Access denied: epistemic obstruction and the distribution of knowledge

John Greco’s The Transmission of Knowledge (2021) argues that sharing knowledge is like passing a ball in sports: it requires teamwork and a hospitable environment. On Greco’s view, for a speaker to successfully transmit knowledge to a hearer, the speaker and hearer must act jointly to achieve the transfer and must trust each other (2021: 57). In acting jointly, the speaker and hearer must also occupy a reliable information channel, which in turn requires that they be in a social and epistemic environment whose norms underwrite such channels (2021: Ch. 4).

Greco’s innovative account aims to solve a problem for achievement theories in virtue epistemology, such as his own (Greco 2010), which argue that knowledge must be an achievement of the agent who has it, in the sense that the agent’s arriving at a true belief must be due to her reliable competences (and not to luck). One problem for achievement theories is that in making the hearer’s knowledge an achievement of her competences, rather than the competences of speakers or other social factors, they make knowledge overly individualistic and neglect its social dimensions. Case in point: Jennifer Lackey’s well-known objection that achievement theories in virtue epistemology overlook speakers’ contributions to testimonial knowledge (2007). Greco’s solution is to ground knowledge transmission in the competency of the joint agent that is formed by the speaker and hearer when they work together to share knowledge in a hospitable social environment (2021: 98). Greco thus makes knowledge transmission via testimony a joint achievement of the speaker and hearer, rather than an individual achievement of the hearer. In other words, Greco defends his virtue-theoretic framework by expanding it to include the competences and achievements of joint agents.

The key upshot of Greco’s argument is that knowledge transmission is irreducibly social. It requires the cooperative joint action of a speaker and hearer, who are working together, within a hospitable social environment that underwrites their reliability. Greco correspondingly argues that knowledge transmission is a distinct phenomenon from knowledge generation. On his view, knowledge transmission cannot be reduced to back-to-back instances of knowledge generation in the individual speaker and hearer. Nor can knowledge generation be reduced to knowledge transmission—since the former does not require joint agency and can be an individual achievement, e.g., in cases of knowledge from one’s own perception, one’s own deductive or inductive inferences, etc. Accordingly, Greco proposes a picture in which there are two different ways of coming to know, either by generation, or by transmission, neither of which is reducible to the other, and both of which can be explained by his expanded virtue-theoretic framework (2021: Ch. 5). Greco subsequently argues that his framework is capacious enough to also explain a third way of coming to know that captures knowledge of hinge propositions (2021: Ch. 6).

In arguing that knowledge transmission is irreducibly social, Greco (2021) makes a crucial contribution to debates in social virtue epistemology. More broadly, in shifting the spotlight from knowledge generation to knowledge transmission, he illuminates a phenomenon that is often overlooked by more traditional forms of epistemology, and makes a compelling case for the study of knowledge transmission in its own right. Along the way, he explores conditions for the transmission of religious knowledge, argues that the epistemic good of understanding can be transmitted through education, and iconoclastically contends that not all knowledge from testimony is transmitted from the speaker to the hearer—some is generated by the hearer. Below, I focus on two of the key features of Greco’s account: joint action, and reliable information channels.

Greco intends his account of knowledge transmission to explain cases in which things go well and cases in which they go badly. When they go well, knowledge is transmitted because agents have trusted each other and worked together in an environment that underwrites the reliability of the channel they occupy. When things go badly, transmission fails because trust, cooperation, and reliability are absent or have been eroded (2021: 183). In developing his account, Greco understandably focuses on cases in which things go well—in which agents do trust each other and cooperate, and do occupy reliable information channels. He likewise avers that it is a matter of luck as to whether we end up in an environment whose channels of communication are reliable (2021: 144, 170). In the below, I focus on cases in which we are less lucky and things are not going very well. Familiar cases in which powerful speakers deliberately obstruct the flow of information, and exclude members of socially marginalized groups from the epistemic community. Cases in which hearers fail to recognize scientific experts for the competent and sincere epistemic agents that they are. And, more broadly, cases in which some of the communication channels in our social and epistemic environments are rooted in unreliable identity prejudices and constrained by epistemic silos and filter bubbles. In short, I ask whether focusing on cases in which things are going rather badly might point us toward a somewhat different picture of knowledge transmission: one that doesn’t require joint action or trust but still requires reliable information channels.

Below, I begin to sketch such a picture. I suggest that knowledge distribution can occur even when joint action and trust are noticeably absent. For it to occur, the environment must allow for reliable information channels in the first place, and the epistemic obstructions blocking those channels must be bypassed or dissolved. To illustrate, it can occur when journalists and lawyers gain access to the private information channels of speakers and their co-conspirators, and in so doing, gain access to the knowledge in those channels—e.g., knowledge that members of the clergy had committed sexual assault, or knowledge that smoking low tar cigarettes is still harmful. Likewise, it can occur when members of socially marginalized groups, who have been excluded from the institution of formal education, gain access to the information channels used in schools and to the knowledge—e.g., of history, science, and literature—in those channels. More tendentiously, I will also suggest it would occur if scientists and tech executives were able to bypass the extant echo chambers of anti-vaxxers and ‘nudge’ them toward knowledge of vaccine effectiveness—perhaps, by removing conspiracy theories from platforms.

In short, I will be arguing that Greco’s analysis of knowledge transmission is too strong. The transmission of knowledge from speakers to hearers does not require speakers to intend to share knowledge with the hearers in question—indeed, speakers might intend to deny access to those would-be hearers. Nor does it require hearers to share such an intention—anti-vaxxers might deny that Dr. Fauci has any knowledge to share (the damage of their echo chamber might have already been done). Nor does it require joint action or trust on the part of speakers and hearers.

But, Greco is right that it does require reliable information channels. Things can’t go so badly wrong that the social environment only engenders unreliable channels. If the channels of co-conspirators, of exclusionary institutions of education, and of vaccine-science weren’t reliable, then they wouldn’t be trading in the epistemic good of knowledge in the first place. Additionally, the bypasses themselves—constructed by journalists, socially marginalized hearers, and the scientific and tech community—must be reliable. If those bypasses aren’t reliable, then even when they succeed in distributing truths, they won’t succeed in distributing knowledge. In other words, to distribute knowledge, the bypasses in question must first tap into channels that are reliable, and must also be reliable channels themselves.

Greco is also right that the distribution of knowledge is like the distribution of material goods in an economic framework (2021: 18). Indeed, on the picture I sketch, knowledge distribution is much more like the distribution of material goods in an economy than it is like the passing of a ball in sports. Even if passing a ball—as opposed to intercepting it—requires joint action, material goods can be reliably intercepted rather than passed, and so can epistemic goods like knowledge.

In sum, while I will object to joint action and trust as necessary conditions for knowledge transmission, I will simultaneously preserve Greco’s insight that reliable information channels are needed. Accordingly, the bulk of the below will point toward a somewhat different picture of knowledge transmission, one that is still inspired by features of Greco’s account but also allows for the distribution of knowledge in conditions where trust and joint action are not in the offing. The final section canvasses some objections to, and implications of, such a picture, including its implications for the sociality of knowledge transmission and for a virtue epistemology of knowledge transmission that grounds it in competences. The opening section, to which we now turn, examines some of the details of Greco’s view.

1 Greco’s account of knowledge transmission

Greco argues that knowledge won’t be transmitted unless a speaker and a hearer trust each other and act jointly, and in so doing, occupy a reliable information channel that is engendered by norms in their environment. Let’s unpack the central features of this account, beginning with joint action.

On Greco’s view, knowledge transmission requires a particular kind of speech act, successful telling, which in turn requires joint action by the speaker and hearer. In his words (2021: 57),

(KT) Knowledge that p is transmitted from a speaker S to a hearer H just in case S successfully tells H that p. And that happens just in case: (1) S knows that p; (2) S asserts that p with the intention of sharing knowledge that p with H; (3) H understands and shares S’s intention; and (4) S and H act jointly so as to bring about their shared intention (i.e., so as to “consummate” the speech act in condition 2).

To explain, Greco follows Elizabeth Fricker (2006) and Richard Moran (2018) in arguing that successful telling is a kind of assertion that requires speaker-intention and hearer-uptake.¹ He emphasizes that on its own, speaker-intention won’t be enough (2021: 60). Thus, I won’t count as successfully telling my students that their paper deadline has been postponed unless: I assert this with the intention of sharing my knowledge and there is uptake on the part of my students.

Importantly, Greco argues that hearer-uptake requires joint action by the speaker and hearer together, rather than action by the hearer alone. In building a conception of joint action, he draws on key features in the accounts of Margaret Gilbert (2014) and Michael Bratman (2014). For Greco, joint action characteristically involves a shared intention to act together, shared knowledge of this intention, and a shared sub-plan for working together to carry out the action. Like Gilbert and Bratman, he also thinks joint action is usually interactive and interdependent (Greco 2021: 56).

What does all of this mean for cases of successful telling? Roughly, it means that in a typical case of successfully telling my students that their deadline has been postponed, I intend to act together with them to share this knowledge. My students likewise intend to act together with me to bring about the sharing of knowledge, and we all understand that this is what we intend. Further, it means that in the typical case, we have a mutual sub-plan for sharing such knowledge—e.g., we may share the sub-plan of posting and reading (respectively) weekly updates on the course website. Finally, it means that we will be responsive to one another and will coordinate our individual actions in our effort to share knowledge—e.g., when I forget to post a weekly update, my students will send me inquiries, to which I will respond by posting a belated update, which they will read. Since this case is one of successful telling, our efforts at coordination will also succeed in sharing knowledge.

Greco avers that while all of the above features are characteristic of joint action, they may not all be necessary. Even so, we might suspect that Greco’s account makes knowledge transmission too hard to attain, putting it out of reach for young children who fail to satisfy many of the conditions. That is a potential problem for Greco, who argues against the reductionist view of testimonial knowledge on somewhat similar grounds (2021: 29–30), and who intends his account to include cases in which parents and teachers transmit knowledge to young children (2021: 36). To be clear, the concern is that Greco’s account removes one epistemic burden on hearers only to replace it with another. It removes the burden of generating inductive evidence for the reliability of speakers, but replaces it with the burden of satisfying the conditions of joint action. Young children may not be able to satisfy those conditions: they may not understand (or know or believe) that speakers intend to cooperate with them to share knowledge; or they may be unable to form such sophisticated intentions themselves. Along similar lines, Deborah Tollefsen (2022) argues that infants younger than 3 years of age, who have yet to develop a theory of mind, are unlikely to understand the intentions of speakers, much less share in any sub-plans or play a substantial role in the coordination of actions. I address further concerns about the strength of Greco’s account below.

The second central feature of his account is trust. Borrowing from Karen Jones’s (1996) work, Greco argues that trust characteristically involves relying on the person who is trusted and expecting them to be dependable (Greco 2021: 65). To briefly illustrate, when my students and I trust each other in sharing knowledge, they behaviorally rely on me to do my part—to post weekly updates, to respond to their inquiries, etc.—and I rely on them to do their part—to check for and read weekly updates, to send inquiries if I fail to post, etc. They don’t do my part for me and I don’t do their part for them. Moreover, when we trust each other, they also expect me to be dependable—they expect me to actually do what they are relying on me to do, to actually make weekly posts, etc.—and I likewise expect them to be dependable—I expect them to actually check for posts and read them, etc.² Greco thinks these features of trust are already entailed by the joint action of sharing knowledge. Accordingly, he argues that since “knowledge transmission essentially involves joint agency, and joint agency essentially involves trust,” knowledge transmission essentially involves trust (2021: 64–65).

Relatedly, it is worth noting that on Greco’s account, successful knowledge transmission can only occur within epistemic communities of cooperating agents. Greco conceives of an epistemic community as “a group of cognitive agents engaged in shared information-dependent tasks, and sharing norms for evaluating information associated with those tasks” (2021: 25). He argues that the two main tasks of epistemic communities—acquiring knowledge and distributing knowledge—should be governed by different norms. The norms governing knowledge acquisition should be stringent and play the role of quality-control, so as to only let high-quality information into the community; whereas the norms governing knowledge distribution should make it easy to move that high-quality information around once it is in the community (2021: 39). Greco argues that the norms of distribution apply when agents are cooperative and share a community—in which case, it can be appropriate for hearers to trust speakers and believe what they say, without any need to generate corroborating evidence of their own or generate knowledge for themselves (2021: 36). In contrast, the norms of acquisition apply when agents are uncooperative and don’t share a community—in which case, hearers will need to generate knowledge for themselves, and thus generate corroborating evidence of their own, and shouldn’t just trust speakers or believe what they say (2021: 32). So, when my students and I are cooperating to share information within an epistemic community, the norms of distribution apply, and it is presumably appropriate for my students to believe what I say about the postponed deadline without any need to generate corroborating evidence of their own. Whereas, when a suspect tells a police detective that he wasn’t at the scene of a crime, the suspect and the detective “are not members of a ‘community of knowers’ and are not ‘information sharers,’ at least in the present context,” and the norms of acquisition apply (2021: 42). Accordingly, the detective shouldn’t just believe what the suspect says, and does need to generate his own corroborating evidence and his own knowledge. In Greco’s words, “the norms that govern testimonial exchanges in cooperative and non-cooperative contexts are for principled reasons different…the kind of trust that is appropriate in the context of information distribution (and so for knowledge transmission) is inappropriate in contexts of information acquisition” (2021: 59–60). For present purposes, Greco’s point is that knowledge can’t be transmitted from speakers to hearers when agents are uncooperative or don’t share an epistemic community; in such cases, knowledge must be generated by hearers. Contra Greco, I will suggest below that agents need not cooperate or share an epistemic community for knowledge to be distributed. What is needed is access to reliable information channels.

This brings us to the third central feature of Greco’s account: in acting jointly, the speaker and hearer must occupy a reliable information channel. On Greco’s view, knowledge “must in fact be reliably formed,” and so the speaker and hearer won’t succeed in transmitting knowledge if their testimonial exchange isn’t reliable (2021: 70). Importantly, Greco argues that the reliability of their testimonial exchange doesn’t solely depend on the speaker and hearer themselves. It also depends on their social environment, and more specifically, on “the social norms that structure the social environment” (2021: 78).³ For Greco, social norms are both prescriptive and descriptive—they tell us what we should do and what we usually do. Further, they are descriptive because they are prescriptive; their prescriptive force leads to their internalization (Greco, 2021: 72; Graham, 2015). Accordingly, social norms such as ‘Listen to your mother’, ‘Listen to your teachers’, and ‘Don’t talk to strangers’ will influence who gets to speak and who doesn’t, who gets listened to and who doesn’t, and who we do and don’t believe. In Greco’s insightful words, “social norms structure the social environment so as to determine the flow of information from mind to mind. The result is an environment that is contoured to include various ‘information channels’—information pathways that both enable and constrain the flow of information” (2021: 76–77). Obviously, social norms need not contour the environment with reliable channels.

Greco argues that when we are lucky and things go well, social norms will contour the environment with reliable channels. We will have the good luck of being born into an environment whose social norms direct us to listen to and believe people who are reliable, and avoid believing people who aren’t (2021: 170). But, when we are unlucky and things go badly, social norms will contour the environment with unreliable channels. We may have the bad luck of being born into an environment whose social norms are saturated with unreliable identity prejudices that direct us to discredit, disregard, and exclude members of socially marginalized groups, and to seek out, listen to, and believe members of socially privileged groups (M. Fricker, 2007; Medina, 2013). Relatedly, we may have the bad luck of being born into an environment whose social norms direct us to avoid talking to ‘strangers’ outside our own bubbles and only talk to people we ‘know’ who are ‘like us’. In short, Greco helpfully argues that since successful knowledge transmission requires reliability, and since it is a matter of luck as to whether we occupy an environment whose norms engender reliable channels, it will also be a matter of luck as to whether our efforts at knowledge transmission are successful.

Now, on Greco’s view, we are lucky enough to be in an environment where at least some social norms engender reliable channels. Whether we are also lucky enough to occupy those reliable channels will depend on our specific social locations with respect to those norms. For Greco, an agent’s social location is determined by a combination of her social and institutional roles in the environment, as well as her personal relationships (2021: 181). Some social locations put agents into reliable channels, whereas others prevent agents from participating in reliable channels, in some cases by putting them into unreliable ones. The upshot is that whether knowledge transmission is successful will partly depend “on the ‘good luck’ of being in a reliable transmission channel,” which in turn depends on luck in one’s social location with respect to norms (2021: 144).

To elaborate, Greco argues that we are sometimes lucky enough to participate in reliable testimonial exchanges, and that the reliability of these exchanges is partly explained by social norms that have contoured our environment. Which social norms are these? To illustrate, Greco explains the reliability of testimonial exchanges between parents and young children, and teachers and primary school students, by invoking social norms that they have internalized. Young children have internalized norms to ask their parents and teachers questions, to listen to their answers, and to believe what they say. Parents and teachers have likewise internalized norms to tell children the truth. Greco points out that parents can be socially sanctioned for failing to tell children the truth (what would the neighbors think if they found out you lied to your children about whether there was milk in the fridge?). In the context of formal education, where such norms have been institutionalized and teachers can be professionally sanctioned for failing to tell students the truth, there is even further incentive to internalize the norm. Something similar holds for doctors and lawyers, since they are incentivized by professional standards and the law to tell their patients and clients the truth (2021: 79; 139). In sum, Greco argues that in all of these testimonial exchanges (and more), speakers and hearers occupy reliable channels that have been carved into our environment by social norms.

He likewise adds that in internalizing these social norms, speakers and hearers gain a dispositional ‘sensitivity’ to which channels are reliable, and which are not. On Greco’s view, and the views of anti-reductionists more generally, some such sensitivity is required for knowledge; i.e., knowledge must not only be reliably formed, “the knower must have some sense that her knowledge is reliably formed” (2021: 70, my emphasis). For Greco, when we internalize social norms that engender reliable channels, we become behaviorally disposed to seek out, listen to, and believe reliable speakers, and avoid, discredit, and disbelieve unreliable ones. As he puts it, “when things go well” and our behavioral dispositions are well-formed, “they reliably put us on to reliable testimonial exchanges. And when that happens, our well-formed dispositions manifest a kind of sensitivity to reliability” (2021: 82). Greco is at pains to argue that this sensitivity is tacit, and does not require explicit belief about the reliability of a source or any corroborating evidence to that effect. Rather, it is a behavioral ‘encoding’ of which sources and testimonial exchanges are reliable, and which are not. I will return to this feature of Greco’s account in the discussion of epistemic engineering below.

At this point, we might wonder whether knowledge transmission requires all of the machinery Greco has introduced. On his picture, information channels in our social environment are already underwriting and explaining the reliability of our testimonial exchanges (and our sensitivity to that reliability). Accordingly, it isn’t clear why joint action and trust would also be needed for knowledge transmission, or whether anything of central importance would be missing from the account if they were absent. If the conditions of joint action prove to be out of reach for young children, this worry will be even more pressing. Below, I suggest that knowledge transmission does not require joint action or trust. But, following Greco’s lead, it does require access to reliable information channels.

2 Access denied: knowledge distribution without joint action

The aim of this section is to mount a case for knowledge distribution without joint action. I begin with cases where speakers are obstructing the flow of information to would-be hearers, followed by cases where would-be hearers are obstructing the flow of information from speakers. All of these are cases in which joint action and trust are absent. Nevertheless, they are cases in which knowledge can still be distributed, or so I will suggest. For knowledge to be distributed in cases like these, the epistemic obstructions in question must be bypassed or dissolved. The bypasses themselves must be reliable, and the channels they gain access to must also be reliable.

There are at least two kinds of cases in which speakers are obstructing the flow of information to would-be hearers. In those that come most easily to mind, speakers inappropriately hide information from one group of would-be hearers in the community, while privately communicating that information to their own allies. Highly publicized cases include, e.g., leaders of the Catholic Church concealing information about sexual assault from their parishioners, the media, law enforcement, and the public while sharing this information amongst themselves on private channels, tobacco executives concealing information about the harm of smoking cigarettes from the public while sharing this information internally, and Donald Trump concealing information about his debt to Michael Cohen from the U.S. government, while communicating privately about this debt with members of the Trump family and with Cohen (the latter of whom, recall, pled guilty to violating campaign finance law for his payments to Stormy Daniels).⁴

In the second kind of case, speakers fail to see socially marginalized persons as potential sources of knowledge, or as potential recipients of knowledge, or as having any epistemic agency at all. As a result, they exclude socially marginalized persons from the epistemic community, and inter alia from its institutions of formal education. As Miranda Fricker (2007) might put it, these are epistemically unjust speakers who systematically assign marginalized persons a severe credibility deficit that results in their exclusion from the epistemic community. Cases are numerous and include, e.g., the historical exclusion of persons with disabilities from formal schooling in the U.S. and elsewhere, the exclusion of Black persons from schools in Southern states in the U.S. until (at least) the 1950’s, and the ongoing exclusion of girls from formal education in Taliban-controlled Afghanistan.⁵

There are two points to note. First, Greco and I are thinking of membership in an epistemic community in somewhat different ways. If the above is correct, speakers can deceive and hide information from hearers who are members of their own epistemic communities—i.e., hearers whom they consistently take to be competent sources and recipients of knowledge, and to be epistemic agents more generally. Indeed, speakers may be hiding information from these particular hearers precisely because they take them to be competent members of the epistemic community. Whereas, Greco seems to think that deception and other forms of obstruction shift speakers and hearers into a context in which they are no longer members of the same community. This commits him to claiming that knowledge cannot be transmitted from one epistemic community to another. Second, I acknowledge that the speakers in question may be acting jointly with their own allies, and within their own institutions of formal education, in ways that satisfy all of the conditions for Greco’s account of knowledge transmission. I am not objecting to any of that, nor am I here denying that those conditions are sufficient for the successful transmission of knowledge.

That said, it should be clear that the speakers and would-be hearers above do not satisfy the conditions for joint action. The speakers in question do not intend to share their knowledge with, respectively, their parishioners, the U.S. government, and persons with disabilities. Quite the contrary, they intend to conceal it from some of these would-be hearers, while failing to register that others are candidates for knowledge-sharing in the first place. Indeed, the speakers in question don’t even tell the would-be hearers above what they know.

And, yet, knowledge does get distributed when these obstructions are bypassed. Knowledge of the details and scope of the sexual abuse and rape committed by priests, and of the role of Catholic leaders in concealing that information, was distributed to journalists (at The Boston Globe and elsewhere), who gained access to thousands of the Church’s internal communications and to previously sealed court records. Knowledge of a variety of topics, including oration, was distributed to Frederick Douglass who gained access to books and teachers during his enslavement, at enormous risk to himself and despite laws against educating slaves (Douglass, 1999/1845). Knowledge of a range of subjects is likewise being distributed to the presumably small subset of girls in Afghanistan who have managed to gain access to books and teachers, and who are studying in secret, despite the Taliban’s most recent (2021) prohibition on educating girls.

Here, too, there are two points to note. First, astute readers will notice that in some real-world cases of exclusion from education, the speakers who are doing the excluding (e.g., the Taliban) will not have given their stamp of approval to the books and teachers to which the would-be hearers in question (e.g., the girls) gain access and from which they get knowledge. Much less will those speakers (e.g., the Taliban) have authored the books themselves. Accordingly, such examples are open to the objection that the speakers and teachers who actually designed the curricula and authored the books did intend to share knowledge with, rather than exclude, the would-be hearers in question (e.g., the girls), and thus that joint action (albeit at a distance) is preserved. I ask these readers to focus on cases where it is indeed the very same speakers who are both (i) excluding the would-be hearers in question and (ii) authoring the books and theories to which those hearers gain access and from which they get knowledge. In this vein, the exclusion of persons with disabilities from education is illustrative.

Second, although Greco addresses the role of the Catholic Church in concealing information about sexual abuse, he has different purposes in mind. He uses this example to argue that the Church’s arrogance and incompetence led to an erosion of its moral authority and of the trust of its members, which in turn prevented the Church from communicating religious knowledge.⁶ In his words, “effective channels of testimony were eroded or destroyed, because members of the Church became less trusting of the institution and its authorities, or simply opted out altogether” (2021: 182 − 83). In short, Greco uses this case to help explain why knowledge transmission fails when it does. Whereas, I am using this case to argue that in bypassing the Church’s obstruction and accessing its private channels of communication, knowledge about the details and scope of the assaults was distributed to journalists, and that the case thus constitutes a counterexample to Greco’s account.

If, as I am claiming, knowledge gets distributed in these cases and its distribution doesn’t require the sort of cooperation Greco envisions, then what does it require? For starters, would-be hearers will need to occupy a bypass channel; i.e., a channel that circumvents or dissolves the obstruction in question and gives them access to the private information channels that had been blocked. Now, in the cases above, would-be hearers had to construct new bypass channels. This involved considerable effort on their part, since they first had to figure out how to bypass the obstructions in question, and then needed to forge channels that did so. We can assume that Frederick Douglass, Spotlight journalists at the Globe, and Afghani girls who are studying in secret all called on epistemic character virtues, such as perseverance, courage, open-mindedness, and creativity, in these endeavors.⁷ They also all worked together with their own allies—e.g., editors and lawyers in the case of the Spotlight journalists, Sophia Auld in the case of Douglass, and teachers and NGO’s in the case of the Afghani girls. Does occupying a bypass channel always take this much work? Perhaps not, since would-be hearers won’t always need to build new bypass channels, and can sometimes avail themselves of bypass channels that have already been built, e.g., a handful of States in the U.S. now have mandatory reporting laws that explicitly apply to pastoral communications of the clergy.⁸ Accordingly, even if epistemic character virtues and joint actions with allies are required for building new bypass channels, they might not be required for occupying bypass channels that already exist. The upshot is that we can’t just assume that epistemic character virtues and joint actions with allies are needed to occupy bypass channels. They might be required to occupy bypass channels, and thus might be required for knowledge distribution in these cases, but we would need an argument to that effect.⁹

Thus far, I have suggested that knowledge distribution in these cases requires would-be hearers to occupy bypass channels. What else does it require? Following Greco, and a long tradition in epistemology that is sympathetic to externalism, I will assume the conditions on reliability also need to remain. Both the bypass channels themselves and the private information channels to which they gain access will need to be reliable in order for knowledge to be distributed from those private channels to would-be hearers. Consider again the case involving Trump, members of his family, Michael Cohen, and Stormy Daniels. For knowledge to be distributed to would-be hearers, the private communications of this ignoble group must first be reliable. If their private communications aren’t reliable to begin with, then even if they are communicating truths to one another, they aren’t trading in knowledge, and so there is no knowledge to be distributed. Further, even if their private communications were reliable and knowledge was being communicated on their internal channels, for that knowledge to be distributed to would-be hearers, the bypasses that access those communications must also be reliable. If the bypass channel consists in a single notoriously dishonest source, who just happens to be telling the truth on this occasion, then the bypass channel itself is unreliable (more on this in the final section) and the knowledge that is in the private channel does not get distributed to would-be hearers. At best, what is distributed are truths. In short, for channels to trade in and move knowledge they must at least be reliable: private channels must be reliable, and bypass channels that gain access to them must be reliable.

In the cases above, we can assume that the private channels and the bypasses themselves are reliable. Frederick Douglass and the Afghani girls gain access to (at least some) reliable books, and the journalists gain access to (at least some) reliable communications among Church leaders.¹⁰ We can likewise assume that the bypass channels in these cases are reliable; i.e., that the teachers to which Douglass and the Afghani girls gain access are competent and sincere, and that the courts and other sources are competent and sincere in conveying documents to the Spotlight journalists.

Now, we might wonder whether the would-be hearers in these cases are merely generating knowledge for themselves. Greco argues that in some testimonial exchanges, knowledge is transmitted from speakers to hearers, but in other testimonial exchanges, knowledge is generated by hearers themselves—e.g., recall the police detective who shouldn’t just believe what the suspect says and needs to generate his own corroborating evidence and his own knowledge. Are the cases above ones in which knowledge is generated by would-be hearers rather than distributed to them?

That depends on what it takes to generate one’s own knowledge. For Greco, in generating one’s own knowledge, one comes to know for oneself, and by one’s own efforts and resources, rather than from someone else (2021: 1, 5, 32, 38). A key thesis of The Transmission of Knowledge is that knowledge generation and transmission are distinct: “In cases of knowledge transmission, the hearer comes to know, but not by coming to know ‘for herself.’ In such cases, rather, it seems that knowledge is…transferred from the speaker to the hearer. And this is opposed to the hearer having to shoulder the usual epistemic burdens associated with coming to know” (2021: 35). Accordingly, Greco seems to think that a hearer in a testimonial exchange generates her own knowledge whenever she incurs the usual evidential burdens in coming to know; i.e., whenever her own corroborating evidence is carrying the justificatory load. Greco doesn’t seem to draw distinctions among different kinds of corroborating evidence a hearer might employ. But, arguably, distinctions among different kinds of evidence will have some bearing on which account of knowledge generation we adopt.

Arguably, there is more than one kind of corroborating evidence a hearer might have—(A) a hearer might have corroborating evidence for a speaker’s reliability but otherwise lack corroborating evidence for any of the specific claims the speaker asserts, or (B) she might have corroborating evidence for the specific claims the speaker asserts but otherwise lack any corroborating evidence for speaker-reliability. To put this differently, the hearer with (A) won’t have any ‘direct’ evidence of her own for the specific claims the speaker asserts but she will have ‘indirect’ evidence for those claims insofar as she has evidence for the speaker’s reliability.¹¹ For instance, she won’t have conducted any vaccine trials herself and thus won’t have any ‘direct’ evidence of her own for Dr. Fauci’s claim that the Moderna vaccine is over 90% effective, but she will have ‘indirect’ evidence for that claim insofar as she has evidence that Fauci is an expert in epidemiology. Whereas, the hearer with (B) will have ‘direct’ evidence of her own that corroborates specific claims the speaker asserts but won’t (otherwise) have any evidence of speaker reliability. As when she has ‘direct’ perceptual evidence of her own that Russia is amassing troops on its Ukrainian border, thus corroborating the claims of a notoriously unreliable propaganda machine for the Russian government. Below I suggest that (A) isn’t enough for generating knowledge oneself.

To see why, suppose (A) were enough. Suppose a hearer counts as generating her own knowledge whenever she employs corroborating evidence for a speaker’s reliability, and thus whenever she employs (A). The would-be hearers above will then count as having generated their own knowledge of a wide range of claims. After all, we can assume that the would-be hearers above have corroborating evidence for the reliability of the speakers in question—e.g., Frederick Douglass knows that he is being denied access to reliable channels—and that if they didn’t have such evidence, they wouldn’t be trying to access the private channels of those speakers in the first place. The problem is that employing corroborating evidence for a speaker’s reliability doesn’t seem sufficient for generating one’s own knowledge of the specific claims the speaker is asserting when one lacks ‘direct’ evidence for any of those specific claims. For instance, employing one’s evidence of speaker-reliability doesn’t seem sufficient for generating, e.g., one’s own knowledge that a specific priest (say, John Geoghan) had been moved from one specific diocese to another, or one’s own knowledge of the details of Cicero’s speeches (in the case of Frederick Douglass), or one’s own knowledge of the details of our planetary system (in the case of the Afghani girls). Generating one’s own knowledge of these claims seems to require ‘direct’ evidence for the claims in question. In other words, in these cases, generating one’s own knowledge that p arguably requires ‘direct’ evidence of one’s own that p; ‘indirect’ evidence for p, in the form of evidence for speaker-reliability, won’t be enough.

If these suggestions about knowledge generation are correct, then in the cases above knowledge isn’t being generated by our would-be hearers, it is being distributed to them. Now, I grant that our would-be hearers generate some of their own knowledge in these cases. For instance, they generate knowledge about how to bypass obstructions, either by themselves or with allies, and by calling on a number of epistemic virtues, including character virtues. They even manage to generate some of the knowledge that appears in private channels. Frederick Douglass taught himself some of the knowledge and skills he would have learned from private channels. And, the Spotlight journalists generated some of their own knowledge of the details and scope of the sexual assaults, by interviewing victims and priests and figuring out via induction and perseverance that the Church was putting priests who had committed sexual assault ‘on leave’. But, if my contentions are correct, when the journalists gained access to the private communications of Church leaders, they learned a whole lot more.¹² Knowledge that they didn’t already have was distributed to them from that private channel, once they gained access to it. That is the key point here. Our would-be hearers didn’t generate all of that knowledge themselves—they gained much of it from the channels they accessed.

In short, my suggestion is that hearers don’t count as generating knowledge that p themselves, if they don’t have any of their own ‘direct’ evidence for p. Their having evidence for speaker-reliability or expertise isn’t enough. Accordingly, hearers can know that speakers are experts and believe what they say on those grounds, but fail to count as generating their own knowledge of the things the experts say.¹³ In these cases and those above, I submit that knowledge is distributed to hearers rather than generated by them.

That said, astute objectors might still insist that the hearer is generating her own knowledge that p.¹⁴ After all, they might point out, the justifying ground for the hearer’s belief that p is the hearer’s own evidence that the speaker is reliable. And, since the justificatory load is borne by the hearer’s own evidence, the hearer still counts as generating her own knowledge that p even though she doesn’t have any direct evidence of her own for p.

In response, I mount an argument for successful knowledge distribution in which hearer evidence for speaker-reliability is also absent. Consider cases of epistemic engineering or ‘nudging’ in which neither joint action nor hearer evidence for speaker-reliability are present—cases in which hearers distrust speakers and even think they are unreliable.¹⁵ Anti-vaxxers in an echo chamber come readily to mind. They distrust vaccine-scientists and dismiss their reliability, fail to believe that vaccines are effective or safe, and occupy echo chambers in which such distrust and doubt are reflected and magnified.¹⁶ Assume that we are restricting our example to extant anti-vaxxers, in whom this damage has already been done.¹⁷ Note that these are also cases of denied access and epistemic obstruction, but it is now would-be hearers (and echo chambers), rather than speakers, who are obstructing the flow of information and denying (themselves) access to reliable channels. Speakers aren’t causing the obstruction, hearers and structures are—vaccine-scientists are trying to communicate openly about the effectiveness and safety of Covid-19 vaccines, and are being obstructed by echo chambers and the would-be hearers in them. Accordingly, it should be clear that anti-vaxxers and vaccine-scientists do not satisfy the conditions for joint action. Nor will anti-vaxxers have evidence of speaker-reliability available to them (the echo chamber has done its work), much less use it in their reasoning.

And yet, I suggest that knowledge would still get distributed to anti-vaxxers if these obstructions were bypassed. Suppose that vaccine-scientists and tech executives were somehow able to dissolve echo chambers and circumvent hearer distrust, perhaps by capitalizing on repetition and fluency effects to facilitate true beliefs about vaccine safety. To briefly explain, it has been suggested that familiarity with a claim increases one’s fluency in processing it, which in turn, makes one more likely to believe it (Begg et al., 1992). Accordingly, it has likewise been suggested that flooding social media sites with knowledge about vaccine safety, and changing content algorithms so that this knowledge repeatedly reaches anti-vaxxers, could gradually lead them to ‘change their minds’ (Battaly, 2021; McIntyre, 2018 and 2019).¹⁸ To reiterate, joint action is absent. Nor, on this picture, would anti-vaxxers be generating their own knowledge, since their beliefs in vaccine safety would not be grounded on any corroborating evidence of their own. Rather, as Thomas Grundman might put it, their doxastic attitudes would be nudged by “triggering automatic non-rational mechanisms of belief-formation” (2021: 5).¹⁹

As above, for knowledge to be distributed in such cases, hearers would need to occupy a bypass channel, and both the bypass channel itself and the information channel it taps into would need to be reliable. In the cases I am envisioning, anti-vaxxers do occupy a bypass channel, but this time it is a channel that has been built by speakers rather than hearers—indeed, it has been custom-designed by scientists and tech executives so that anti-vaxxers will occupy it, it has been built up around them (as it were). I will also assume that the channels of vaccine-science are reliable and that they are trading in knowledge about vaccine effectiveness and safety.

The tricky and tendentious bit is whether the bypass channel itself is a reliable one. If it isn’t, then at best the bypass channel will distribute truths but not knowledge. Ultimately, I suspect this will turn on how we delineate the relevant bypass channel. If we take the relevant bypass channel to be ‘believing as a result of repetition and fluency’ then it will be unreliable, given that there are masses of environments that use the repetition of falsehoods to facilitate false beliefs. But, if we take the relevant bypass channel to be ‘believing as a result of repetition and fluency in an environment designed by epistemically benevolent vaccine scientists and tech executives,’ then it will be reliable, given that the channels of vaccine-science (and epistemic benevolence) are themselves reliable.²⁰ I think an argument can be made for the latter (or something close to it), though I won’t pretend to make one myself: Grundman has insightfully argued that “belief-forming methods should be individuated externally; i.e., partly in terms of the epistemic agent’s environment” (2021: 12). For my part, I add that our educational systems likewise seem to be benevolently engineered to use repetition and fluency effects to distribute knowledge.

While this is an important objection, and more would need to be said to allay it, it may be unavailable to Greco who also recognizes the role of engineering in education and its success in transmitting knowledge to young children. On Greco’s view, the gullibility of young children does not prevent knowledge from being transmitted to them because “children are rarely left to themselves. On the contrary, we construct and monitor their social environments so as to keep them safe from insincere and incompetent speakers. Put differently, we engineer environments that enable the transmission of knowledge that their care and upbringing require” (2021: 140). But, if knowledge—and not just truth—is distributed to children in the engineered channels they occupy, then won’t knowledge—and not just truth—likewise be distributed to anti-vaxxers in their engineered channels? Arguably, it would, and these cases of engineering will stand or fall together.

Relatedly, one might object that knowledge requires not just reliability, but a sensitivity to it (Greco, 2021: 70), and that this sensitivity is missing in the case of anti-vaxxers. It is missing because anti-vaxxers won’t be behaviorally disposed to avoid and disbelieve unreliable sources about vaccines or seek out and believe reliable ones. This, too, is a fair objection. But, again, one that may be unavailable to Greco for the reasons already cited above, viz., he argues that we engineer environments to shield children from unreliable speakers whom they would otherwise believe. To put this reply differently, the gullibility of children arguably renders them no more sensitive to reliability than anti-vaxxers. But, if knowledge is nevertheless transmitted to children in their engineered environments as Greco attests, then it appears that knowledge would likewise be distributed to anti-vaxxers in their engineered environment. Again, the cases seem to stand or fall together.

Taking stock, I have argued that cases of denied access point us toward a picture of knowledge distribution that is broader than Greco’s—one that does not require joint action or trust, but does still require reliable transmission channels. Many questions remain, the role of sensitivity in cases of epistemic engineering needs to be decided, and further necessary (and sufficient) conditions need to be identified. That said, I hope to have shown that a broader account of knowledge distribution is needed. The concluding section addresses some further objections and implications.

3 Objections and implications

There are several worries about the arguments in the preceding section. For starters, we might think Greco never intended his account to apply to every case of knowledge distribution, and thus never intended to exclude cases of denied access in the first place. In other words, we might suspect that Greco’s account allows for other methods of knowledge distribution that don’t involve testimonial exchanges. Indeed, Greco indicates as much in a note: “The framework leaves open whether there are non-testimonial means of transmission” (2021: 68n2). The problem is that elsewhere, his view is committed to a closure clause, according to which “S knows that p only if S knows that p by Generation or Transmission,” a base clause of Generation whereby S knows that p if S’s arriving at a true belief that p is attributable to S’s reliable competences, and a recursive clause of Transmission whereby “H knows that p if, for some speaker S, S and H satisfy the conditions specified in KT” (2021: 62).²¹ If the suggestions in Sect. 2 prove viable, then either Transmission is false and there are other ways for knowledge to be transmitted and distributed, or the closure clause is false and Generation and Transmission are not the only ways of coming to know.

Second, we might think that Greco’s view already accommodates cases of denied access. Recall that Greco grounds the reliability of information channels not just in the sincerity and competence of speakers and hearers, but in contributions made by the social environment and its structures. Accordingly, we might think that there are already structures in our social environment that can explain the cases I have in mind. Along these lines, Greco argues that “contributions to reliable transmission by speaker and hearer might be bolstered by licensing agencies, supervisors, fact checkers, eavesdroppers, and the like” (2021: 196). My reply is this: eavesdroppers, fact checkers, and regulatory agencies are not acting jointly with speakers. Quite the contrary, they are needed because speakers obstruct the flow of information through exclusion and deception. Now, I agree that such structures contour our environment with bypass channels. But, I fail to see how this necessitates joint action as a condition on knowledge transmission. From my point of view, it explains why joint action isn’t needed.

Third, we might wonder whether the picture is committed to claiming that knowledge is distributed from the suspect to the police detective, rather than generated by the detective himself. In reply, I argued above that knowledge distribution still requires reliable information channels. So, if the suspect is unreliable (insincere) in communicating with the police, then his knowledge will not get distributed for that reason. In which case, the detective will acquire knowledge by generating it himself via direct evidence that corroborates the suspect’s claim (see (B) above). We might think of the unreliable suspect as kin to the Quirky Liar, whom Greco invokes in his answer to the garbage problem (2021: 190). On Greco’s view, the Quirky Liar is insincere relative to the general range of information in which the hearer is interested, and for that reason doesn’t transmit knowledge to the hearer (2021: 198). A suspect who usually lies to the police would (presumably) fail to transmit knowledge for the same reason, viz., we can attribute this failure of transmission to the unreliability of the suspect rather than to distrust.²²

Of course, if it is stipulated that the suspect is reliable, a different reply is needed. Here, I point to two alternatives that may be worth pursuing. First, a reply might argue that the detective does not acquire knowledge via transmission because he has a defeater, but nor does he acquire knowledge via generation if he relies solely on corroborating evidence for the suspect’s reliability (see (A) above). Instead, he acquires knowledge via some hybrid route that combines features of transmission with features of generation. Here, we would be denying Greco’s closure clause (2021: 62). Alternatively, our reply might retreat to biting the reliabilist bullet; i.e., we might argue that some sort of ‘lower-level’ knowledge does get transmitted from the suspect to the detective since the suspect is reliable, even if the detective lacks ‘higher-level’ knowledge because he has a defeater.²³ We might likewise resist the implication that this line of reasoning makes knowledge transmission too easy. Arguably, it won’t make the transmission of ‘lower-level’ knowledge too easy since reliability would be required. Nor would it make the transmission of ‘higher-level’ knowledge too easy since reliability at both levels, the belief that one knows, the absence (or defeat) of defeaters, and perhaps more, would be required.

In closing, I briefly clarify what is, and is not, implied by the picture of knowledge distribution I have proposed. First, although I have argued that knowledge distribution does not require joint action, I have not argued that joint action is unimportant. Greco’s conditions might still be sufficient for knowledge distribution, and even if they aren’t, we can expect joint action to play an important role in other tasks of the epistemic community, including bypassing epistemic obstructions and changing unjust epistemic norms. Second, nor have I argued that knowledge distribution is asocial. Quite the contrary, I have preserved the insight that social norms structure our epistemic environments and determine our locations in them. Whether that is enough to render it “a distinctive phenomenon of important epistemological interest” is well worth exploring (Greco, 2021: 5).

Finally, what are the implications of this picture for virtue epistemology? Recall that Greco responds to Lackey’s (2007) objection by expanding his achievement theory to include the competences and achievements of joint agents. Crucially, Greco’s response won’t be available to my picture of knowledge distribution, since on my picture, knowledge distribution does not entail joint agency. But, that needn’t prevent us from pursuing achievement theories, if institutions and structures in the environment can have virtues, or, to put this differently, if agency, competence, and achievement can be extended even further beyond the individual and into the environment.²⁴ Nor will it prevent us from pursuing virtue epistemology more generally, since there are many ways in which the competences and character virtues of agents will be important for our epistemic tasks, even if they prove unnecessary for knowledge distribution.