1.1. The chamber: a social/epistemic structure

The first part of any definition of an echo chamber concerns the setting in which the phenomenon occurs. This setting is usually described as a social/epistemic structure, context, or community. For example, Lackey speaks about ‘an enclosed system or “chamber”, such as a social network’,Footnote ⁷ and describes an ‘enclosed system of agreement’Footnote ⁸ when she talks about examples like Donald Trump’s staff keeping the channels tuned to Fox News. Jon Robson lists a key feature of echo chambers as being that ‘Echo chambers are in some respect – be it geographical, cultural, or otherwise – enclosed spaces’.Footnote ⁹ Nguyen refers to ‘a social epistemic structure’, and later ‘an epistemic community’.Footnote ¹⁰ Bert Baumgaertner describes ‘a sociological setting’ and gives examples of ‘media networks…blogs and online niche forums…and even the right setting in a classroom of peers’.Footnote ¹¹ For Eric Gilbert et al., giving an earlier definition, it was simply ‘a blog’.Footnote ¹² Kathleen Hall Jamieson and Joseph N. Capella’s analysis of echo chambers applies to ‘a bounded, enclosed media space’.Footnote ¹³ Although these latter two definitions are somewhat more restricted, aiming to talk specifically about blogs in one case and media in another, we will look more widely than either online or mediated echo chambers, and we take it to be true to the epistemic/social phenomenon to say that it can also occur in real life, such as in political movements, discussions in real-life friendship groups or down the local pub, etc.

In some ways, this will be the simplest and most basic aspect of the definition: it’s just where the echo chamber is located – the setting in which the supposed problems occur. Think of an example of a paradigmatic case of an echo chamber: news consumption on the internet. Suppose Mary has given up on buying newspapers or listening to the radio, and she gets most of her news from articles that she sees shared on social media. However, because of the way the social media platform works, this limits her to only really reading articles and headlines that her chosen friends on that platform have decided to share. If there are articles out there and news events that her friends don’t see the need to share, then Mary won’t see them. One way to understand the social epistemic structure in this echo chamber, then, is to see it as the structure of Mary’s friends on social media. Together, they make up a group of people who (through the platform) filter what news and headlines Mary will see on a day-to-day basis. If Mary is ‘trapped in an echo chamber’, then this is where the echo chamber is – the social media, and the group of friends that make it up. It is epistemic because the group can be a source of information, and it is described as social because it is a group, although we note that term might be misleading because the information exchange in some paradigmatic cases (such as a charismatic videoblogger and their audience) might not always be reciprocal.

Many of the above snippets from the literature also specify that the space in question is ‘closed’ in some way. We turn to that feature next.

1.2. The voice: a shared core belief

To begin a process of narrowing down the phenomenon supposedly picked out by the term ‘Echo Chamber’, the next step is to look at what theorists mean when they describe the relevant structures as being ‘closed’. This is because the structure is based around a ‘core belief’, as Nguyen calls it,Footnote ¹⁴ or an ‘accepted view’ in the words of Robson.Footnote ¹⁵ This core belief is what defines the boundaries of the echo chamber, and what determines whether you are an insider or an outsider.

Let’s think back to Mary’s social media. Of course, Mary is likely to be friends with only certain types of people – her friends may overwhelmingly be people of her own age, from the country she lives in, who speak her language(s) and share some of her interests. But Mary might also select – intentionally or not – for those who share at least some important aspects of her political beliefs. She might be hesitant to befriend anyone who doesn’t agree with her stance on her favourite local political party, for example. And even if she is friends with some who disagree with her on these issues, it’s still likely to affect the numbers overall. Membership of Mary’s echo chamber, then, is made up of people who largely share certain core beliefs. The definition of echo chambers gets a bit narrower.

1.3. The echoing: amplification of the core belief

The third feature concerns the way the system handles information. Echo chambers, in virtue of their structure, operate in a way that amplifies the core belief, while ignoring or undermining contradictory beliefs. Perhaps the clearest explanation for this in the literature comes from Lackey, who describes it in the context of the ‘echo chamber’ metaphor:

Just as an echo from a single voice can bounce around a chamber and come back to the hearer, so too can information bounce around an epistemic structure and come back to the speaker reinforced by the shape of that structure. For example, this might happen if the members of the structure repeat similar views to each other while excluding contrary views.

Pablo Barberá et al. also describe this selective exposure:

Florian Justwan et al. make a similar point, describing echo chambers as ‘ideologically congruent and homogeneous environments in which political views are not debated but instead reinforced and amplified thus paving the way to increased polarization’.Footnote ¹⁸ Baumgaertner describes them as settings in which ‘peoples’ prior beliefs are “echoed back” giving the impression that their beliefs are correct’.Footnote ¹⁹ Robson talks about views being repeated and reinforced.Footnote ²⁰

Let’s think about Mary again. Mary’s social media is made up of people who share a positive disposition towards Mary’s favourite local political party. The shape of this social structure means that there are several ways that core belief(s) and other related beliefs might be reinforced. For example, each individual friend will – because they share that belief – be more likely to share news stories that in some way support or are at least coherent with this belief. News stories and opinion pieces about things that members of the party have done well, and things that members of opposing political parties have done badly – will be more likely to be shared by Mary’s friends and therefore appear more frequently on her social media. This might be due to biases in which articles they enjoy, it might be due to what they think will interest their friends or might just be that they’re more likely themselves to look for articles in places which provide content with a certain leaning. Here the core belief reinforces itself – the more such articles are shared, the more the belief may become magnified, and so on. Mary might think her love for her local political party is more justified than it actually is because the information she surrounds herself with artificially reinforces that idea.

Note that in this example there is not necessarily any deliberate attempt on the part of those involved to misrepresent the facts. Nevertheless, the process described above is distinct from ways in which information might get reinforced ‘naturally’. It’s not the case that Mary is looking at an objectively neutral dataset and then happens to see a lot of information that reinforces her views. Rather, the dataset she has access to is already filtering for certain kinds of information. This brings us to our final element of the definition, which looks in more detail at the way this information can be manipulated.

1.4. The manipulation of trust

The final feature of a definition of echo chambers comes primarily from Nguyen, who does good work in developing and clarifying the definition of echo chambers. He says:

According to this definition, then, an echo chamber is not just a social space of some kind where information echoes, and then we make mistakes in the way we interpret the information that the social space provides us. There is an added layer built into echo chambers: they actively give undue credit to certain information and actively discredit other kinds of information. The chamber does something to give disproportionate credit to views that match the core belief and discredit those which go against it.

Nguyen gives some examples of how this might work. He says echo chambers can manipulate the data by making particular use of:

• Charismatic personalities (such as Rush Limbaugh)

• Consistent attacks on outside sources (those that give opposing views to those shared by the chamber)

• Selectively limiting approved sources of information

• Exaggerating differences between insiders and outsiders of the group by creating new terms (such as ‘SJWs’ – a derogatory term to label people with outsider views)

• Giving stories and explanations that counter any contrary views, ones that paint the givers of those views as untrustworthy for various reasonsFootnote ²²

Each of these methods work to make the shared beliefs more credible, and any opposing beliefs untrustworthy, no matter what their objective epistemic merit might be.

Let’s imagine that Mary’s social media group also meets this final criterion for being an echo chamber. This means that in addition to there being a select group that shares and amplifies a core belief about a local political party, it also does something further to manipulate the available information. This might take a number of forms. For example, when someone posts a link to a news story that supports the group’s shared political beliefs, a number of members of the group might share it again, or leave comments or ‘likes’ that praise the sharing of the article. When someone in the group posts a dissenting view instead, they might be the object of ridicule, or have excuses given as to why they’d make such a mistake. In some cases, the people posting dissenting views might be criticised as immoral, or presented as trying to manipulate others or profit in some way from spreading lies. In other cases, they might be presented as victims – easily manipulated and tricked by some others into believing and sharing things that aren’t true.

The chamber’s active crediting and discrediting might be done by group members on an individual level, or by other features of the structure itself. In Mary’s case above, there are individuals who will each try to explain away contrary views, for example, or bolster friendlier ones. But sometimes this undue credit/discredit will happen because of (entirely or in part) something more structural like an algorithm, or like the prerequisites that might be required to become a member of a certain epistemic group. These might not just filter out certain views (which leads to the echoing in 1.3) but also present outsider views in a different way to insider views, contributing to (and at least partially constituting) their different treatment.

Even when the bolstering does result from the actions of individuals, it often won’t be something that the individuals are consciously or intentionally doing. After all, Mary’s friends will often have good reasons to bolster or discredit the information that they do, that has nothing to do with intentional manipulation. Echo chambers, then, can be manipulative even when the individuals inside don’t know it. The active crediting and discrediting of sources is one of the most challenging elements for our paper. If we are to argue that echo chambers are not so problematic after all, then this particular element will be the toughest part to argue against. But even this, as we’ll argue in S3, is not a problem in the way that people generally take it to be. To make the strongest case we can, then, for the rest of the paper we’ll treat echo chambers as structures that meet all four of the defining features we’ve listed above, even if the definition in recent scholarship is sometimes a little muddier than this. But since we’ll go through the four criteria separately, our argument will also work against other definitions of echo chambers.

By now it should be starting to become clear why echo chambers are thought to be so problematic. In S2, we will outline the criticisms more specifically.

2.1. Restricted or unfairly presented information

Perhaps most obviously, echo chambers make it more likely that we will come to hold false beliefs by restricting our access to diverse viewpoints. Objections to this aspect of echo chambers often invoke one of the classical Millian defences of the value of freedom of speech according to which exposure to diverse viewpoints is important for determining the truth since this maximises our chances of being exposed to the correct viewpoint in case ours turns out to be wrong.Footnote ²³ Our access to diverse viewpoints may be restricted in different ways, depending on the echo chamber in question. Some echo chambers may simply fail to supply us with the relevant information, by consistently excluding perspectives that dissent from the core belief(s) around which membership of the echo chamber is based (that such exclusion is intentional rather than accidental marks the difference between an echo chamber and an ‘epistemic bubble’). Alternatively, perhaps some of these perspectives do in fact make it through the walls of the echo chamberFootnote ²⁴ but they are nevertheless presented in such a way that fails to convey the true strength of the arguments to those within. For example, suppose Mary reads an article posted by one of her friends in which the strongest arguments against their views are discussed to determine the best way to counter them in the course of public debate. We might imagine that these arguments, selected and presented by someone who is completely unconvinced by them, might be delivered more effectively by a ‘true believer’. Of course, there is no necessary connection between believing in the truth of an argument and the skills needed to make an effective presentation of the argument (it would be bad news for university professors if this were so), but it seems plausible to suppose that there is a risk that one will not be exposed to the best forms of arguments if one only ever hears those arguments articulated by people who aren’t convinced by them (and especially in cases where they are articulated by people who find the conclusions not merely implausible but deeply morally objectionable).

Finally, we can imagine cases where dissenting views are accessible and where they are transmitted via those who sincerely endorse those views, but where the proponents of these views have been pre-emptively discredited. Perhaps one of Mary’s friends, spurred by a sense of outrage and incredulity, shares an article authored by someone who is critical of one of the members of her preferred political party, but before she can even click on it her eyes are first drawn to her friend’s denigration of the article, and a slew of scornful comments by other residents of her echo chamber. Even if Mary decides to bother reading the article at this point, she will have been primed to reject it. Each of these examples tracks ways in which access to diverse viewpoints can be restricted, with the end result in each case being that those within the echo chamber are less likely to hold true beliefs as a result.

In addition to helping us find the truth, exposure to diverse viewpoints is thought to promote good epistemic practices by ensuring that we are able to defend our views rather than merely assert them without reflection such that they become Millian ‘dead dogmas’.Footnote ²⁵ We may think that it is not enough to simply believe things that are true if we are unable to support those beliefs with good arguments. Thus echo chambers can serve to undermine our role as good epistemic agents even if our beliefs are entirely correct and held with an appropriate degree of confidence.

2.2. Epistemic agency

This point is emphasised and developed by Christopher Ranalli and Finlay Malcom, who suggest that echo chambers are bad for us because of their ‘reasons-undermining features’ (Reference Sheeks2023, p.1) even in cases where an echo chamber tends to cultivate true beliefs among those within. They argue (plausibly, in our view) that being a good epistemic agent is about more than just holding true beliefs and (less plausibly, in our view) that echo chambers are always bad for us because they inhibit our ability to strengthen and diversify the reasons that support our true beliefs. One key insight from Ranalli and Malcom’s view is their criticism of what they call a ‘broadly veritistic’ (Reference Sheeks2023, p. 9) accounts of echo chambers, according to which the goodness or badness or a particular echo chamber depends on whether it tends to lead those within to hold true beliefs. They identify Jennifer Lackey’s approach as one such account, which they call the ‘Right Inputs View’, according to which the goodness or badness of an echo chamber depends on what kind of information is fed into it: true information results in true beliefs (and good echo chambers), false information results in false beliefs (and bad echo chambers). We agree with Ranalli and Malcom that veritistic accounts fail, since holding true beliefs (while obviously important) is not the only epistemic obligation we must fulfil, and so we cannot declare a particular echo chamber to be (all things considered) epistemically good for us merely because it tends to cultivate true beliefs for those within. Similarly, Meredith Sheeks (Reference Sheeks2023) criticises epistemic bubbles (and by extension, echo chambers) on the grounds that it is never rational to remain inside epistemic bubbles – including those that tend to inculcate true beliefs.

2.3. Complacency

A third similar worry is expressed by Eli Pariser (Reference Pariser2011) and Kenneth Boyd (Reference Boyd2019), who suggest that echo chambers make us intellectually complacent, ‘[removing] from our environment some of the key prompts that make us want to learn’Footnote ²⁶ and encouraging ‘the stagnation of processes of inquiry’.Footnote ²⁷ One interesting feature of this objection is that it applies not only to our pre-existing views but also to those that we might acquire in future (were we to be inclined to seek out perspectives beyond those we find within an echo chamber). Nguyen (Reference Nguyen2021) argues that echo chambers (especially those that promote certain kinds of conspiracy theories) can offer their members enticingly (and overly) simplistic narratives, delivering a kind of ‘thought-terminating’ clarity which causes members to cease their inquiries too early. Such objections might be best understood as worries about the kinds of ‘epistemic vices’Footnote ²⁸ that echo chambers encourageFootnote ²⁹ and suggest that echo chambers may be bad for us (as epistemic agents) even if the beliefs we hold are entirely correct and held with an appropriate degree of confidence.

2.4. Overconfidence

Fourthly, some worry that those in echo chambers ‘will tend to have an unwarranted degree of confidence in their beliefs’.Footnote ³⁰ This may be so if we take the mere fact that a fellow member believes P to be a reason for increasing our confidence in P, without further regard to the evidence on which their belief is based. The basic worry here is that echo chambers tend to restrict the information available to their members without their members realising this (or if they do realise, without responding appropriately), which leads in turn to members granting undue weight to consenting opinions among fellow members as they treat them as if they have come to their conclusions using different or additional sources of information. Boyd develops this worry further, arguing that echo chambers also give rise to a ‘groupstrapping’ problem in which an individual increases her confidence in P once she learns that the group endorses P, causing the group to increase its confidence in P on the basis that one additional person now endorses P.

2.5. Duties of civility

Although the literature focuses on epistemic problems with echo chambers, this isn’t the only problem they’re thought to have. Specifically, in encouraging excessive levels of trust among group members, and excessive levels of mistrust in non-group members, echo chambers may encourage us to violate the duties of civility that we owe to others in virtue of our role as public deliberators. This strand of thought is perhaps most evident in recent work by Nguyen, for whom the manipulation of trust marks the fundamental distinction between echo chambers and the comparatively less serious phenomenon of epistemic bubbles: ‘When you are in a bubble, you don’t hear the other side. When you’re in an echo chamber, you don’t trust the other side. Echo chambers don’t cut off lines of communication from the outside world; rather, they isolate their members by manipulating their members’ trust’.Footnote ³¹

In Nguyen’s case, this focus on trust is primarily relevant insofar as trust relates to credence and broader debates about the role of trust in social epistemology. However, it seems to us that questions of trust also carry significant implications for our duties of civility toward one another as public deliberators. For example, public reason liberals hold that we are sometimes required to engage with those who hold diverse viewpoints to our own (so that we can provide them with the public reasons that justify coercive political policies we support) and engage with them in ways that presuppose good faith and a degree of reasonableness and rationality, by default.Footnote ³² In encouraging us to mistrust non-group members, we are encouraged to regard such people as unwilling or unable to engage with us in this kind of good faith, reasonable dialogue. In addition to our duties toward individual co-deliberators, mistrust may undermine our democratic institutions more generally. According to Sunstein, ‘Democracy does best with what James Madison called a “yielding and accommodating spirit”, and that spirit is at risk whenever people sort themselves into enclaves in which their own views and commitments are constantly reaffirmed’.Footnote ³³ Echo chambers may thus undermine our role as good public deliberators and/or our democratic institutions, in addition to undermining our role as good epistemic agents.

3.1. The chamber

We argued above that echo chambers can occur outside blogs or spaces like online or televised media, they can exist in physical spaces and communities as well. So what does that mean for what can count as the social/epistemic groups that make a core part of the definition of an echo chamber?

Let’s think about what groups we could talk about on a very general level first. It’s likely that we can describe a person as being in any large number of groups. Radko might be a chemistry student, and we could describe him as being a member of his class (perhaps a different structure for each of his modules, with a different makeup of other students), his year group, or his degree programme. He’s a member of the student body, and he might be a member of several different societies. He’ll be in a number of different chat groups in a variety of messaging services, and a number of groups on social media or other websites. There are groups of people he works with, lives with, eats with, and socialises with. We could class him as being a member of the group of people who all watch The Simpsons after dinner on a weeknight, people who walk past a certain library each day, people who are all friends and acquaintances of a person he works with called Wolfgang.

But just because something is a group, that doesn’t mean it’s a social structure, nor an epistemic structure. Presumably, to be in one of these, they must be grouped in some way that’s not just arbitrary, but that has an element of social interaction in the former case, or passing on of information in the latter case. It wouldn’t be accurate to say that Radko is in a social/epistemic structure based on people whose names begin with R, for example.

However, despite this limitation, we think that the rest of the examples of groups listed above could still count. Some of them more obviously so – Radko socialises to an extent with the various student groups and classes, and information is regularly shared in those circles as well. But even if we think about a group of people who watch a certain show – these people will share that show as a source of knowledge and values. They’ll form beliefs based on the show, the adverts that come up in the break. The show might even influence how they come to understand other things in their lives. The same might be true for friends of Wolfgang – they’ll share a source of information that informs their other beliefs. This is even the case for groups like ‘people who walk past the library’ – who will see the same building, on the same street, and be shown the same fliers on the side of the building.

There are a number of ways that we could count ourselves as social or epistemic groups, but there’s still one worry we’d like to tackle straight away. Are we already being too inclusive with our definition, given the examples of groups we’ve listed? After all, some of these groups are not particularly social, and in some of them, the flow of information only goes one way. Can a number of people who watch the Simpsons at the same time any day really count as the sort of epistemic or social group that philosophers mean when they aim to give the first component of a definition of echo chambers?

We would argue that yes, these should count, for two reasons. Firstly, because this mirrors how some of the paradigmatic cases of echo chambers work. The social/epistemic structures of the right-wing echo chambers of news stations or radio shows by charismatic personalities like Rush Limbaugh, for example, have similar lopsided directions for the flow of information. Secondly, we take this breadth to be true to the definition because groups being structured in that way is adequate to meet the rest of the needs of the definition. As long as information is flowing towards you, it doesn’t always matter that you’re not interacting much with it yourself.

As we go through each of the parts of the definition, we’ll have a look at three different cases and show how each of these come under the definition of echo chambers. Hopefully, they should each demonstrate a different kind of case: echo chambers that are epistemically important, echo chambers that are morally important, and echo chambers that are fairly mundane.

• UNIVERSITY: One of the social/epistemic groups that Radko is in is his chemistry department at university. It’s a group of all of the members of his course and his lecturers. Although not everyone will interact with everyone, there’s a lot of socialisation and interaction between a lot of members of the group, and they all share sources of information such as the information given in the courses and discussed in class.

• LGBT: Mimi is a member of their local LGBT group. They meet up weekly and all socialise with each other. They have a reading group and share a number of recipes regularly.

• CORNER SHOP: Anna is one of the people who shops at her local corner shop. One of the epistemic/social structures she’s in is the group of people who regularly shop here.

So far this is just the first, and perhaps the simplest, part of the definition of echo chambers that we aim to look at. Next, we will narrow our scope a little more, and look not just at social/epistemic groups, but at ones which share at least one core belief.

3.2. The voice

To get closer to understanding how widely the concept of echo chambers applies, we need to start being more specific about what form the epistemic/social structures take. The next move, then, is to demonstrate that many of these structures exist with closed membership and around core beliefs.

Remember what these criteria looked like to begin with. In many of the paradigmatic cases, the shared belief is something political and usually something false. But when we look beyond those parameters, it seems like the existence of a shared belief doesn’t actually do much to narrow down the range of things that count for our definition so far. After all, just about every group of people will share a number of important key beliefs that are the foundations for our actions, our communications, our shared projects, etc. These might be beliefs about the value of shared hobbies, for example, or simple facts about the flatness of the earth or the way the world is laid out.

Let’s take some examples of core beliefs that might underpin the three cases I listed above.

• UNIVERSITY: The members of Radko’s chemistry department will share a number of core beliefs that they need to believe in order to be members of this group. This will include beliefs such as positive beliefs about the value of science and education, the value of studying in the particular city of their university, and a number of beliefs about what chemistry is and what some of its basic rules are.

• LGBT: Many of the core beliefs of Mimi’s local LGBT club are going to be important moral and political beliefs. Because of the membership being made of members of marginalised communities, the group will centre around beliefs about things like justice, freedom, and equality.

• CORNER SHOP: Even a relatively uninteresting group like Anna’s will have a number of shared beliefs that are still necessary for anyone to become a member of that group. For example, beliefs about the ability to exchange money for goods, or the fact that the building is, indeed, a corner shop, or that gravity will allow us to walk around to get there.

It might be the case that we underestimate how many social and epistemic groups we are in that have core beliefs as a requisite part of membership because firstly we don’t tend to think of these groups at all unless we have some reason to, and we don’t tend to think too much about the sheer number of core beliefs that we all share in order to get by most of the time. After all, it’s often only when a belief is questioned when it becomes interesting to discuss – maybe why cases like the roundness of the earth is currently an interesting thing to think about.

Before moving on we’ll make one further brief point in defence of this way of understanding the criterion of a core shared belief. It might be a concern to some that many of the beliefs are underlying, implicit, or even unconscious. But once again we think this is in line with the paradigmatic cases of echo chambers. Many of the problematic right-wing or cult-like echo chambers will also have their core beliefs and values be implicitly held rather than explicitly so, sometimes even as a tool for plausible deniability to others, or to avoid their being questioned (take dog-whistling as a form of racism, for example).

3.3. The echoing

The third criterion for a definition of echo chambers was the way the echoes can reverberate around the chamber – that is, the way that the core beliefs can get repeated and reinforced by the structure and its members. So how does this work for many of the core beliefs that are discussed in the cases above? Particularly when many of those beliefs can be implicitly shared, rather than spoken about explicitly.

There are a number of ways that core beliefs can be reinforced beyond something as obvious as explicitly uttering the beliefs out loud. In the social media example, the beliefs were upheld by way of increased sharing of other news, commentary, or information which, if true, supported those core beliefs. This will also happen in the cases we’ve been discussing – for any of those shared beliefs, the other things that members of the group do or say will support and reinforce those core beliefs in virtue of being underpinned by them. The phenomenon of people being more likely to share information that supports their underlying core beliefs in that way will happen even when the core beliefs aren’t wrong, or politically charged. We still have biases towards things we believe, for example. Let’s look through the cases from above.

• UNIVERSITY: The core beliefs that Radko’s course shares about the underpinnings of chemistry and the value of science will be repeated just via most of the other ways that the people on the course will share and talk about the science that they do, the experiments that they run, the chemistry facts they learn.

• LGBT: Members of Mimi’s LGBT group will reinforce and repeat beliefs about the value of equality through a lot of their activities. Any discussions around politics will likely make use of this underlying assumption. Furthermore, just as in the paradigmatic bad cases, the group will be more likely to share and discuss news stories or opinions that support those beliefs.

• CORNER SHOP: Because most of the core beliefs listed here are very ordinary, but also staples for some very basic behaviours, the core beliefs of the members of the corner shop will be reinforced through everyday actions. Simply purchasing something from the shop will reinforce beliefs about the workings of capitalism, for example. Walking into the shop will reinforce the beliefs about how gravity works, etc.

3.4. The manipulation

The final criterion presents the biggest problem for our account. Since echo chambers are inherently manipulative, on the face of it it seems like this feature will be difficult to identify in the kinds of benign and mundane cases we outlined above. Nonetheless, we aim to show that manipulation is in fact incredibly widespread. The reason for this, we think, is the imperfect nature of our epistemic abilities. In a world in which humans had the capacity for infinite and only fully rational reasoning, then we might not ever benefit from echo chambers. But we are biased and finite beings, who depend upon those biases in order to gain knowledge.

Before we look at some of the ways in which this kind of manipulation can be wide-ranging, let’s think back to what sorts of examples of manipulation there were in the paradigmatic cases. We have mockery and humour, opponents being described as bad, and stories and explanations asserting that one’s opponents are either manipulating themselves or have been manipulated, in order to think what they think. However, this is structurally the same sort of thing as we do in good and neutral cases. We will sometimes mock outsiders to groups for espousing beliefs that contradict our core beliefs – whether that’s because we find them so ridiculous or out of a form of solidarity with other members of that group. Other times we’ll offer explanations for why those people hold the views that are so different from our own – pointing out that they stand to gain from saying those things in other ways, or that they’ve been tricked by others. Sometimes we’ll even accuse people of holding those beliefs because they’re in echo chambers. Let’s look at our examples one more time:

• UNIVERSITY: Some students in Radko’s class might question the core beliefs of the values of science or the underlying assumptions of scientific paradigms. But when they do, they might well be treated in a way which is analogous to people with outsider beliefs in the paradigmatic cases of echo chambers. For example, they might tell each other stories of how those students had been looking up untrustworthy videos online, and that they’d fallen for clickbait content aimed at generating hysteria and ad revenue instead of the truth.

• LGBT: Within this group, anyone who expresses views hostile to the core beliefs of the group is likely to be treated as an unreliable source. Nguyen concedes this point himself in moral cases. He says, ‘the fact that any person or group is in favor of, say, sexual orientation conversion therapy is enough for me to discredit them on any social or moral topics’.Footnote ³⁴

• CORNER SHOP: Suppose when Anna is in the shop one day, Wilhelm walks in and declares that the shop items should be free, and that money isn’t real. The response of anyone else in the shop at the time might well be mockery, particularly when that person is no longer around. Anna might go home to her partner and laugh – ‘you’ll never guess what someone said in the shop today!’. That mockery will serve to reinforce the core beliefs about the way that money and the shops work.

Although we have focused on only three main examples as we go through these criteria, we aimed to show something different with each. As such, we have shown that echo chambers are far more wide-ranging than the paradigmatic cases would suggest.

Of course, just because echo chambers are wide-ranging, that doesn’t do enough to show that they’re not, in themselves, epistemically or morally bad. And there are certainly some cases in which they are bad – in which the various features of the structures cause significant harm to people in the way that they prevent the truth from being seen. We only need to look as far as the paradigmatic cases in the literature to see that much. But we hope to have shown through the above examples that it’s not that these structures themselves are inherently the problem. It’s not just that echo chambers are everywhere and it’s impossible to get any fresh air, it’s also that we need echo chambers. We wouldn’t be able to function properly if we stopped to question every belief, every shared belief, and every assumption. We need to be able to take some things for granted – and in order to do that, not only do we need to have social and epistemic groups and shared beliefs, but sometimes we also need epistemic structures that filter out information and views that run contrary to these. Otherwise, we’d simply never get anything done.

It is worth dwelling on this point since it represents one of the chief benefits of echo chambers. It is true, of course, that advances in knowledge may come when orthodox positions are challenged and overturned, but building on any advance in knowledge becomes extremely difficult if every new conjecture must be preceded by an argument from first principles. Similarly, when communities engage in decision-making procedures it is obviously important that those procedures themselves can be challenged, but if every decision were to be preceded by deliberation about the legitimacy of the decision-making procedures themselves, things would very quickly grind to a halt. In practice, the way we deal with these kinds of cases is to compartmentalise – some scientists work with given theoretical models, while others theorise alternatives, for example. In politics, national discourse provides space to argue about principles of political legitimacy or voting laws, while local councillors take these things for granted in order to be able to do their jobs effectively. In such cases problems may arise not due to the existence of an echo chamber as such, but precisely in those cases where there aren’t enough echo chambers to go around and thus not enough space to provide for specialised deliberative communities.

To underline the point, consider the ‘Simple Sabotage Field Manual’ – a document produced in 1944 by the Office of Strategic Services (a predecessor of the CIA) which was declassified around 2008. One section of the document offers advice to would-be saboteurs on ‘general interference with organisations and production’. One is advised to ‘never permit short-cuts to be taken to expedite decisions’, ‘talk as frequently as possible and at great length’, ‘refer all matters to committees “for further study and consideration”’, ‘haggle over precise wordings of communications, minutes, resolutions’, and to ‘refer back to matters decided upon at the last meeting and attempt to re-open the question of the advisability of that decision’. One might well imagine a modern saboteur (wittingly or unwittingly) engaging in the same kinds of strategies on the basis that it is important not to become trapped in an echo chamber.

Attending to the positive features of echo chambers shows that members of an echo chamber need not be irrational, and indeed it may sometimes be irrational to try to escape or dismantle an echo chamber where doing so will likely leave one in a worse epistemic position overall. Members simply need to be finite beings, who can only do so much research, think through so many lines of argument, and hold so many different ideas in their heads at once. This also helps to explain why fully rational and epistemically virtuous individualsFootnote ³⁵ may nevertheless find themselves in echo chambers that are epistemically bad for them – even when we behave as we ought to (as both moral and epistemic agents) we may nevertheless make mistakes or be led astray by third parties through no fault of our own.Footnote ³⁶

Echo chambers, then, are not bad in themselves. In that way, they’re similar to a lot of other things that can be good or bad in different contexts. Rhetoric, for example, can be used to persuade people round to a good point of view or a bad one. Trust can be used for good or misused for evil. And whether structures (or tools) like these end up being good, bad, or neutral, depend not on the structures themselves but on further facts about the people or information in question.

Finally, it is worth returning to assess Ranalli and Malcom’s view in the light of the analysis above. Recall that they defend the claim that echo chambers are always (epistemically) bad for us because they ‘have a social-epistemic structure which systematically pre-empts its members from critically engaging with serious objections’ (Reference Sheeks2023, p.24). However, if echo chambers are as ubiquitous (and often trivial) as we have suggested, this highlights the inadequacy of the reasons-undermining view as a basis for action-guiding recommendations. Note first that the reasons-undermining view makes no distinction between relatively trivial and non-trivial forms of knowledge. While Ranalli and Malcom suggest that one can be in a better or worse epistemic position depending on how deeply embedded one is within an echo chamber (Reference Sheeks2023, p.26), the actual content of these beliefs apparently plays no part in this determination, and all of the examples they rely upon involve forms of knowledge that seem quite important (e.g. facts about climate change or political conspiracy theories). However, some beliefs in addition to being trivial are comparatively isolated: someone deeply entrenched in an echo chamber who believes that Nintendo makes better products than Sony is unlikely to be in a worse epistemic position overall than someone deeply entrenched in an echo chamber who believes in some vast global conspiracy theory.

A second problem for the reasons undermining view as an action-guiding view is that it offers no indication as to how much we are obliged to develop our epistemic capacities. Ranalli and Malcom (Reference Ranalli and Malcom2023, pp. 36–37) explicitly address this when they discuss the opportunity costs associated with improving one’s epistemic position, and their response (which also serves as a response to our worry about triviality) is to insist that theirs is not a normative thesis in the sense that it does not recommend particular courses of action for particular kinds of epistemic agents. Rather, their aim is only to identify how echo chambers are epistemically bad for us, not to tell us what we ought to do about this. They point to an asymmetry between someone within an echo chamber and someone without – the latter is epistemically better because she has opportunities to develop her capacities that the former lacks, even if she chooses (and even if she is justified in choosing) not to do so. We believe a similar line of response can be taken to other non-veritistic accounts such as those of Sheeks – while we accept Sheeks’ argument that it is not possible to know whether one is in a ‘good’ epistemic bubble without first exiting that bubble (at least temporarily), we reject the premise that being a good epistemic agent necessarily requires one to have justified beliefs about the goodness of each and every epistemic bubble that one might inhabit.

If our account is correct, however, then any person standing outside one echo chamber is likely to reside within others, and we simply cannot say without more information whether they enjoy a wider range of epistemic opportunities compared to anyone else. Ranalli and Malcom’s view depends on the assumption that it is possible to place oneself outside of any echo chamber, but if this is mistaken then the choice we face is not whether to extract ourselves from all echo chambers, but which echo chambers we are comfortable remaining within, and which we should avoid or escape.