while the individuals and the group as a whole consider precisely the same evidence and they assign the same weight to the evidence, the group reaches its threshold for acceptance while no individual member has reached her threshold for acceptance” (Mathiesen, 2011, pg 41). If this is right then we may find, say, an academic hiring committee reviewing applications. The committee is given specific criteria that a candidate must meet to be considered qualified. However, each of the individual members “…may prefer to be very skeptical that anyone is truly qualified. But given that as a group they need to present a set of names to the department, such skepticism would be out of place in group reasoning” (pg. 40, italics original).

The problem with Mathiesen’s explanation of these different risk settings is that it is not clear whether the skeptical risk settings of the committee members are appropriate. Are there non-arbitrary, legitimate grounds for their stubborn skepticism? All we are told is that the committee members prefer to personally be skeptical that any candidate can be truly qualified. Without any further explanation that justifies this personal preference, the committee members seem unreasonable in their high standards (no candidate can be qualified?). At some point it will begin to seem like they are stubbornly refusing to admit what the evidence before them indicates. While a justified belief is one that is permissible given the evidence, it is also possible to possess evidence that is strong enough that withholding belief is itself impermissible and therefore unjustified. And this can happen because having such high standards is arbitrary, unwarranted, or even morally problematic.

This lack of legitimate reason to have such high standards is, I think, what leaves Mathiesen’s case open to an objection offered by Lackey. Lackey asks, “If epistemic risk settings can be determined by practical interests [such as the need to submit the name of a candidate in a job search], and such settings can justify different doxastic states, what prevents groups from manipulating their risk settings precisely to suit their unwarranted practical purposes?” (Lackey, 2021, pg. 64). The case is open to this objection because the individuals seem to be doing just that, manipulating their risk settings in an arbitrary and unwarranted way, and if the individuals can do this then so can the group agent. Because of this Mathiesen’s view that different risk settings can make a difference to group justification is vulnerable to Lackey’s insistence that it has no way of explaining why companies like Phillip Morris of ‘big tobacco’ infamy should not be extraordinarily cautious about listening to the testimony of scientists when it comes to the dangers of smoking, given that they have an extremely large financial interest at stake. Thus, if Mathiesen is right, Phillip Morris would be justified in withholding belief, “even when belief is clearly called for” (ibid.), just as the individual members of the hiring committee would be justified (individually) in withholding belief even when belief is clearly called for. This obviously won’t do.

Fortunately, I think pragmatic encroachment can help Mathiesen avoid Lackey’s objection. All that is needed is to point out that the practical stakes for Phillip Morris do not just include their private financial interests. They also include the dangers of being wrong in general both to themselves and to others. On the one hand, all else equal it is clearly true that Phillip Morris’s standards of evidence ought to increase when the company’s financial future is on the line. On the other hand, all else is not equal. It is even more clear that the risk to the company’s bottom line pales in comparison to the risks they would be undertaking if they continue to insist their products are safe. So, what prevents groups from manipulating their risk settings to suit their unwarranted practical purposes? Precisely the fact that those practical purposes are unwarranted. What makes them unwarranted is just the stakes—if they are wrong about the safety of their products millions of people will suffer or die, which obviously outweighs the financial interests of the company. Saying “We’re not sure it’s safe, we need more evidence before we can say that” is the response reasonable prudence dictates, not “We’re not absolutely sure it’s dangerous, so keep selling it until someone proves beyond a shadow of a doubt that it isn’t safe.” The latter of these is just straightforward negligence.

Similar things can be said about the hiring committee members only in the opposite direction. What prevents the individuals from manipulating their risk settings to suit their arbitrary or unwarranted personal preferences? Just the fact that they are arbitrary or unwarranted. What is needed, then, is an adjustment to Mathiesen’s case where the risk settings are not determined by an unwarranted or arbitrary over-caution or mere personal preferences.

With all of this in mind, imagine a philosophy department is growing in popularity and really needs to hire more people to fill department needs in ethics, epistemology, and ancient philosophy. Unfortunately, the administration only approves one new tenure track position. It is decided that the greatest need is in ancient philosophy, so they advertise for that position. They are a very small department of just three people so they are all on the hiring committee. Now suppose that two of the committee members will be forced to pick up the slack and teach extra courses in ethics and epistemology that are not really in their realm of expertise and that will increase their workload significantly. This will take up valuable time that will cause a significant hindrance to their research output and career prospects. As such, from their point of view, they think that if the department is going to hire someone in ancient philosophy, then that person had better be the best of the best to improve the department’s international reputation and make their personal sacrifice worth it. One of the candidates, Kaiya, is clearly more qualified than all the rest, but is she really qualified enough for the job? We can imagine the two members who are sacrificing a lot remaining skeptical that Kaiya is really qualified enough, even while also admitting that from the perspective of the department, there’s good evidence that she can at least meet the department’s needs, and hence is qualified enough for the job.

This version of the case maintains the spirit of Mathiesen’s example because the group agent that is ‘the committee’ holds a belief that the individual members are skeptical about. However, their skepticism is more warranted in this version because the individual committee members have a lot on the line. As such, from their personal point of view (I-mode), it makes sense to be a bit skeptical that the candidate is really qualified enough for the job but given that they must hire someone to meet the department’s most pressing need, as Mathiesen puts it, “such skepticism would be out of place in group reasoning” (pg. 40). Hence, we have a case of divergence where the stakes for the group are lower than the stakes for the individual members.

4 Conclusion

In this paper I have I defended an argument I labeled Pragmatic Encroachment Divergence (PED), which states (i) that practical stakes make a difference to what an agent (group or individual) is justified in believing and (ii) that the practical stakes for group agents can come apart from the practical stakes of the members. I have argued that (i) and (ii) are both true and if I am right, it follows that (iii) a group’s justified beliefs can diverge from the justified beliefs of the individual members. Hence, if pragmatic encroachment is true of group belief, then it provides two different and novel forms of divergence argument, and as such, it also provides a challenge for deflationary views like Lackey’s.