In this paper, I explore what gives collective testimony its epistemic credentials, through a critical discussion of three competing accounts of the epistemology of collective testimony. According to the first view, collective testimony inherits its epistemic credentials from the beliefs the testimony expresses—where this can be seen either as the beliefs of all or some of the group’s members, or as the beliefs of group itself. The second view denies any necessary connection to belief, claiming instead that the epistemic credentials of collective testimony derive from the reliability or truth-conduciveness of the statement that expresses the testimony. Finally, the third view claims that the epistemic credentials of collective testimony derive from the fact that it involves undertaking a collective commitment to trustworthiness, which makes the group susceptible to rebuke and blame if its testimony is not trustworthy. I argue that this last account holds the most promise for preserving what is distinctive about testimonial knowledge while still underwriting a robust epistemology of collective testimony.
We often learn, that is, we acquire knowledge, from groups. In fact, there are various kinds of groups whose primary function is to serve as a source of knowledge—think, for example, of the weather bureau, or a commission of inquiry, or a scientific research team. For groups like these, a central way of discharging their epistemic role of disseminating knowledge is by means of testimony: we learn from these groups, when we do, by believing them in what they tell us. How they tell us things may vary, of course: perhaps the weather bureau maintains a website and a Twitter account; the commission of inquiry holds a press conference to announce its findings; and the scientific research team publishes co-authored papers in academic journals. I take it that all of this counts as collective testimony, and all of it might, in principle, be a source of knowledge.
The fact that we acquire knowledge from group testimony raises the question of what gives group testimony its epistemic credentials. Different philosophers have given different answers to this question, and my aim in this paper is to critically examine three leading views on the epistemology of collective testimony. According to the first view, collective testimony inherits its epistemic credentials from the beliefs the testimony expresses—where this can be seen either as the beliefs of all or some of the group’s members, or as the beliefs of group itself. The second view denies any necessary connection to belief, claiming instead that the epistemic credentials of collective testimony derive from the reliability or truth-conduciveness of the statement that expresses the testimony. Finally, the third view claims that the epistemic credentials of collective testimony derive from the fact that it involves undertaking a collective commitment to trustworthiness, which makes the group susceptible to rebuke and blame if its testimony is not trustworthy. I shall argue that this last account holds the most promise for preserving what is distinctive about testimonial knowledge while still underwriting a robust epistemology of collective testimony.
2.1 Testimony and Testimonial Knowledge
Before we examine the three accounts of the epistemology of collective testimony, it may be worth briefly explaining and illustrating the guiding assumption of this paper, that we can gain testimonial knowledge from groups. Testimonial knowledge is knowledge acquired, in the right sort of way, from testimony. But there is no consensus amongst epistemologists as to what testimony is exactly, nor, for that matter, what the right way of gaining knowledge from it might be (or even whether there is a “right way”). In this paper, I will adopt the widespread (but still not uncontroversial) view that testimony should be identified with the illocutionary speech act of “telling”: there is an act of testimony just when a speaker, S, tells an audience, A, that p.
The distinctiveness of the speech act of telling can be brought out with reference to the idea that, in telling someone that p, a speaker aims at “being believed” (Moran 2005a, p. 26). There are two aspects of this that are especially worth noting. First, testimony is an inherently relational, and more specifically, addressive act: a speaker cannot perform the speech act of telling except towards some audience—whether particular (e.g., a particular person or group) or general (e.g., “the general public” or “to whom it may it may concern”). We cannot simply say, for instance, James told that p, and just leave it at that—the description will be incomplete until we specify to whom James was speaking. This distinguishes telling from certain other assertoric speech acts such as “asserting”, “stating”, which may be performed without any audience at all. Second, to say that a testifier aims at being believed is not merely to say that she aims at inducing belief in her audience, but that she aims at inducing belief in a particular way. Specifically, the speaker does not offer her word as something for her audience to make up her own mind about but instead invites her audience to take it from her that p. This distinguishes telling from, for example, the speech act of “arguing”, which also seeks to induce belief in the audience but on the grounds of independently cogent considerations, rather than the speaker’s say-so.,
This conception of testimony places certain important restrictions on what counts as “testimonial knowledge”. First of all, you cannot get testimonial knowledge from someone’s testimony without being the addressee of that testimony. Imagine that on the bus to work I overhear one person telling another that Magnus Carlsen successfully defended his world chess title the night before. Since I recognise the speaker is Norwegian, and I know that Norwegians are renowned not only for their unimpeachable sincerity but also their utter reliability with respect to the accomplishments of Magnus Carlsen, I may thereby come to know that Carlsen defended his title. But although I attain this knowledge on the basis of someone’s testimony, my knowledge is not testimonial knowledge. This is because I was not the addressee of the testimony, and so was in no position to accept the speaker’s implicit invitation to “take it from me”. Insofar as someone’s word was offered, it was not offered to me. Nonetheless, the fact of someone’s offering their word—albeit to someone else—is often itself good reason to believe what they say. So I am not denying that I can come to know what I overheard, only that my knowledge is testimonial knowledge.
Note that even when testimony is addressed to you, it is possible to come to know what you are told on the basis of the testimony without your knowledge counting as testimonial knowledge. Suppose I know that my enemy wishes to deceive me about p, and so will tell me the opposite of what she really thinks when I ask her whether p. But I also know that she is always wrong about p, to the point of being entirely reliable on the matter: she believes not-p when p is the case, and p when not-p is the case. Hence, when she tells me that p, the fact of her telling me can function, for me, as a good reason in support of what she says, and so, taking it in this way, I come to know that p on the basis of her testimony. But this surely does not count as testimonial knowledge, because I have not accepted her testimony in the spirit in which it was offered: I certainly did not believe her, or take it from her, that p.
Here, then, is how I shall construe testimony and testimonial knowledge in this paper. An act of testimony is a distinctive sort of speech act, in which one person tells another that p, therein (implicitly) inviting her audience to take it from her that p. If the audience does take it from her, then we may say that she (the audience) believed the speaker, as opposed to simply believing what the speaker said. Testimonial knowledge, then, is knowledge an addressee of testimony gets by believing the speaker in what she says, by taking it from her that p.
2.2 Group testimony: An Example
As mentioned at the outset, I think it is clear that groups often tell us things and that we often gain testimonial knowledge from their testimony. Let me illustrate this now by giving an example. The example involves what is perhaps the central mechanism through which groups give testimony, that is, through the deployment of an authorised spokesperson:
TERROR ALERT: On 24 July 2014, Benedicte Bjørnland, the head of the Norwegian Police Service, gave the following statement at an emergency press conference: “There is a serious, credible threat of a terrorist attack on Norway to take place within days.” Bjørnland was speaking on behalf of the Norwegian Police Service, who had received the threat, and had judged it to be serious and credible. The purpose of her statement was to inform the Norwegian public of the threat, so that they would exercise extra caution and vigilance.
It may seem peculiar, at first blush, to classify spokesperson cases like this as cases of collective testimony. After all, they tend to involve an individual person uttering certain words on behalf of the organisation, so it may seem more apt to conceive of this in terms of proxy agency than group agency. But to conclude that would, I think, be a mistake: specifically, it would be to confuse the locution, or utterance, with the illocutionary speech act of telling. Though it is Bjørnland who comes out with the string of words, it is the Norwegian Police Service’s mode of organisation—the fact that the Service has a particular public role, that Bjørnland holds a particular office within the Service, that she makes her statement in a particular conventional setting, and so on—that constitutes her utterance as an instance of the speech act of telling. Moreover, insofar as there are certain norms relevant to the speech act of telling—such as, for instance, the injunction to tell others only what you know—it is clear that these are not norms to which the spokesperson herself is beholden. Bjørnland herself may have no view on the matter of which she speaks, or may even hold the opposite view, without this meaning that the norm of testimony has been violated. As far the testimony being offered goes, then, Bjørnland is not the one who offers it: she is but the “mouthpiece” of the Norwegian Police Service.
So much for the question of whose speech act is involved in Norway’s Terror Alert—it is, by hypothesis, the Norwegian Police Service’s speech act and no-one else’s. But why accept that it is the speech act of telling? It seems to me that the context makes this clear enough. Specifically, the statement is not made in an effort to persuade the public or mount an argument to the effect that there is a serious, credible threat of an imminent attack. Rather, it seems that what the Norwegian Police Service offers is an implicit invitation for its audience to take it from them that the terror threat is real. As we have seen, this is the hallmark of testimony.
Assuming that this is, as it appears to be, a collective speech act of telling, let us assume further that someone (maybe someone watching the live broadcast of the press conference) takes up the invitation to believe, and in this way comes to know of the terror threat. So someone acquires testimonial knowledge from the testimony of the Norwegian Police Service. The question for epistemological theories of collective testimony, then, is what gives this testimony its epistemic credentials—what makes it worthy of being believed?.
2.3 Key Desiderata for Epistemological Theories of Collective Testimony
In what follows, I will critically examine three ways of answering this question. The first focuses on the epistemic credentials of the beliefs expressed by collective testimony; the second focuses on the reliability of the statement through which collective testimony is given; and the third focuses on the epistemic significance of the commitment the group makes in testifying.
In evaluating these different epistemological theories of collective testimony, I will make use of two key desiderata. The first desideratum is what I will call epistemological adequacy, by which I simply mean that the theory should cohere with our intuitive judgements as to whether or not knowledge is acquired by an audience in particular cases of collective testimony. The second desideratum is what I will call testimonial distinctiveness, by which I mean that the theory should reflect—or at least not undermine—what is distinctive of testimony and testimonial knowledge. More specifically, an adequate theory of (collective) testimony should not overlook the fact that testimony is a distinctive speech act, and that testimonial knowledge involves the audience responding to that speech act in a distinctive way, by believing the speaker.
3 The Belief View of Collective Testimony
The first approach to collective testimony I will explore is a version of the so-called “Belief View of Testimony” (BVT). According to the BVT, the epistemologically salient thing about testimony is that it expresses belief, and hence what gives credible testimony its epistemic credentials is the epistemic credentials of the beliefs it expresses. Philosophers who go in for this kind of view tend to see the epistemological status of a piece of testimony as a product of the speaker’s sincerity and competence. According to them, we acquire testimonial knowledge by being faithfully put in touch (this is the sincerity condition) with someone else’s reliably-held belief (this is the competence condition).
There are two broad ways in which a Belief View of Collective Testimony (BVCT) can be fleshed out: a “summativist” way and a “non-summativist” way. Although summativist approach does not have many contemporary defenders, it can be traced back to a broadly Hobbesian approach to group speech. Hobbes held that one central way in which groups can speak is by authorising a spokesperson to express the views held by a majority of the group’s members. Put specifically in terms of the speech act of testimony, a summativist view is one that holds that “a group’s testimony p expresses the views of all or some of the members of the group” (Tollefsen 2007, p. 300). On this kind of view, then, when we believe a group in what it says, what we are doing is relying the doxastic competence of (all or some of) the members of the group.
In the context of individual testimony, the BVT holds that, when all goes well in testimony, the epistemic credentials of the speaker’s beliefs are somehow carried over, or “transmitted”, from speaker to audience. Summativism about collective testimony introduces a small twist on this formula, exploiting the fact that whenever a group is on the scene, there are other, smaller agents on the scene too: the group members. So according to the summativist version of the BVCT, in a case of collective testimony it is not the speaker’s own beliefs that are expressed, but the beliefs of those the speaker represents or speaks for—i.e., the group members. But this need not detract from the credibility of the testimony, since what ultimately shoulders the epistemological burden is just p’s being competently believed, regardless of whether the one who so believes is also the speaker. In the case of TERROR ALERT, a viewer might come to know that there is a serious, credible terror threat because this is something that (all or some of) the individual people in the Norwegian Police themselves know, this member-level knowledge being what is expressed by the words of the group’s spokesperson.
There is, however, a fundamental problem with the summativist version of the BVCT. The problem is that it seems we can acquire testimonial knowledge that p from collective testimony that p without any individual member of the group believing that p. To see this, let us embellish our earlier example with the following backstory:
DIVISION OF EPISTEMIC LABOUR: Imagine that the Norwegian Police Service has a policy governing whether to inform the public of national security threats. The policy states that the public should be informed of the threat if and only if the threat is serious, credible and imminent. Imagine, further, that the task of deciding each of these issues is assigned to a different expert within the Norwegian Police Service. So if Expert A judges that the threat is serious, and Expert B judges that it is credible, and Expert C that it is imminent, then the collectively endorsed view of the Service is that the threat is serious, credible and imminent. If so, then, given their policy, they are committed to informing the public of the threat, and this is something they do by having Benedicte Bjørnland make an announcement at a press conference. Now imagine that Bjørnland herself has no settled view on any of the issues, much less their conjunction (she prefers to “leave it to the experts”, and is happy to go along with their findings). Furthermore, the experts themselves have differing views on the various issues, all the while acknowledging one another’s expertise, and hence authority, to decide for the group on the issue each has been assigned. What this means is that when the Norwegian Police Service announces, through Bjørnland, that there is serious, credible threat of an imminent terrorist attack, they testify to something which none of the individual members personally believes.
The problem for the summativist version of BVCT is that it fails to fulfil the desideratum of epistemological adequacy. While no member believes what the group tells the public, the group’s testimony is still, intuitively, worthy of being believed. The distribution of epistemic labour and deference of non-experts to experts—i.e., the very organisational structures that produce the lack of belief amongst the members—seem to be impeccable from an epistemic standpoint. So the Norwegian Police Service in this case exemplifies a well-designed, epistemically responsible organisation, functioning exactly as it should, and yet the summativist is bound to deny that their statement is a potential source of testimonial knowledge. For if no member believes what the group tells the public then, on the summativist approach, the group’s say-so is left completely without epistemological credentials: it no longer expresses anyone’s competently believing that p.
This kind of problem for the summativist version of the BVCT is what leads Deborah Tollefsen (2007, 2009, 2011) to develop a non-summativist approach in its stead. Whereas the summativist version holds that collective testimony expresses the views of group members, Tollefsen’s non-summativist view sees it as expressing the view of the group itself. Collective testimony can be a source of testimonial knowledge, then, when it faithfully puts one in touch with a competent collective belief.
According to Tollefsen, in order for someone to be a source of testimonial knowledge, they must exhibit doxastic stability: “our practice of relying on the word of another presupposes that the other has a steady mind” (Tollefsen 2009, p. 17). But groups, just like individuals, can satisfy this requirement. That is, they can undergo a process of “steadying the mind”, with a view to speaking their minds — or, to put this in terms of belief, they can undergo a process of settling and then expressing, via testimony, a collective belief. To illustrate this, Tollefsen borrows an instructive example from John Beatty’s (2006) paper, “Masking disagreement among experts”. The example involves a report produced in 1956 by a panel of geneticists appointed by the US National Academies of Science, concerning the genetic risks associated with exposure to radiation. Although there was substantial disagreement amongst the members of the panel concerning those risks, the panel recognised the need “to provide a unified story to the public rather than reveal their dissenting opinions” (Tollefsen 2011, p. 17). Thus they undertook a lengthy, arduous process of group deliberation, of discussion, debate, drafting and even voting, in order to arrive at a report they could all sign off on—a jointly acceptable view, which they then communicated to the public through the publication of the report in popular media.
According to Tollefsen, the panel’s report “is a paradigm case of group testimony” (Tollefsen 2011, p. 18) because, through this process of group deliberation, the panel successfully steadied its mind, reaching “a consensus regarding the information they wanted to convey to the public” (Tollefsen 2011, p. 18). But it is important to note that the consensus reached was not a consensus of opinion on what the report was about, i.e., the genetic risks associated with radiation exposure. Rather, the process of group deliberation was aimed producing consensus with respect to which set of claims the panel would let stand, publicly, as their collective view. So what the “doxastic stability” of the group in this case really amounts to is their having settled for themselves the question of what their collective view will be, or what view they will represent themselves as having.
One might think that this betrays a degree of insincerity in the panel’s report, making it strange that Tollefsen should call it a “paradigm case” of collective testimony. This impression may be encouraged by Beatty’s original article, which highlights the fact that information about internal disagreements in the panel was intentionally withheld from the final report (hence the paper’s title, “Masking disagreement…”). But Tollefsen does not see any of this as undermining the credibility of the report. This is because she conceives of collective testimony as expressing a collective belief, rather than the beliefs of the members, and she holds that a collective belief is the sort of thing that can indeed be produced via the sort of essentially negotiative process that Beatty describes. Hence, in settling the question of what view they would let stand as their collective view, the panel thus actually makes up its collective mind.
I think this is where the problem with Tollefsen’s account lies. The problem is not that groups are generally insincere speakers, nor that they are routinely incompetent believers. Rather, the problem concerns the epistemic credentials of competent collective belief, on Tollefsen’s model of it. To see this, note first that in viewing the panel of geneticists as having formed a genuine collective belief via a process of negotiation, Tollefsen tacitly adopts what is sometimes called a “joint acceptance” account of collective belief, according to which collective belief is produced when a set of people openly let some proposition stand as their collective view. In fact, this is probably the most widely held account in the literature on collective belief, with versions of it having been defended by Margaret Gilbert, Raimo Tuomela, Frederick Schmitt, Philip Pettit and Tollefsen herself. But, whatever its independent merits, this view of collective belief cannot help to vindicate the epistemology of collective testimony, because it allows collective belief to be properly, competently held for non-epistemic reasons. And this means that the guiding principle of the BVT, that we are entitled to believe what others competently believe, does not apply at the collective level.
To see why this is so, consider what the right kind of reasons are, for individual and collective belief respectively. As far as individual belief goes, it seems clear that the only reasons which are of the right kind for individually believing that p are reasons which bear, positively, on the question of p’s truth. This means that a competent individual believer of p is one who is sufficiently sensitive to considerations that bear on p’s truth, and this, in turn, helps to explain why the fact of one person’s competently believing that p can feature as a reason of the right kind for another person to believe that p.
What, then, of collective belief? Well, when collective belief is understood as what a group of people openly let stand as their collective view, as the joint acceptance account has it, then the right kind of reasons for the collective belief that p will be any reasons that bear positively on the question of whether to let p stand as the group’s view. This means that amongst the right kinds of reasons for collectively believing that p are reasons that do not bear on the truth of p—such as reasons of politeness, of solidarity, of political advantage, of financial gain, etc. But this poses a serious problem for Tollefsen’s non-summativist account of collective testimony, because if collective belief can be properly, competently held for non-epistemic reasons, then being put in touch with a competent collective belief is not a proper epistemic reason to adopt that belief oneself.
We can illustrate this by altering once more the backstory in our case of the Norwegian Police Service:
MANUFACTURED COLLECTIVE BELIEF: Imagine that the Norwegian Police Service is facing severe cuts to its public funding, which will result in the withdrawal of key resources, and many police officers losing their jobs. The leadership of the Police Force recognise, however, that the plan to reduce public funding would not be implemented if the country is perceived as subject to any kind of concrete terror threat. So when the latest terror threat is received, the Police Force comes to an internal “understanding” that the threat is serious, credible and imminent. In other words, the (relevant, or “operative”) members agree to let the proposition “There is a serious, credible threat of an imminent terrorist attack on Norway” stand as their collective view, i.e., the view of the Norwegian Police Service. And, of course, they decide to express this view publicly, since it is only by publicising the view that they will achieve their aim of forestalling the funding cuts.
There are at least two quite different ways that this sort of scenario might be seen to pose a problem for Tollefsen’s account. One way is to see it as impugning the account of collective belief that Tollefsen tacitly incorporates—the joint acceptance account of collective belief. This is how Lackey (2020), who describes a similar case, sees its significance. According to her, the group in such a scenario is obviously lying, but a joint acceptance account of collective belief prevents us from reaching that verdict. That is because, on a joint acceptance of collective belief, the group in this scenario would actually believe what it asserts, and so would fail to meet the insincerity condition on lying.
A different approach—the one I take here—is to grant the joint acceptance account of collective belief and focus instead on the implications this has for the epistemic credentials of testimony that expresses collective belief. So we grant that the view expressed is indeed the view of the group (a group view may be formed by agreeing to let some proposition stand as the group’s view), and that view is supported by the right kind of reasons for collective belief (since the right kind of reasons for collective belief are any reasons that bear positively on the question of whether to let some view stand as the group view). This means that the group is sincere and competent, which should be sufficient, on a belief view of testimony, to make its testimony credible. But, of course, the group’s testimony in this case is not really worthy of being believed, since it is not based on sound epistemic reasons. Hence, Tollefsen’s non-summativist version of the BVCT does not fare much better than its summativist counterpart: it too fails to fulfil the desideratum of epistemological adequacy.
In short, the problem for Tollefsen is that the epistemic credentials of competent collective belief, at least on one leading account of collective belief, may be considerably weaker than those of competent individual belief. And this means that the fact of p’s being competently collectively believed does not give someone apprised of that fact any reason to herself believe that p. It doesn’t help to insist, as Tollefsen (2007, 2009) does, that groups may need to be monitored more closely than individuals for trustworthiness and credibility; the problem runs deeper than that. For even when groups are competent in belief and sincere in the way they relay their beliefs through testimony, their testimony may still not provide an audience with a reason to believe what they say. This is because the fact that p is competently collectively believed is not, in general, a reason to believe that p.
4 Lackey’s Statement View of Collective Testimony
It seems, then, that neither version of the BVCT provides a satisfactory account of the epistemology of collective testimony. The problem with the summativist version is that it is too restrictive: it forces us to discount some collective testimony that is intuitively worthy of being believed. By contrast, the problem with the non-summativist version is that it is too permissive: it provides no resources for discounting collective testimony that is intuitively unworthy of being believed. It is worth noting that aside from failing to fulfil this epistemological desideratum, these two views also fail to satisfy the other key desideratum mentioned earlier, testimonial distinctiveness. On both the summativist and non-summativist versions of the BVCT, the fact that testimony is a distinctive speech act is relegated to a somewhat incidental, or at least merely instrumental, role in the epistemological story. On these accounts, the speaker’s words function only to put us in touch with competently held belief, this belief being what shoulders the epistemic burden of collective testimony. But this, I suggested earlier, threatens to undermine the distinctiveness of testimonial knowledge.
It is considerations such as these that have led Jennifer Lackey to argue that, in both individual and collective testimony, it is not the speaker’s beliefs but her words or statements that give credible testimony its epistemic credentials:
“The process of communicating via testimony does not involve a speaker transmitting her belief to a hearer, along with the epistemic properties it possesses. Instead, a speaker offers a statement to a hearer, along with the epistemic properties it possesses, and a hearer forms the corresponding belief on the basis of understanding and accepting the statement in question. Statements are not, therefore, merely vehicles for expressing beliefs but, rather, they are the central bearers of epistemic significance themselves.” (2006, p. 93)
In line with this, Lackey (2014) develops a “deflationary account of collective testimony”—“deflationary” because it accommodates collective testimony within the Statement View of (individual) Testimony (SVT) that Lackey has given elsewhere. According to the Statement View, a hearer acquires testimonial knowledge from a speaker’s testimony not by being put in touch with a competently held belief, but by being presented with a “reliable or otherwise truth-conducive” statement. Group testimony, then, will be apt to produce knowledge (will be worthy of belief) just when the statement put forward in the name of the group is suitably reliable. There is thus no need to ascribe any doxastic attitudes or abilities to the group itself.
To illustrate her view, Lackey gives the example of Sam, who is a spokesperson for, but not a member of, the UN Population Commission. Sam’s job is to compile data submitted by various members of the commission, and then make statements reflecting the data she has compiled. Now of course, because of the division of epistemic labour within the Commission, it is possible that some of these statements that Sam makes will not be claims that are believed by any of the members of the commission. This is because, in compiling the data, Sam has the discretion to conjoin hitherto separate pieces of data or make other kinds of straightforward inferences from the data she receives. In the example, one such statement is produced when Sam connects the data concerning the historical birth rate of Latinos in the US with the latest figures. So she states that “the birth rate of Latinos in the US is on the rise”, and since this statement is pieced together from two separate sources, this is something that no member is even aware of, let alone believes.
The question is, what could give this piece of collective testimony its epistemic credentials? According to Lackey, there is no need to posit group belief or group knowledge, as the non-summativist version of the BVCT has it. Instead, we can see the epistemic credentials of this collective testimony as “determined by the reliability of the proffered statement” (Lackey 2014, p. 79), where this reliability could be further explicated in variety of different ways. For instance, we could think the statement is reliable if all the statements that are produced in this way, i.e., by people who occupy Sam’s role, as spokesperson for this group, tend to be truth-tracking. Alternatively, we could examine whether Sam herself has a good track record of coming out with true statements, or perhaps whether she is a virtuous epistemic agent. Lackey does not commit herself to any of these ways of spelling out the relevant notion of reliability. Her point is simply that neither Sam’s belief, nor the beliefs of the group’s members, nor (if there could be such a thing) the group’s own beliefs, are necessarily relevant to the epistemic credentials of this collective testimony. All that matters is that the statement itself can be relied upon.
From a purely epistemic standpoint, Lackey’s account surely cannot be faulted. If a speaker’s statements are reliable then of course they should, or at least may, be believed. So Lackey would have no trouble accounting for the case of DIVISION OF EPISTEMIC LABOUR, since in that scenario the organisational design of the Norwegian Police Service ensures that the group’s statement is indeed reliable (despite the lack of member belief). Moreover, on Lackey’s view statements that are not reliable should not be believed. So Lackey would have no trouble with the example of MANUFACTURED COLLECTIVE BELIEF either, since in that scenario the cynical production of a certain convenient collective belief would clearly undermine the reliability of the statement that expresses that belief. So it seems that Lackey’s account fulfils the desideratum of epistemological adequacy.
The problem for Lackey’s account, as I see it, lies in its failure to satisfy the second of the desiderata laid out earlier—testimonial distinctiveness. Lackey’s account recommends that we believe testimony just when we take the speaker’s words to be a reliable indication of the truth, and of course treating someone’s words in this way is indeed a perfectly sound way to attain knowledge. But note this is the same way we attain knowledge when we believe the readings on our thermometers and kitchen scales, or when we eavesdrop on the soliloquys of our mortal enemies, and so we might question whether the knowledge one attains in this way is what I earlier characterised as “testimonial knowledge”. One person’s speech can obviously be the occasion and cause of another person’s acquisition of knowledge, including knowledge of what was said. But not all such knowledge deserves to be called testimonial knowledge. Testimonial knowledge is the special sort of knowledge that an addressee of testimony gets by believing the speaker, by taking it from her that p, and it is possible to treat someone’s speech as reliable—and hence to believe what they say—without believing the speaker herself. The problem for Lackey, then, is that her account threatens to undermine the distinctiveness of testimonial knowledge, by reducing the epistemic credentials of testimony to the reliability of the speaker’s statement.
To see this, let us consider yet another version of the case of the Norwegian Police Service:
LEAKED EMAIL: Imagine once again that a terror threat is received, and is judged by the Norwegian Police Service to be credible, imminent and serious. However, in this scenario, the Police Service is required by law to consult with the Prime Minister before any information about the threat can be made public. So the head of the Police Service sends an email to the Office of the Prime Minster, stating that there is a credible, serious threat of an imminent terrorist attack. But before any decision can be made about whether the public should be informed, this email is leaked to the press, which means that news of the threat gets widely circulated.
It seems correct to say that both the Prime Minister and the public can acquire knowledge of the terror threat from the testimony contained in the email from the head of the Police Service. But intuitively the way they acquire that knowledge, and as a result, the kind of knowledge they each acquire, seems markedly different. This is because only the Prime Minister was the addressee of the Police Service’s testimony: only she was being invited to take it from them that there is a terror threat. The public, by contrast, was decidedly not meant to hear of the terror threat, and so was not invited to take it from them. So while both the public and the Prime Minister can clearly acquire knowledge of the threat from the statement of the Police Service, only the latter acquires testimonial knowledge.
The problem for Lackey is that no such distinction is available on her Statement View of Collective Testimony. Since the reliability of the Police Service’s statement is unaffected by the fact that it is later leaked, it seems that everyone who encounters the statement—both the intended and unintended recipients—can acquire testimonial knowledge from it, provided it meets some suitable standard of reliability. Hence, if we wish to preserve the distinctiveness of testimonial knowledge, and distinguish it from knowledge that is merely based on someone’s testimony, then Lackey’s account is also unsatisfactory.
Now, of course, it should be acknowledged that this objection to Lackey’s account of collective testimony is only as strong as the case that can be made for the testimonial distinctiveness desideratum—and I have not really attempted (nor shall I now attempt) to properly justify that desideratum. What this means is that anyone who rejects the testimonial distinctiveness desideratum, either in general or only in the case of collective testimony, would be unmoved by the objection. This would no doubt include Lackey herself, who clearly rejects the notions of testimony and testimonial knowledge adopted earlier that motivate the testimonial distinctiveness desideratum. According to Lackey, testimony is any statement that either is intended to convey, or can be taken as conveying, information, and testimonial knowledge is the knowledge someone can glean from such a statement. This means that, for Lackey, an eavesdropper or overhearer can attain testimonial knowledge from a speaker’s testimony just as easily and in essentially the same way as the addressee—and this suggests she would see nothing at all untoward about counting the Norwegian public’s knowledge in LEAKED EMAIL as genuinely testimonial knowledge.
I am happy to accept that this dispute with Lackey boils down to a more fundamental disagreement over the distinctiveness of testimonial knowledge that cannot be resolved in this paper. A more troubling concern, however, is the thought that while individual testimony and the testimonial knowledge it produces may well be distinctive in the manner outlined earlier, collective testimony and the testimonial knowledge it produces may not be distinctive in this same way. In particular, it might be thought that, in contrast to individual testimony, collective testimony need not involve a speaker making a tacit invitation to “take it from me”, and knowledge from collective testimony need not involve an audience accepting such an invitation.
Something along these lines has recently been argued by Katherine Hawley (2017). Hawley’s suggestion is that there is no particular reason to think that collective testimony, like individual testimony, involves a speaker inviting the audience to “take from me”. Indeed, she suggests that groups may lack the epistemic and linguistic agency needed to do this. Moreover, in contrast to the individual case, Hawley claims that there is no real cost—neither epistemological nor ethical—to viewing group testimony in purely reliabilist terms:
“We can see groups as producing statements via the functioning of various internal mechanisms: we may then consider whether those mechanisms lend themselves to the production of truth or false statements” (Hawley 2017, p.242).
In effect, Hawley is proposing a bifurcated view of the epistemology of testimony—a view that treats individual testimony as epistemologically distinctive in certain ways that collective testimony is not. Perhaps that is, ultimately, the right approach. However, before resigning ourselves to that explanatorily inelegant conclusion, I think we should first investigate the prospects for an epistemologically robust theory of collective testimony that treats collective testimony as distinctive in essentially the same way as individual testimony. In the following and final section, I tentatively suggest that Miranda Fricker’s “trust-based” view of collective testimony is a promising candidate for such a theory.
5 Fricker’s Trust View of Collective Testimony
As I see it, Fricker’s account of collective testimony lies somewhere between Tollefsen’s non-summative version of the BVCT and Lackey’s statement view. According to Fricker, it is neither the beliefs of the group nor the reliability of the group’s statement that gives group testimony its epistemic credentials. Instead, it is something in-between the two, namely the fact that, in testifying, a speaker makes a certain kind of interpersonal epistemic commitment to her audience. This is something done in uttering words but it is not equivalent to those words themselves; and, though it bears important connections to belief (which is an intrapersonal kind of epistemic commitment) it cannot be identified with belief.
Fricker’s account of the epistemology of collective testimony is rooted in her “trust based conception of the speech act of testimony” (Fricker 2012, p. 259). According to her, the act of telling (testifying) is an illocutionary act that essentially involves the speaker making a commitment of epistemic trustworthiness to her audience, with the aim of establishing a relation or “deal” of epistemic trust. What this deal of trust allows is that the audience acquires knowledge by trusting the speaker for the truth of what she says.
The anatomy of a deal of trust is simple. The speaker first perceives a certain epistemic need on the part of her interlocutor (which need not be based on an explicit request for information), and then answers that need by means of testimony. The speaker’s testimony does not simply furnish the audience with the needed information, but offers that information in a particular spirit: the speaker implicitly invites the audience to trust her with respect to what she states (“Take it from me”). In this way, testimony commits the speaker to trustworthiness with respect to that information, which means that the audience acquires the standing to blame the speaker or hold her responsible in other ways, should the testimony turn out to be false.
One of main motivations for a trust-based view of testimony such as Fricker’s is that it promises to do justice what is distinctive about testimonial knowledge. Testimonial knowledge is knowledge an audience gets by accepting a speaker’s invitation to trust her. This is, as Edward Hinchman points out, completely different from merely treating someone’s words as reliable, since “when you have evidence of a speaker’s reliability you don’t need to trust her […] You can treat her as a truth-gauge” (Hinchman 2005, p. 580). The challenge for proponents of this kind of view, however, is to say how reasons of trust could be good epistemic reasons—and hence how testimonial knowledge, understood in these terms, could ever amount to knowledge at all. Why should someone’s commitment of trustworthiness, their invitation to “take it from me”, be a sound epistemic reason to believe them in what they say?.
One part of the answer to this question is that, in making such commitments of trustworthiness, testifiers make themselves liable to normative sanctions if they fail to fulfil them—i.e., if they fail to prove trustworthy in what they say. So perhaps it is the threat of being held accountable that gets testifiers, for the most part, to speak responsibly, telling people only what they themselves know, or have good grounds for. The problem with this explanation, as it stands, is that it does not seem sufficient. The mere fact that people are accountable for abiding by certain norms does not always (or even for the most part) mean that they will abide by them—especially in contexts where their chances of being found out are low. Consider, for instance, how drivers routinely break speed limits when they know their chances of being caught are slim, or how many people illegally download copyrighted material online, knowing that they are very unlikely to be found out and prosecuted. In such contexts, people knowingly undertake accountability for their compliance with norms, but they nonetheless cannot be relied upon to comply with those norms. Why then should things be different in the case of testimony? More specifically, if irresponsible testimony is likely to pass unsanctioned, why should the mere fact that testifying involves undertaking responsibility mean that testimony can, in general, be relied upon?
It is partly in response to these concerns that Fricker stresses the distinctively second-personal dimension of the speaker’s commitment. In testifying, the speaker commits to her audience to proving trustworthy. This means that she incurs a directed epistemic obligation: she in some sense owes the truth of what she says to the audience. And, correlatively, the audience acquires a directed epistemic right: she has the standing to blame the speaker, or epistemically sanction her in other ways, if the speaker does not live up to her commitment of trustworthiness. This makes the normative situation of testimony quite different from the normative contexts considered above, in which people (drivers and internet users) are knowingly accountable in terms of an impersonal and independent normative standard (speed limits and copyright regulations). The illegal downloader does not owe it to anyone to respect copyright laws, and the speeding driver does not betray anyone with her speeding. So while these people are accountable for their actions in the sense of liable to various kinds of normative sanctions, they are not liable to the distinctively second-personal forms of holding responsible that are involved in testimony.
Fricker’s thought is that it is the threat of these deeply interpersonal reactions and sanctions that keeps testifiers in line (for the most part), ensuring that speakers tell others only what they know, or have good grounds for, on pain of being blamed or shunned. This in turn creates a default entitlement for us to believe what others tell us even when we lack an independent reason to think they are reliable. As Fricker puts it:
“the testifier’s very commitment to truthfulness is shored up by the implicit threat of a powerful kind of blame (‘You betrayed me!’). This is one of the many ways that the addressee’s holding the testifier to the trust invested is intertwined with the relevant epistemic relations: in holding testifiers to the truthfulness of their word in the manner of second-personal trust, addressees far and wide massively boost the veritistic energies that drive a well-functioning testimonial practice” (Fricker 2012, p. 269).
Fricker thinks that groups too can make commitments of second personal epistemic trustworthiness in their speech, and to support her view she appropriates Margaret Gilbert’s concept of “joint commitment”. A joint commitment, as it is defined by Gilbert, is a commitment of and by several people to doing something as a single body, where possible substitutions for “doing something” include both actions, such as walking together, and attitudes, such as believing that p, or intending to phi. So Fricker’s idea is simply that the characteristic object of a testimonial commitment, namely, proving worthy of an audience’s second personal epistemic trust is the possible object of a Gilbertian joint commitment. Just as an individual testifier makes a personal commitment of epistemic trustworthiness in her testimony, so too can a collective speaker—a “plural subject”, as Gilbert would call it—make a joint commitment of epistemic trustworthiness in its speech. And when such a commitment is met with an investment of epistemic trust on the part of the group’s audience, there will be a deal of second-personal epistemic trust on one side of which is a group: there will be a we-thou deal of epistemic trust.
To illustrate her account of collective testimony, Fricker gives the example of a government-appointed committee that is given the task of producing a report on the health risks of certain food additives—a report which will then be published so that the public can learn of the findings. Such a group would have to be constituted, Fricker thinks, by way of a joint commitment to epistemic trustworthiness with respect to what they will say about those health risks in their report (such a commitment “comes with the job”, Fricker (2012, p. 272) claims), and this joint commitment forms the basis for the invitation to trust that the report communicates to the public. This means that when a member of the public reads and believes the report, a deal of trust is sealed between the committee and this citizen, and so, were it to later transpire that certain members of the committee had been swayed by bribes from certain food companies, this would amount to a “personal betrayal” of the citizen by the committee. It is the threat of being blamed for such a betrayal, or being exposed to other epistemic sanctions or penalties, that ensures that groups such as this one tend to be epistemically responsible in their testimonial practice.
I cannot here give a complete defence of Fricker’s account of collective testimony. Instead I will conclude by briefly saying how I think it satisfies the main desiderata for a theory of collective testimony, and how it avoids the difficulties faced by the other accounts I have considered. The primary desideratum for an account of the epistemology of collective testimony is, of course, that it provides a satisfying answer to the question of what gives credible collective testimony its epistemic credentials. For Fricker, the answer is not competent belief, as it was with both the summativist and non-summativist versions of the BVCT, nor is it reliable words or statements, as it was with Lackey’s statement view. Instead, for Fricker the answer lies in the way the joint commitment a group speaker makes in testifying renders the group responsible to its audience, and vulnerable to blame and charges of betrayal if its testimony turns out to be false. The prospect of being held to account in these distinctively second-personal ways promotes good epistemic behaviour, “massively boost[ing] the veritistic energies of a well-functioning testimonial practice”, and thereby providing a kind of default, albeit defeasible, entitlement to take speakers, including group speakers, at their word.
The other desideratum for an account of testimony concern the distinctiveness of testimony as a source of knowledge: an account of testimony should connect the epistemological story on offer to the fact that testimony is a speech act, as well as to the fact that learning from testimony requires the audience to respond to that act in a certain way, by believing the speaker, rather than simply believing what the speaker says. Fricker’s account is tailor-made to fulfil this desideratum. According to her, the forging of a commitment to second personal epistemic trustworthiness is just what is done in the speech act of telling—in other words, this kind of commitment is definitive of that illocutionary act. And testimonial knowledge is acquired, on Fricker’s account, only by an addressee of testimony taking the speaker’s word on trust, i.e., by being a party to a deal of second personal epistemic trust. This is her way of fleshing out what is involved in “believing the speaker” herself, rather than simply believing what the speaker says.
How would Fricker’s account of collective testimony handle the cases that posed problems for the other views I have discussed? Consider first the case of DIVISION OF EPISTEMIC LABOUR in which no member of the Norwegian Police Service believed what they as a group stated, yet their collective testimony still seemed, intuitively, to be belief-worthy. For Fricker, what matters in this case is simply that the joint commitment the Norwegian Police Service makes to proving epistemically trustworthy vis-à-vis the terror threat be a credible commitment, and it seems that despite the lack of belief among the members, there is no reason to doubt that it is. Quite the opposite: the fact that the organisation would be roundly blamed and censured if their testimony turned out to be false, as well as the fact that the division of epistemic labour within the organisation follows from epistemically impeccable principles of institutional design, suggests that the group’s commitment is indeed credible, and hence that their testimony is belief-worthy.
In the case of MANUFACTURED COLLECTIVE BELIEF a collective belief is fabricated for non-epistemic reasons and then expressed as group testimony. Fricker’s account allows us to deny that this amounts to good or credible collective testimony. This is because the cynical motives involved in the production of collective belief would impugn the credibility of the group’s commitment to proving epistemically trustworthy, even if they do not impugn the collective belief thus produced. Should it come to light that the Norwegian Police Service is expressing a view they arrived at for non-epistemic reasons, the default entitlement to take speakers at their word would be defeated in this case. There would no longer be any reason to think their commitment to proving epistemically trustworthy with respect to the terror threat is credible, and hence this piece of collective testimony would not be worthy of belief.
In the case of LEAKED EMAIL, the Norwegian public only comes to know of the terror threat because of the leak, rather than by being the addressee of the Norwegian Police Service’s testimony. Fricker’s account allows us to admit that this counts as knowledge, while denying that it is testimonial knowledge. This is because the Norwegian Police Service made no commitment of second personal epistemic trustworthiness to the public, and so the public was not a party to the kind of “deal of trust” that makes testimonial knowledge possible. If it turns out that the information in the leaked email is false, the public would not have the standing to blame or otherwise sanction the Norwegian Police Service; they may feel disappointed or concerned about the epistemic competence of this institution, but they would not have been betrayed. So while the public may come to know of the terror threat by encountering the Norwegian Police Service’s testimony about the terror threat, the fact that this testimony was not publicly addressed means that their knowledge is not testimonial knowledge.
Fricker’s account thus not only satisfies the key desiderata for a theory of collective testimony, it also delivers the intuitively correct verdict on these three cases that posed problems for the other views considered. Like Lackey’s statement view, it manages to do this without requiring that the testifying group, or its members, believe what it tells its audience. But unlike Lackey’s view, it still requires that the testifying group itself be an epistemic agent, capable of making interpersonal epistemic commitments, and undertaking distinctively epistemic responsibilities. On this account, then, collective testimony is not a matter of the group making its internal epistemic commitments public, so much as a matter of its making public epistemic commitments—commitments for which the group’s addressees may hold it responsible.
In this paper, I have critically discussed three views of the epistemology of collective testimony and suggested that the last of these is the most promising. The Belief View of Collective Testimony holds that the epistemic credentials of collective testimony should be tied to the beliefs that testimony expresses—where these might be either the beliefs of group members (summativist version) or the beliefs of the group itself (non-summativist version). I argued that that both versions of the BVCT fail on grounds of epistemological adequacy. The summativist version fails because there are cases of credible collective testimony (such as DIVISION OF EPISTEMIC LABOUR) where no group member believes what the group testifies, and the non-summativist version fails because there are cases (such as MANUFACTURED COLLECTIVE BELIEF) in which competent collective belief is expressed through group testimony but the testimony is not credible. This suggests that we should look beyond belief—both individual and collective—in order to understand what is epistemologically salient about collective testimony. Lackey’s Statement View of Collective Testimony attempts to do exactly this by focusing not on beliefs but on statements, and specifically their reliability. But this view is also inadequate, I argued, because it fails to do justice to the distinctiveness of testimonial knowledge—the sense in which testimonial knowledge is knowledge acquired by believing the speaker or “taking it from her”.
The final view considered, Fricker’s “trust view”, seems to me the most promising. It offers a robust epistemology of collective testimony without undermining what is distinctive of testimonial knowledge. According to this view, what is epistemologically salient about collective testimony is not that it expresses competent belief, nor that it is constituted by reliable words. Rather, it is the fact that, in testifying, a group commits itself to proving worthy of its audience’s epistemic trust.
Funding source: Austrian Science Fund
Award Identifier / Grant number: I 3068-G24
For their kind and generous feedback, I am grateful to Olav Gjelsvik, Jennifer Hornsby, Deborah Tollefsen, Hans Bernhard Schmid, and Dina Townsend, as well as audiences in Copenhagen, Lund and Vienna, and several anonymous reviewers. This research was supported by the Austrian Science Fund (FWF), grant number I-3068-G24.
Anscombe, E. 1979. “What Is It to Believe Someone?” In Rationality and Religious Belief, edited by C. F. Delaney. Notre Dame: University of Notre Dame Press.Search in Google Scholar
Coady, A. 1992. Testimony: A Philosophical Study. Oxford: Oxford University Press.Search in Google Scholar
Gilbert, M. 2002. “Belief and Acceptance as Features of Groups.” In Chapter 6 in Gilbert, Margaret (2014): Joint Commitment, New York: Oxford University Press.10.1093/acprof:oso/9780199970148.003.0007Search in Google Scholar
Graham, P. 2015. “Testimony as Speech Act, Testimony as Source.” In Moral and Intellectual Virtues in Western and Chinese Philosophy: The Turn toward Virtue, edited by S. Mi and Slote. London: Routledge.Search in Google Scholar
Hawley, K. 2017. “Trustworthy Groups and Organisations.” In The Philosophy of Trust, edited by P. Faulkner and T. Simpson. Oxford: Oxford University Press.10.1093/acprof:oso/9780198732549.003.0014Search in Google Scholar
Hinchman, E. 2014. “Assurance and Warrant.” Philosophers’ Imprint 14 (17).Search in Google Scholar
Huebner, B. 2014. Macrocognition. Oxford: Oxford University Press.Search in Google Scholar
Kulik, J., and P. Carlino. 1987. “The Effect of Verbal Commitment and Treatment Choice on Medication Compliance in a Pediatric Setting.” Journal of Behavioral Medicine 10 (4): 367–76.10.1007/BF00846476Search in Google Scholar
Lackey, J. 2014. “A Deflationary Account of Group Testimony.” In Essays in Collective Epistemology, edited by Lackey. Oxford: Oxford University Press.10.1093/acprof:oso/9780199665792.003.0004Search in Google Scholar
Moran, R. 2005a. “Getting Told and Being Believed.” Philosophers’ Imprint 5: 1–29.Search in Google Scholar
Pettit, P. 2003. “Groups with Minds of Their Own.” In Socialising Metaphysics: The Nature of Social Reality, edited by F. Schmitt. London: Rowman & Littlefield.Search in Google Scholar
Pettit, P. 2008. Made with Words. New Jersey: Princeton University Press.Search in Google Scholar
Schmitt, F. 1994. “The Justification of Group Beliefs.” In Socializing Epistemology: The Social Dimensions of Knowledge, edited by Schmitt, 257–87. Lanham, MD: Rowman and Littlefield.Search in Google Scholar
Schultz, W., S. Oskamp, and T. Mainieri. 1995. “Who Recycles and When? A Review of Personal and Situational Factors.” Journal of Environmental Psychology 15 (2): 105–21.10.1016/0272-4944(95)90019-5Search in Google Scholar
Tollefsen, D. 2015. Groups as Agents. Cambridge: Polity Press.Search in Google Scholar
© 2020 Leo Townsend, published by De Gruyter, Berlin/Boston
This work is licensed under the Creative Commons Attribution 4.0 International License.