Justified Group Belief, Group Knowledge and Being in a Position to Know

Abstract

Jennifer Lackey has recently presented a new and lucid analysis of the notion of justified group belief, i.e. a set of individually necessary and jointly sufficient conditions for a group to justifiedly believe some proposition. In this paper, however, I argue that the analysans she proposes is too narrow: one of the conditions she takes to be necessary for justified group belief is not necessary. To substantiate this claim, I present a potential counterexample to Lackey's analysis where a group knows and thus justifiedly believes some proposition but there is no single group member who actually believes that proposition. I close by defending the example against the objection that the group in question does not know but is at most in a position to know the target proposition.

1. Introduction

Some philosophers think that just like an individual subject, a group of subjects can have beliefs that may or may not be justified. But if this is correct, what are necessary or sufficient conditions for a group to be justified in believing something? Jennifer Lackey (2016) has recently presented a novel account of justified group belief according to which groups can have justified beliefs responding to both evidence and normative requirements that arise only at the group level but that are constrained by the epistemic statuses of the beliefs of their individual members. The heart of Lackey's account is a lucid analysis of the notion of justified group belief, i.e. a set of individually necessary and jointly sufficient conditions for a group to justifiedly believe some proposition. It runs as follows:

Group Epistemic Agent Account. A group G justifiedly believes that p if and only if the following conditions hold:

1. 1. A significant percentage of the operative members of G

   1. (a) justifiedly believe that p

   2. (b) are such that adding together the bases of their justified beliefs that p yields a belief set that is coherent.

2. 2. Full disclosure of the evidence relevant to the proposition that p, accompanied by rational deliberation about that evidence among the members of G in accordance with their individual and group epistemic normative requirements, would not 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 sufficiently probable that p.

As Lackey convincingly argues, her account improves upon a number of rivals such as the joint acceptance accounts she attributes to Schmitt (1994) and Hakli (2011) (see also Gilbert 1989 and Tuomela 2004 for this account), Goldman's (2014) gradual account of group justification and List's (2005) Condorcet-inspired account. In particular, she points out that her account yields correct verdicts concerning justified group belief in a number of cases in which the other accounts apparently fail.

Lackey's account thus seems attractive at first sight. But the analysans she proposes is too narrow, as the present paper argues. This is because one of the conditions she takes to be necessary for justified group belief is not necessary: there are cases in which a group justifiedly believes some proposition despite condition 1 of Lackey's account not being satisfied, i.e. despite it not being the case that a significant percentage of the operative group members has the justified belief that the proposition in question is true. I will present and discuss such an example in the following.

More precisely, and to put things into context, the example to be discussed in this paper is supposed to be a case in which a group knows and thus justifiedly believes that p while no single group member believes that p. If such cases exist, then – given the well-established view that knowledge entails belief – there are also cases where a group knows that p while no single group member knows that p, as famously argued by Hutchins (1995) and more recently by Bird (2010, 2014). It also entails – again given the view that knowledge entails belief – that there are cases where a group believes that p while no single group member believes that p, as suggested by Gilbert (1987, 1989) and more recently by Silva (2019a). Thus, the example presented here neatly unifies a number of views from the social epistemology literature.¹

2. Group knowledge without individual belief

Consider the following example:

The Distributed Password.

Mrs Smith, the founder of a lemonade company wants to protect the company's secret recipe by using a complex password whose length corresponds to the number n of employees of the company. The recipe shall only be revealed if the whole group of employees decides to do so. To achieve this, Mrs Smith distributes all the parts of the password in a sophisticated way. She truthfully tells:

• • Employee 1 that character 1 of the password is such-and-such.

• • Employee 2 that character 1 being such-and-such entails character 2 being such-and-such.

  . . .

• • Employee n that character n-1 being such-and-such entails character n being such-and-such.

Now, let p be the proposition stating that character number n, i.e. the last character of the password is such-and-such. Then, the following three observations should be uncontroversial:

1. (i) By construction of the example, no employee believes that p and a fortiori no employee justifiedly believes that p. It is therefore not the case that a significant percentage of the group of employees justifiedly believes that p. In other words, Lackey's condition 1a is not satisfied.²

2. (ii) There are no bases for the justified beliefs that p simply because, as stated in observation (i), none of the employees has the justified belief that p. Hence, the set of the bases is empty and therefore cannot be coherent.³ Thus, Lackey's condition 1b is not satisfied.

3. (iii) If the employees were to put all their evidence relevant to p on the table, accompanied by rational deliberation and in accordance with their individual and group epistemic normative requirements, the evidence would make p maximally probable. This is because the total evidence entails p and hence, p is maximally and thus sufficiently probable conditional on the evidence – notice that it is a theorem of probability theory that if E entails p, then P(p|E) = 1. Thus, Lackey's condition 2 is satisfied.

The crucial, fourth observation is perhaps less uncontroversial but still very natural:

• (iv) There is a clear, intuitive sense in which the group knows the password. Consequently, the group justifiedly believes the target proposition p, i.e. that the last character of the password is such-and-such.⁴

But if observations (i) to (iv) are correct, then we have a case of justified group belief – via group knowledge – despite condition 1 of Lackey's account not being satisfied. This is just what we wanted to show. Let us turn to the next section for a discussion of this example.⁵

3. Knowing versus being in a position to know

There is an obvious defence strategy for proponents of Lackey's account: they can question the truth of observation (iv) by arguing that the group in the password example does not know the password but is at most in a position to know it. If this objection is correct, then the password example is not necessarily a case where a group justifiedly believes the target proposition p while condition 1 of Lackey's account being violated. This is because unlike knowing that p, being in a position to know that p does not entail having the belief that p, let alone the justified belief that p.

In the following, I will take a look at arguments for and against the claim that the group of employees knows the password. I will consider three arguments in support of this claim in section 3.1 and discuss two potential objections against the claim in section 3.2.

3.1. In support of knowing

The first argument in support of the view that the group of employees knows the password is a variation of the Searle's (1980) well-known Chinese Room argument: imagine a room such that only written requests can be passed into and out of the room. There are two scenarios: in scenario A, there is a single agent in the room who knows the password, while in scenario B, there is a group of agents in the room possessing the information described in the example. Now, suppose we would like to find out, only by passing requests into the room and by getting responses, whether or not whoever is in the room knows the password. In both scenarios, the responses will be roughly the same. Since by assumption, the single subject in scenario A knows the password, the same should hold for the group in scenario B, too. But then the group knows and thus justifiedly believes that p.

The second argument is based on the intimate connection between knowledge and rational action as emphasized by authors such as Hyman (1999), Williamson (2000), Fantl and McGrath (2002) or Hawthorne and Stanley (2008). Various proposals for what the exact relationship between knowledge and rational action is have been discussed, but the core idea is that whenever an action depends on the truth or falsity of some proposition p, knowing that p is necessary or sufficient for it being appropriate to treat the proposition p as a reason for action. The argument for the view that in the password example the group of employees knows the password only relies on one direction of the claim, namely that knowing that p is necessary for it being appropriate to treat the proposition that p as a reason for acting. It is not difficult to see that together with the claim that the group does not know the password it follows that it is not appropriate for the group to treat the proposition that the password is such-and-such as a reason for acting. But intuitively, it is appropriate for the group to treat the proposition that the password is such-and-such as a reason for acting – for instance, when the group's action is to reveal the secret recipe. We thus have three jointly inconsistent claims: first, the claim that knowing the password is necessary for appropriately acting on the proposition that the password is such-and-such; second, the claim that the group in our example does not know the password; and third, the claim that it is appropriate for the group to treat the proposition that the password is such-and-such as a reason for acting. It seems that the easiest way to solve this conflict is by giving up the second claim and to embrace the view that the group knows the password, thus knows that p and hence, justifiedly believes that p.

The third argument is more pre-theoretic and based on reflections upon our everyday linguistic practices: suppose that someone asked Mrs Smith what she is trying to achieve by distributing the password as described in the example. She would probably respond along the following lines: ‘Well, it is important to me that the whole group of employees knows the password so that they can reveal the recipe when needed, but also that there is no proper subset of employees that already knws it – and no single employee in particular’. From an ordinary-language point of view, this seems like a perfectly natural and reasonable thing for Mrs Smith to say. She is trying to pass on her knowledge to the group without passing it on to any proper subset of the group. And our example is to be understood as a case in which she succeeds to do so. But if this is correct, then it is also correct to say that the group knows the password, thus knows that p and hence justifiedly believes that p.

3.2. Against knowing

Let us also take a look at the arguments suggesting that the group of employees does not know the password. As far as I can see, there are two such arguments. The first is due to Lackey herself. In her (2014) discussion of Bird's (2010) conception of social knowing, she writes:⁶

there is a clear difference between knowing that p and being in a position to know that p, which itself is grounded, at least in part, in the difference between information that has been accessed and information that is merely accessible. For instance, if I have an unopened letter on my desk that contains a confession from my friend to a crime, we wouldn't say that I know that my friend committed the crime prior to my opening it and reading its contents. … Instead, we would say that I am in a position to know this. Indeed, it would be bizarre for me to assert that my friend committed the crime or to act on this by reporting her to the police if the information is merely accessible, but not accessed. (Lackey 2014: 294–5)

And she continues:

But why would the situation be any different when groups are concerned? If the discovery of an enzyme that plays a role in the development of cancer cells is published in a journal article that is accessible to, but not accessed by, any living scientist, why wouldn't we provide the same verdict as we did in the individual case: the scientific community is in a position to know this discovery, but it doesn't know it? (Lackey 2014: 295)

Lackey's argument against knowing can thus be reconstructed as two steps. The first is to establish the following conditional claim:

Knowledge-Access Principle.

If some piece of information p is accessible to a subject S but has not been accessed by S, then S does not have the knowledge that p but is only in a position to know that p.

Lackey's unopened letter example illustrates the idea very well. The second step is an argument by analogy: since the individual situation is similar enough to the group situation, the Knowledge-Access Principle should also be expected to hold in the group situation. But if the Knowledge-Access Principle holds and the password in our example is accessible but has not been accessed by the group of employees, it follows that the group does not know the password but is at most in a position to know it.

Lackey's argument is interesting, but I think it fails. This is because the Knowledge-Access Principle neither seems to hold in the individual nor in the group case. To see this more clearly, suppose that I know that today is Thursday. It then makes sense to say that I also know the disjunction consisting of the proposition that today is Thursday and some other arbitrary proposition which I have never even thought about. In other words, I never accessed this disjunction but it makes perfect sense to say that I know that it is true. This carries over group situations. Let a group know that today is Thursday and thus, know the disjunction consisting of the proposition that today is Thursday and some other arbitrary proposition the group has never even considered. We thus seem to have group knowledge without access of the relevant proposition. Hence, the Knowledge-Access Principle cannot be true.

This verdict is also supported by an observation made by Bird (2010) who seems to have anticipated the issue of group knowledge and access:

we should not expect direct access even ‘in principle’ in every case of social knowing. It is true to say that in 1945 that the Americans, British, and Canadians knew how to build an atomic bomb but the Germans, Soviets, and Japanese did not. Yet that knowledge was secret and far from accessible, even to those who would have understood it. (Bird 2010: 34)

If Bird is correct, then there are cases where a group knows something despite the fact that what it knows is not accessible. And since it is a conceptual truth that what is not accessible is not accessed, there are cases in which a group knows something that is not accessed. Hence, again, the Knowledge-Access Principle fails: Lackey's idea that it is the difference between p being accessed and p being accessible that grounds the difference between knowing that p and merely being in a position to know that p seems to be on the wrong track. Accordingly, the fact that the group of employees has not accessed the password does not rule out that observation (iv) is true.⁷

The second argument against the truth of observation (iv) goes via an objection to the argument based on the variation of Searle's Chinese Room argument. The basic idea is that there is a difference between scenario A in which a single subject knows the password and scenario B in which a group of subjects is equipped with the information described in the example. The difference is that the group in scenario B will need more time to retrieve the relevant information than the subject in scenario A who knows the password. If correct, this objection would undermine the claim that we should attribute knowledge to the group of employees.⁸

This objection is tempting, but I do not think it succeeds. I grant that there are cases in which a group like the one described in scenario B needs more time to retrieve the relevant information than the single subject from scenario A who knows the password. But similarly, critics will agree there are cases in which this is not the case: in particular, for large enough n, i.e. for sufficiently large groups and correspondingly, for sufficiently long passwords, it is obviously harder for a single agent to retrieve the relevant information than it is for a large group. Now, since our example does not depend on a specific value of n, we can simply pick an n such that the behaviour of the group matches the behaviour of the single agent who knows the password by assumption and let the corresponding case be our example. It then holds that there is no difference regarding the time needed for information retrieval. Accordingly, the objection against the variation of the Chinese-Room argument fails.

Let me close by summarizing the two preceding sections 3.1 and 3.2. Proponents of Lackey's account might try to defend the account by rejecting observation (iv). But as I have argued, the two most salient arguments for doing so seem to fail and there are at least three arguments in support of observation (iv). To clarify, I do not think that any of these arguments are conclusive. Still, it seems to me that on balance, the prospects for the view that the group knows the password are better than the prospect for the view that the group does not.

4. Conclusion

This paper was concerned with Lackey's account of justified group belief. I presented an example in which, from an intuitive point of view, a group knows and thus has the justified belief that p while Lackey's condition 1 is not satisfied. I offered three arguments in defence of the claim that the group knows the target proposition and argued that two potential objections to this claim fail. I also pointed out that the example unifies previous work in social epistemology since it entails two other claims that have been discussed in the literature, namely that there are cases in which a group knows that p but no single member knows that p and that there are cases in which a group believes that p while no single member believes that p.

Where does this leave us? I think there are two potential conclusions: one more radical and one more cautious. The radical conclusion is that Lackey's account must be given up in favour of an account of justified group belief that adequately captures the phenomenon of distributed cognition exemplified by the password case. The more cautious conclusion is that there is still some value in Lackey's account. Just like other approximately correct analyses that are just a bit too narrow or too wide, her analysis correctly captures a variety of cases of justified group belief – think, for instance, of the concept of knowledge and Gettier cases. Lackey's account might just be another analysis of that kind. Now, which conclusion is the right one? That decision is left to the readers.⁹