2. The mistaken-report argument

Having clarified the scope of my overall argument, let me turn to an argument for a mild form of testimonial skepticism that was originally introduced by Keith DeRose in the context of a discussion of traditional skepticism, which I will call “the mistaken-report argument.” DeRose notes:

Like most skeptical arguments, the mistaken-report argument consists of two premises. The first premise is that the protagonist of DeRose’s example (whom I will refer to as “Keith”) does not know that the report is not mistaken. Keith’s belief that the report is not mistaken is insensitive – had the report been mistaken (and his belief, thus, false), Keith would have still believed that the report was not mistaken, as he would have had no reason to believe otherwise.Footnote ⁷ Thus, the mere fact that Keith has no reason to believe the report is mistaken does not give him (sufficient) reason to believe that it is not mistaken, which means that his belief that the report is not mistaken is unjustified and, as such, it is not knowledge.Footnote ⁸ The second premise is that, if Keith does not know that the report is not mistaken, then he does not know that the Bulls won. If Keith does not have reason to believe that the report is not mistaken, then he does not have reason to believe that the Bulls did win either, which means that his belief that the Bulls won is also unjustified, and, as such, it is not knowledge either.

Why does DeRose take the mistaken-report argument to be only mildly skeptical? Presumably, it is because it scores lower than traditional skeptical arguments along two dimensions. The first dimension is what we might call its reach. The mistaken-report argument only threatens the epistemic status of a handful of Keith’s testimonial beliefs, such as his belief that the Bulls won or his belief that their opponents lost. In contrast, traditional skeptical arguments (such as the brain-in-a-vat argument I briefly mentioned above) threaten the status of most of our beliefs, including many non-testimonial beliefs that are completely out of the reach of the mistaken-report argument, such as Keith’s perceptual belief that he has hands.

The second dimension is what we might call its indefeasibility. If the brain-in-a-vat hypothesis were true, Keith would never be able to realize that he is envatted, as the hypothesis is designed to be completely indefeasible – envatted brains are not given any reason to believe that they are envatted. The mistaken-report hypothesis, on the other hand, seems highly defeasible. If it were true, it is likely that Keith would eventually find out, as, for instance, he might learn that the Bulls actually lost through a friend who attended the game or by reading a report in a different newspaper.

3. Testimonial triangulation

In this section and the next, I argue that, on closer scrutiny, the mistaken-report hypothesis is not as defeasible as I just suggested. Keith’s belief that the Bulls won last night is an inherently testimonial belief. Since the game took place last night, it is now impossible for Keith to independently confirm that the Bulls won, as to do so, he would have to either have attended the game (which he did not) or travel back in time (which he cannot). The inherently testimonial nature of that belief means that the mistaken-report hypothesis cannot be defeated as easily as one might think. To see why, suppose that Keith does come across a second newspaper that reports that the Bulls lost. What should he do? A first possible answer is that he should simply withhold belief. However, this answer seems to ignore the fact that, typically, we do not trust all informants equally. If I trust a reputable newspaper much more than a disreputable gossip magazine, I do not withhold belief about whether, say, a famous couple is separating when the two print conflicting reports on the topic. I would give the report by the newspaper much more epistemic weight than the one by the magazine.Footnote ⁹ Similarly, if Keith happens to trust the Tribune more than the Herald (when it comes to their coverage of the Bulls),Footnote ¹⁰ he is likely to give proportionally more weight to the Tribune’s report than to the Herald’s when the two conflict.

However, triangulating between informants is an unmanageable feat. The most serious problem, as far as we are concerned, is that it is unclear how one can be justified in trusting one informant more than another (when it comes to remote topics). In Keith’s case, individualists might suggest that he could justify his belief that the Tribune is more trustworthy than the Herald by independently assessing the trustworthiness of each informant (or at least its reliability).Footnote ¹¹ If, for example, the Tribune has correctly reported the result of each Bulls game that Keith has ever attended but the Herald has not, Keith might be able to justify his belief that the Tribune is more trustworthy than the Herald. However, this response presupposes that Keith has attended enough Bulls games and that each time he attended a game he checked whether both newspapers correctly reported the score of the game, which means that the topic in question is not really remote for Keith. But what if Keith has never attended a Bulls game (as I will assume here)? After all, even if, contrary to what I just assumed, Keith did happen to have plenty of first-hand evidence to independently assess the reliability of each newspaper’s coverage of the Bulls, he would still not have enough first-hand evidence to independently assess the reliability of their coverage of the vast majority of topics they cover, including the President’s whereabouts, the current rate of inflation, the number of deaths caused by drug overdoses each year, the magnitude of a recent earthquake in Japan, or the damage inflicted by an airstrike on Kyiv.

Individualists might reply that, when confronted with conflicting testimonies from two informants, Keith should simply turn to a third informant, which would essentially act as a testimonial tie-breaker. The underlying idea is, presumably, that it is highly unlikely that a majority of informants would be wrong. However, this is only true if two crucial conditions obtain.Footnote ¹² First, each informant’s testimony must be independent from that of the others. If, unbeknownst to Keith, the Herald gets all its basketball scores from the NBA website (because it is cheaper than sending its own reporters to the game), then the chances that both the NBA website and the Herald got the score wrong might be higher than the chances that the Tribune got it wrong, as the Tribune might be more reliable than the NBA website.Footnote ¹³ Second, each informant must be sufficiently reliable. If the scores reported by the Herald, the Tribune, and the Times are fully independent from one another, but, unbeknownst to Keith, the Tribune’s reporting is more reliable than either the Herald’s or the Times’, then, if the score reported by the Tribune differs from the one reported by both the Herald and the Times, it might still be more likely that the Tribune got the score right than that both the Herald and the Times did.Footnote ¹⁴ The problem, of course, is that, even if both conditions hold, it is unclear how Keith could know that they do, for, as we saw, he cannot independently assess the reliability or independence of each informant. Expanding Keith’s range of informants, therefore, does not make testimonial triangulation any more likely to succeed.

4. Testimonial concordance

Individualists might object that the sort of case I discussed in the previous section is rare. Informants that cover basketball usually agree on whether the Bulls won, so we rarely find ourselves in Keith’s predicament. However, there are two serious problems with this reply.

The first and more abstract problem is that we have already seen some reasons why Keith should not take the concordance among informants to be evidence of their reliability. If, for example, several newspapers get their basketball scores from the NBA website, then their concordance is not evidence of their reliability, as each of them is only as reliable as the NBA website, which, for all Keith knows, might not be reliable at all. More importantly, even if the informants’ testimonies are independent from one another, the concordance among them would not be evidence of their reliability, as, for all Keith knows, most of them might be unreliable. Since, as we have assumed, Keith has no way to independently check the reliability or independence of each informant, there seems to be no way for him to use their concordance as evidence of their reliability. Moreover, if Keith could independently assess the reliability of each informant, he would not have to rely on their concordance in the first place, which makes the reply viciously circular.

The more concrete problem is that, at best, this solution applies only to topics that are undisputed. The Bulls game score is a good example of such a topic. Fans who attended the game might disagree as to whether the final score reflected the performance of each team (or whether, say, it was partly the result of bad calls by the referees), but they do not typically disagree about what the final score was in the first place. The score was there on the scoreboard for everyone in attendance to see, whether they liked it or not. However, these days undisputed topics are relatively rare in our lives, and the endless discussions about sports show that even when a topic is undisputed (e.g., who won a game), there are often many disputed topics in its vicinity (e.g., whether the win was deserved).

Consider, for example, what many seem to consider an indisputable topic – the size of the crowd that attended Donald Trump’s first presidential inauguration. The day following that inauguration, the then-White House Press Secretary, Sean Spicer, claimed that it was the largest crowd to ever attend an inauguration.Footnote ¹⁵ Even if the question of whether Trump’s inauguration was the most widely attended were as unambiguous as the question of whether the Bulls won, there would still be a significant difference between them. While most fans who attended the Bulls game would be able to independently confirm that the Bulls won, not even those who attended Trump’s inauguration would be able to independently confirm whether it was the most widely attended ever. In fact, nobody would be able to independently confirm or disconfirm Spicer’s claim regardless of whether they attended Trump’s inauguration. This is because not even those in attendance would be able to independently ascertain how many people were in attendance or whether more people had attended any previous presidential inauguration (including those that occurred before they were born).Footnote ¹⁶ This means that all of us must outsource (wholly or partly) the task to others, which requires us to rely on an intricate web of testimonial chains.

Suppose, for example, that I read an article by the editor of the Washington Post’s fact-checking division, Glenn Kessler, which concludes that Spicer’s claim was false (Kessler Reference Kessler2017). Kessler’s conclusion is itself based on the testimony of a variety of informants on which Kessler relied in his investigation of the topic. Among them are two crowd scientists who, from photographs and videos taken in the hour before Trump’s speech, estimated that approximately 160,000 people were in attendance at that time. However, the crowd scientists’ testimony, too, was arguably based on the testimony of others. For one thing, each photograph and video they examined was itself a piece of testimony. To believe that a given photograph is a veridical depiction of Trump’s inauguration in the hour before Trump’s speech requires believing that the photographer who took it is testifying truthfully about when and where the photograph was taken and whether the photograph accurately depicts the scene it portrays (or, for example, was altered). And this is only one of the many testimonial chains that contributed to Kessler’s conclusion, each only as strong as its weakest link.Footnote ¹⁷

One might argue that, if I justifiably trust Kessler, then I need not trust his sources, as it was his responsibility to assess their trustworthiness (and it was their responsibility to assess the trustworthiness of their sources, and so on). But how can my trust in Kessler be justified in the first place?Footnote ¹⁸ Personally, I do not know anything about him other than the fact that he works for the Post. Insofar as I trust Kessler (or, indeed, anyone else at the Post), it is because I believe that the Post is a trustworthy newspaper, and I assume that trustworthy newspapers tend to employ journalists who are themselves trustworthy and to adhere to strict editorial processes. In other words, it is the Post’s responsibility, not mine, to monitor the trustworthiness of its editors and reporters and to act as a guarantor of their trustworthiness to its readers.

However, this response only pushes the question back one step. As we have seen in Keith’s case, none of us can independently check the trustworthiness of informants such as the Post, so it is unclear how I could ever justify my trust in the Post, let alone convince someone who does not already trust the Post that Kessler’s article gives them a decisive reason to believe that Spicer’s claim was false. This means that even a seemingly indisputable topic like the size of the crowd attending Trump’s first inauguration can become disputed among people who disagree as to which informants are worthy of trust.

5. Reputation

In this section, I outline a different strategy to resist the mistaken-report argument. Suppose that the report Keith read in the Tribune was mistaken and that, as a result, he falsely believes that the Bulls won. Suppose further that Keith later learns from a long-time friend who attended the game that, in fact, the Bulls lost. Keith now has sufficient reason to believe that the Bulls lost, as he has enough first-hand evidence to independently justify his belief that his friend is a highly trustworthy source. He also has reason to believe that the Tribune is not as reliable as he thought. In fact, after a few such incidents, Keith might start to realize that the Tribune is not particularly reliable, and he might start warning others. If a co-worker tells him that she read in the Tribune that the Bulls won again, Keith might warn her that the Tribune is not particularly reliable. Furthermore, it is likely that others in Keith’s community will also come to realize that the Tribune is not very reliable and communicate it to others. As a result of all this, the Tribune will eventually develop a bad reputation within the community, at which point it will cease to be a trusted or well-reputed informant (within that community).Footnote ¹⁹

This suggests the following response to the mistaken-report argument. If the Tribune were an unreliable informant, it would eventually develop a bad reputation within Keith’s community.Footnote ²⁰ So, the fact that the Tribune is still considered a trusted informant in Keith’s community gives Keith reason to believe that it is reliable, which, in turn, gives him reason to believe that the report is not mistaken. Thus, Keith can be justified in believing that the report is not mistaken even if he does not have any way to independently confirm that it is true or to independently assess the Tribune’s trustworthiness (or its reliability).

While this response is promising, it has extremely problematic consequences. If Keith is justified in believing that informants that are well-reputed in his community are trustworthy, then so are neo-Nazis, QAnon followers, flat-Earthers, anti-vaxxers, incels, or cult devotees, who inhabit communities in which a variety of extremely untrustworthy informants are well-reputed. To avoid this consequence, individualists might insist that the principle that people can justifiably trust informants that are trusted within their own community applies only to people who belong to communities that are testimonially discerning – i.e., communities in which the level of trust in an informant tracks the actual level of trustworthiness of that informant. It does not apply to neo-Nazis, QAnon followers, flat-Earthers, anti-vaxxers, incels, and cult devotees. This is because their communities are echo chambers – they are communities that insulate their predominantly false and inherently testimonial core beliefs from testimonial counterevidence by promoting trust in untrustworthy informants that support those beliefs and distrust in informants that do not – and, by definition, echo chambers are testimonially undiscerning.Footnote ²¹ The problem with this response, however, is that, even if it succeeds against the mistaken-report argument, it exposes those who adopt it to a more radically skeptical argument, which I call the echo-chamber argument.

6. The echo-chamber argument

Like other skeptical arguments, the echo-chamber argument too relies on two premises. Its first premise is that Keith does not know that he is not trapped in an echo chamber. The reason why this is so is similar to the reason why, according to the brain-in-a-vat argument, he does not know that he is not an envatted brain. Both beliefs are insensitive, as the underlying skeptical hypotheses are indefeasible. The brain-in-a-vat hypothesis is, as we have seen, indefeasible by design. If it were true, Keith would still believe that he is not envatted, as he would have no reason to believe that he is envatted, which means that the fact that he has no reason to believe that he is envatted gives him no reason to believe that he is not envatted. Keith’s belief that he is not trapped in an echo chamber is similarly insensitive, as the echo-chamber hypothesis (i.e., the hypothesis that he is trapped in an echo chamber) seems almost impossible to defeat as well. If it were true, Keith would still believe that he is not trapped in an echo chamber. This is because, to realize that he is in an echo chamber, Keith would have to be able to independently establish that his community is testimonially undiscerning, which would require him to independently determine whether informants that are trusted in his community are actually trustworthy (and informants that are distrusted are untrustworthy). And, to do that, Keith would have to be able to independently assess the trustworthiness of a variety of informants. However, the very reason why Keith must rely on the reputation of informants as evidence of their trustworthiness is that he cannot independently assess their trustworthiness. This means that, if Keith were trapped in an echo chamber, he would still not believe that he is trapped in one, as he would have no reason to believe that he is, which means that the fact that he has no reason to believe that he is trapped in an echo chamber gives him no reason to believe that he is not trapped in one. The echo-chamber hypothesis, thus, appears to be (almost) as indefeasible as the brain-in-a-vat hypothesis.

The second premise of the echo-chamber argument is that, if Keith does not know that he is not trapped in an echo chamber, then he might not know many of the things that he takes himself to know about the world. This is because most of his beliefs about the world are inherently testimonial, and, for any subset of Keith’s inherently testimonial beliefs, we can devise a version of the echo-chamber hypothesis that undermines the epistemic status of each of the beliefs in that subset. The general schema is simply: “I am trapped in an echo chamber whose core beliefs include the inherently testimonial beliefs in the relevant subset even if each of those beliefs is false.” This means that the echo-chamber hypothesis can threaten any subset of Keith’s inherently testimonial beliefs. While the reach of the echo-chamber argument might not be as wide as that of the brain-in-a-vat argument, it can, thus, extend as far as all of Keith’s inherently testimonial beliefs.

While the echo-chamber argument is not as radical as the brain-in-a-vat argument (due to its narrower reach), it might still pose a significant (skeptical) threat. In contemporary epistemology, the brain-in-a-vat argument has largely supplanted the evil-demon argument despite the fact that the two are roughly equally radical (where their radicality is a combination of their reach and their indefeasibility). In my view, this is because the former poses a greater threat than the latter, as the brain-in-a-vat hypothesis is more plausible than the evil-demon hypothesis (for it does not presuppose a Cartesian-style mind-body dualism). And the echo-chamber argument poses a much greater threat than the brain-in-a-vat argument. Aside from the few people who genuinely believe the Simulation Hypothesis, everyone agrees that no actual person is currently envatted – the brain-in-a-vat hypothesis is merely a fanciful piece of science fiction employed to make an epistemological point. The case of the echo-chamber hypothesis, however, is different. Nowadays, it is common to believe that many people are, in fact, trapped in echo chambers.Footnote ²² The reason why the plausibility of a skeptical hypothesis matters to the threat it poses is that it is much harder to dismiss a plausible hypothesis than an implausible one. If we had reason to believe that many people are, in fact, envatted, we would have to take much more seriously the possibility that we too are envatted. However, we do seem to have reason to believe that many people are, in fact, trapped in echo chambers, which makes the echo-chamber hypothesis much harder to dismiss.Footnote ²³ Regardless of whether the echo-chamber argument is as threatening as traditional skeptical arguments, it seems undeniable that it poses a significant threat to our inherently testimonial knowledge, one that lends support to a radical form of testimonial skepticism. Over the next three sections, I evaluate three general strategies for trying to resist its conclusion.

7. Structural clues

A first general strategy to try to resist the conclusion of the echo-chamber argument is to insist that the echo-chamber hypothesis is not as indefeasible as I claim it to be. Individualists might try to do so by maintaining that echo chambers can be recognized by identifying features that are both distinctive (i.e., they are individually necessary and jointly sufficient for a community to qualify as an echo chamber) and structural (i.e., they are epistemically neutral features of the community that members and non-members can recognize equally well).Footnote ²⁴ This means that, even if Keith cannot independently determine that his community is not testimonially undiscerning, he can still know that he is not trapped in an echo chamber by recognizing that his community lacks at least one of the distinctive structural features of an echo chamber.

Consider, for example, Thi Nguyen’s prominent account of echo chambers. On a structural (mis)interpretation of it, the account maintains that echo chambers are characterized by two distinctive structural features. The first is that they are characterized by “a significant disparity in trust between members and non-members” (Nguyen Reference Nguyen2020: 146; emphasis in original) – its members assign high levels of trust to the other members and high levels of distrust to non-members, who are “actively assigned some epistemic demerit, such as unreliability, epistemic maliciousness, or dishonesty” (Nguyen Reference Nguyen2020: 146). The second is that the acceptance of “a core set of beliefs [(including beliefs that support the disparity in trust)] is a prerequisite for membership” (Nguyen Reference Nguyen2020: 146; emphasis in original).

These two features are structural because both members and non-members are able to recognize that the community exhibits them – i.e., they can all agree that its members share a core set of beliefs and that its members assign high levels of trust to other members and high levels of distrust to non-members. What members and non-members are likely to disagree on is the epistemological evaluation of those features. Its members presumably believe that their community’s core beliefs are true and that the disparity in trust is warranted, while most non-members are likely to disagree with them on both counts.

This structural interpretation of Nguyen’s account, however, faces (at least) three problems. The first is that (as Nguyen himself acknowledges)Footnote ²⁵ it is possible for a community to exhibit both structural features without being an echo chamber. For instance, political movements operating under repressive socio-political circumstances (such as the resistance movements in Nazi-era Germany, Fascist-era Italy, or Apartheid-era South Africa) tend to display the two structural features of an echo chamber without qualifying as echo chambers. The general lesson seems to be that it can be epistemically healthy for a community experiencing oppression or repression to exhibit the two structural features identified by Nguyen, which means that a community exemplifying those two features is not necessarily an echo chamber.

The second problem is that (as Nguyen also acknowledges)Footnote ²⁶ echo chambers simply exploit the fact that we must all rely on the reputation of informants within our community as evidence of their trustworthiness. Nguyen, thus, eventually concedes that it is not sufficient for a community to instantiate the two structural features he had initially identified for it to be an echo chamber. His considered view seems to be that the truly distinctive feature of an echo chamber is not its disparity of trust in different informants but the fact that the level of trust in different informants does not reflect their actual level of trustworthiness. If this interpretation is correct, then a community must display a third non-structural feature in order to be an echo chamber – it must also be (what I call) testimonially undiscerning.

The third problem is that it is unlikely that most real-world echo chambers exhibit the first feature. On the one hand, real-world echo chambers likely include members who distrust one another for all sorts of reasons (e.g., a member of a neo-Nazi group might distrust another who he falsely believes to be an undercover agent). On the other hand, real-world echo chambers likely include members who trust non-members on a variety of topics (e.g., an incel might trust the cashier when she tells him that the total is $53.45, and the anti-vaxxer might trust the ER doctor when she diagnoses him with a broken leg).

Supporters of a structural interpretation might reply that the disparity in trust between members and non-members only applies to topics related to the community’s core beliefs. So, for example, a QAnon follower might trust the Washington Post to truthfully report the score of a basketball game but not the results of a presidential election. However, this topic-restricted version of the account would seem to classify many epistemically healthy communities as echo chambers. For instance, it would misclassify as an echo chamber a community whose members believe that anthropogenic climate change is real because they trust the vast majority of climate scientists and distrust informants that deny its reality (even if such a community is neither epistemically unhealthy nor testimonially undiscerning).

Of course, those who still wish to maintain that echo chambers exhibit distinctive structural features have a variety of options available. They could modify the structural account I discussed so far or offer an entirely new account. However, the problem is that, even if echo chambers are characterized by distinctive structural features, we still do not know what they are, and thus, we cannot use them to rule out the possibility that we are trapped in an echo chamber. Moreover, in trying to identify the distinctive structural features of echo chambers, we usually start from our own pre-theoretical intuitions about which communities are echo chambers and which are not, which include the intuition that our own community is not an echo chamber. This means that any structural account of echo chambers is going to beg the question when it comes to our own community.

One might still wonder whether it is necessary to identify structural features shared by all echo chambers in all possible worlds in order to avoid the echo-chamber argument. Wouldn’t it be sufficient to just identify structural features typically shared by echo chambers in socio-epistemic contexts relevantly like our own? For example, one might argue that a community that has certain structural features might not be an echo chamber if it operates in the inhospitable epistemic environment of a totalitarian dictatorship (as the anti-Fascist resistance), but it would be one if it operated in the much more hospitable epistemic environment of a liberal democracy. However, this suggestion presupposes that we can independently assess how hospitable our epistemic environment is regardless of whether we are trapped in an echo chamber, which seems highly dubious. Presumably, those who believe the propaganda of totalitarian regimes mistakenly believe themselves to be in a hospitable epistemic environment.Footnote ²⁷ Moreover, some would argue that the epistemic environments of liberal democracies are not as hospitable as this suggestion takes them to be.Footnote ²⁸

8. Narrow escapes

Another general strategy to resist the conclusion of the echo-chamber argument is to argue that the echo-chamber hypothesis is not as indefeasible as I claim it to be because, as a matter of fact, people do sometimes manage to escape echo chambers. Nguyen, for instance, discusses the case of R. Derek Black, who was raised in the Ku Klux Klan but, eventually, came to reject its ideology. Individualists might argue that people like Black are living proof of the fact that people who inhabit echo chambers are not trapped in them – they can come to realize that their communities are echo chambers and escape them.

Here, we must carefully distinguish two points individualists could be making by bringing up these alleged escapes. The first point is that the echo-chamber hypothesis is not as indefeasible as traditional skeptical hypotheses. However, this point is not particularly controversial. As we saw, the brain-in-a-vat hypothesis is designed to be completely indefeasible, but this is part of the reason why it is implausible. Current technologies have bugs and glitches, and we have no reason to believe that any actual technology will ever be as infallible as the supercomputers running the brain-in-a-vat simulation. The echo-chamber hypothesis, on the other hand, is not designed to be indefeasible – it presents us with a live possibility. The fact that there are plenty of people who inhabit echo chambers (as Black once did) but who (unlike Black) will never realize it should make us more, not less, inclined to suspect that we too might be unaware that we are in an echo chamber.

The second point individualists might be trying to make is that escapees prove that it is possible for those trapped in an echo chamber to come to know that they are trapped in one, which means that, if Keith were in an echo chamber, he too might be able to realize that he is. However, the existence of escapees does not actually prove that much. First, even if we were to concede that it is sometimes possible for some people who inhabit some echo chambers to realize that they are (or were) in an echo chamber, this does not mean that it is always possible for anyone who inhabits any echo chamber to do so. In particular, it does not show that Keith (or you or I) would be able to do so if, indeed, we happen to be trapped in one. In fact, cases like Black’s seem to show how unlikely such escapes are. After all, Black only left the Klan after going to college and being befriended by a Jewish student – a fortuitous chain of events that illustrates how escaping an echo chamber requires a great deal of luck.

Second (and more importantly), individualists must concede that, just like it is possible to exit (what we consider) echo chambers, it is also possible to exit (what we consider) epistemically healthy communities. For example, those who fall into the rabbit hole of online conspiracy theories or those who are recruited by white nationalist groups often end up believing that they are the ones who have escaped an echo chamber. The fact that there are people who exit a community is, thus, no reason to assume that those people are thereby in a position to know that their old community was an echo chamber.

9. Traditional maneuvers

A third general strategy to resist the conclusion of the echo-chamber argument is to perform one of the many philosophical maneuvers that have been developed over the centuries to respond to traditional skeptical arguments. However, there are at least two reasons to doubt that this strategy is likely to succeed. The first is that, after centuries of debate, there is still no agreement about whether any of these maneuvers is effective against traditional skepticism. The second reason is that, even if some of those maneuvers are effective against traditional skeptical arguments, it is unclear that they would be equally effective against the echo-chamber argument. To see why, let me briefly consider three prominent such maneuvers.

The Moorean maneuver claims that traditional skeptical arguments rely on premises that are less justified (and more dubitable) than the negations of their conclusions. My belief that I know that I have hands, for instance, seems much more justified (and less dubitable to me) than the premise that I do not know that I am not envatted, so, instead of inferring that I do not know that I have hands from the premise that I do not know that I am not envatted, I should infer that I know that I am not envatted from the premise that I know that I have hands. The gap between the premises of the echo-chamber argument and the negation of its conclusion, however, is nowhere near as wide. My belief that I know that Napoleon died in 1821 is much less justified (and more dubitable to me) than my belief that I know that I have hands, while the premise that I do not know that I am trapped in an echo chamber seems more justified (and less dubitable to me) than the premise that I do not know that I am envatted (for I believe that many people are trapped in echo chambers but nobody is envatted).

The contextualist maneuver concedes that it would be correct to deny that I know that I have hands in conversational contexts in which the possibility that I am envatted is salient (such as a philosophy seminar) but not in contexts in which such a remote possibility is not salient (such as most everyday conversations).Footnote ²⁹ However, the same does not apply to the echo-chamber argument, for, as we saw, the echo-chamber hypothesis is nowhere near as remote a possibility, and it is highly salient in many ordinary conversational contexts (such as in animated political debates).

The third maneuver, which appeals to epistemic “hinges,” claims that certain fundamental commitments – such as the belief that the external world exists – function as “hinges” on which our epistemic practices turn. These commitments are not themselves justified or unjustified in the ordinary sense; rather, they are presupposed by the very practice of justifying beliefs and cannot themselves be subjected to justification without circularity. One might argue that the assumption that one’s own community is testimonially discerning functions as an epistemic “hinge” – it is a commitment that is presupposed by the practice of relying on testimony and evaluating the trustworthiness of informants, and as such it cannot itself be required to meet ordinary standards of justification.Footnote ³⁰ However, when applied to the echo-chamber argument, the “hinge” maneuver does not avoid testimonial skepticism. While we (almost) unanimously agree about who is envatted (i.e., nobody), we strongly disagree about who is trapped in an echo chamber (“we” think it’s “them,” while “they” think it’s “us”). But this means that, if we are entitled to presuppose that our own community is discerning, so are neo-Nazis, QAnon followers, or flat-Earthers about their own respective communities. Whether my inherently testimonial beliefs are justified thus entirely depends on whether they are being judged by the members of my own community or by the members of one of those communities. However, each community would want to insist that their standards are objectively correct and that their community is discerning even if neither we nor they can know (or justifiably believe) that they are.

While these brief remarks are far from conclusive, they do suggest that, even if the standard maneuvers might be effective against traditional skeptical arguments, they are unlikely to be as effective against the echo-chamber argument.

10. Communitarianism

So far, I have argued that it is difficult to resist the conclusion of the echo-chamber argument. But what if, instead of resisting it, we were to embrace it? Communitarians do just that. They concede that, individually, each of us might not know much, while insisting that we can still know a great deal collectively.Footnote ³¹ This is because, according to them, if our inherently testimonial beliefs amount to knowledge, it is only in a derivative sense – it is only because we belong to a community that collectively knows those things.

To illustrate the point, consider again the case of Keith. As we have seen, Keith does not know that the Bulls won because his belief that the Tribune is a reliable informant is unjustified. This is because, even if Keith knows that the Tribune is well-reputed within his community, he cannot know that informants that are well-reputed in his community are typically trustworthy. However, even if, individually, Keith cannot know that the Tribune is trustworthy or that the Bulls won, his community can collectively know both things.

The communitarian solution comes in two steps. The first step is to maintain that Keith’s community can know that the Tribune is trustworthy even if Keith cannot know it (except in a derivative sense). The basic idea is that the Tribune’s reputation within the community is the functional analogue of a collective belief about the Tribune’s trustworthiness and that it is justified insofar (and only insofar) as it fits the evidence collectively possessed by Keith’s community. This requires (i) that enough community members have evidence about the Tribune’s trustworthiness, (ii) that they share it with the community, and (iii) that the shared evidence informs the Tribune’s reputation within the community.Footnote ³² So, even if Keith’s belief that the Tribune is trustworthy is unjustified, the community’s collective “belief” is not, provided that it fits the community’s shared evidence. Moreover, if the Tribune’s reputation within the community is not only justified but also well-deserved (i.e., if it reflects the Tribune’s actual trustworthiness), then the community collectively knows that the Tribune is trustworthy. When Keith truly believes that the Tribune is trustworthy on the basis of the fact that it is well-reputed in his community, his belief is not true merely due to a stroke of epistemic luck – it is true because Keith belongs to a community that collectively knows that the Tribune is trustworthy. If his belief amounts at all to knowledge, it is thus only in a derivative sense – it is because he belongs to such a community.

The second step is to maintain that Keith’s community can collectively know that the Bulls won the game even if Keith cannot know it (except in a derivative sense). If the Tribune happens to be the only informant known to be trustworthy (about the Bulls) by Keith’s community, the community knows that the Bulls won insofar as the Tribune’s reporter knows that the Bulls won and she shares her knowledge with the community by publishing her report in the Tribune.Footnote ³³ When Keith believes that the Bulls won on the basis of reading about it in the Tribune, his belief is, once again, true not merely due to a stroke of epistemic luck – it is true because Keith belongs to a community that, collectively, knows that the Bulls won. If his belief amounts at all to knowledge, it is thus only in a derivative sense – it is because he belongs to such a community. Strictly speaking, then, it is only his community (not Keith) that knows that the Bulls won because the justification for both beliefs is irretrievably distributed within the community.

As this example shows, communitarians take advantage of the fact that communities extend in space and time. Any sufficiently large community is likely to include members for whom a given topic is not remote. This means that a topic can be remote for most individual members of a community without being remote for the community as a whole. However, not all cases are as straightforward. Some topics are deeply remote – i.e., they are remote for every individual.Footnote ³⁴ For example, no individual physicist can independently confirm that uranium has no stable isotopes.Footnote ³⁵ Experts do not typically know what they know because each has independently examined all the relevant first-order evidence. Even if it were feasible to do so, it would be wildly inefficient. Much expert knowledge is both coproduced (i.e., it is the product of a collective effort to gather and evaluate evidence) and distributed (i.e., no individual member can independently examine all the evidence). If experts’ beliefs about deeply remote topics are justified at all, they too are only justified in a derivative sense – they are justified in virtue of the fact that the community has collective justification for those beliefs. Beliefs on deeply remote topics are like pictures on an enormous jigsaw puzzle, and each piece of evidence is like a piece of that puzzle. Different members of the expert community hold different pieces of the puzzle, and it is only by working together that they can assemble enough of the puzzle for a relatively unambiguous picture to emerge.Footnote ³⁶

The same is even more clearly true for the rest of us. If I know that uranium has no stable isotopes, it is only in a derivative sense – it is because the physics community collectively knows this; its collective knowledge has reached my community through informants that my community knows to be trustworthy, and I have partaken in that knowledge by relying on the testimony of some of those informants. In this case, the physics community is playing the role played by the reporter in Keith’s case, and the informants through which my community partakes in that knowledge play the role played by the Tribune in that case. But who plays the role of the community members who attend the games? In other words, how can my community know who the trustworthy informants about physics are? The answer lies in the fact that any sufficiently large community is likely to include dual members – i.e., members of the community who are physicists and, thus, are also members of the (global) community of physicists. These dual members will be able to determine which informants are trustworthy about physics and, if everything works properly, share that with the broader community, thereby grounding its collective reputation tracking for physics informants – just as game attendees ground the community’s reputation tracking for the Tribune’s sports coverage. If, for example, a story in the Tribune mentions that a second stable isotope of uranium has been discovered, those dual members will quickly identify the error, and, if my community’s reputation tracking mechanisms work properly, this will affect the reputation of the Tribune as an informant on scientific matters within it. The upshot is that even for topics that are deeply remote, the communitarian account still applies provided the community includes a subcommunity of members who also belong to the relevant expert community and can perform the independent verification on everyone else’s behalf thanks to the knowledge they have in virtue of their membership in that community.Footnote ³⁷

It is helpful to contrast these cases of collective knowledge with one of collective ignorance. Consider an anti-vax community that believes vaccines cause autism. Such a community fails on both steps of the communitarian account. First, the reputation of its informants (about vaccines) fails to track their actual trustworthiness (on that topic). In fact, the community is not simply undiscerning but anti-discerning on this topic – its well-reputed informants are untrustworthy, and its ill-reputed ones are trustworthy. Even if the community includes doctors or other dual members, they fail to help the community identify the trustworthy informants on the topic. Second, while the public health community knows that vaccines do not cause autism, the anti-vax community fails to partake in that knowledge, due to the first failure.

Let me conclude this brief sketch of communitarianism by emphasizing three points.

The first point is that communitarians are internalists about justification. They maintain that my beliefs that are not inherently testimonial are justified in virtue of the reasons I individually possess. If I know that there is coffee in my mug, it is because I individually have reason to believe it. By the same internalist standard, my inherently testimonial beliefs are not justified – individually, I do not have reason to believe that uranium has no stable isotopes. Collective justification only kicks in when individual justification runs out. It is only when it comes to inherently testimonial beliefs that the primary bearers of justification are communities rather than individuals. However, just as what justifies an individual’s beliefs must be accessible to the individual, what justifies a community’s collective beliefs must be accessible to the community – it must consist of reasons that have been shared within the community rather than merely privately held by individual members.Footnote ³⁸ A community that trusted informants depending on their astrological signs would fall short of the communitarian standard even if, serendipitously, the informants’ reputations happened to be well-deserved. This is because its beliefs about the informants’ trustworthiness, while accidentally true, would not be grounded in the evidence collectively possessed by the community.

The second point concerns the epistemic responsibilities of individuals. If communitarians are right, then I do not know much of what I take myself to know unless I am fortunate enough to belong to a community that knows those things. By the same token, people who hold inherently testimonial beliefs that, to us, seem bizarre or unfounded typically do so through no fault of their own – they have the misfortune of belonging to testimonially undiscerning communities.Footnote ³⁹ Individually, they are no more epistemically blameworthy for their unjustified inherently testimonial beliefs than we are for our equally unjustified ones, for the difference between us and them is largely a matter of epistemic luck. This, however, does not mean that communitarians take individuals to bear no epistemic responsibilities. First, communitarianism exempts individuals from epistemic blame for their beliefs only insofar as those beliefs are inherently testimonial. However, many problematic beliefs are not fully inherently testimonial. Individuals who hold racist or sexist beliefs, for example, are epistemically blameworthy insofar as they have non-testimonial evidence that weighs against those beliefs. Second, epistemic responsibility is not exclusively a matter of belief. For example, individuals are epistemically responsible for how their epistemic conduct contributes to (or detracts from) the epistemic health of their communities. We have seen a simple example of this earlier. The Tribune’s reputation functions as a kind of public “scoreboard” of its trustworthiness. However, the community can “keep score” only insofar as its members contribute to this collective effort. If too many of them have excessive trust in the Tribune (or an interest in protecting its reputation or that of the Bulls), the community might fail to hold the Tribune epistemically accountable, failing to downgrade its reputation even after it repeatedly publishes mistaken reports.

The third point concerns a distinction that social epistemologists have not always made explicit. Social epistemologists rightly emphasize the social division of epistemic labor, but they typically focus on zetetic labor (i.e., the labor that goes into producing (new) knowledge) and gnostic labor (i.e., the labor that goes into acquiring and retaining (old) knowledge).Footnote ⁴⁰ Communities divide their gnostic and zetetic labor because doing so is efficient. Producing, acquiring, and retaining knowledge is costly, and it is far more efficient to delegate the production, acquisition, and retention of specialized knowledge to expert communities. Not as much attention, however, has been paid to the epistemic labor that goes into retrieving that knowledge by figuring out which informants can be trusted. We might call this third kind of epistemic labor fiducial labor. Communitarians maintain that reputation helps divide fiducial labor by collectivizing it, distributing the task of tracking the trustworthiness of informants across the whole community rather than leaving it to each individual. This division of fiducial labor, however, has consequences that many are likely to find uncomfortable. One of the most important is that it means that the belief that our own community is testimonially discerning is, at the individual level, a form of faith, as none of us has adequate epistemic reasons to hold it. However, this is not an undesirable conclusion but an honest acknowledgment of our epistemic condition. We are not the self-sufficient epistemic agents that individualists imagine us to be. Much of our knowledge is derivative. Our collective knowledge is like the collection of a public library, whose books individual users can only ever borrow but never own.

Much work remains to be done to further develop communitarianism. What makes communities epistemically healthy or unhealthy? And how can individuals contribute to the epistemic health of their communities? How can we improve the broader epistemic infrastructure on which communities depend? Unfortunately, I cannot address these questions here, but I hope that the arguments in this article might persuade others to contribute to this broader project, for, as a communitarian, I take this effort too to be a collective endeavor.