Epistemic Bystander

Abstract

Epistemic bystanding occurs when an agent has all the competences, knowledge and opportunity to prevent another person from forming a false or risky belief, but does not prevent the belief formation. I provide a definition of an epistemic bystander and explain the mechanism that makes someone an epistemic bystander. I argue that the phenomenon is genuinely epistemic and not merely linguistic. Moreover, I propose an account of the mechanism of epistemic bystanding building on Ishani Maitra’s notion of licensing. An epistemic bystander licenses a risky belief-forming process in another person and thereby performs a blameworthy epistemic action. This form of licensing explains the distinctive wrong of being an epistemic bystander.

1 Introduction

Suppose you are at a dinner party with some friends. At some point, the topic of discussion comes to books people currently read and the following conversation takes place between your friends Linh and Martin:

• Linh: I’ve been reading Cloud Atlas last week. It’s by David Mitchell.

• Martin: David Mitchell? The comedian? I didn’t know he’s a fiction author.

• Linh: Neither did I until I saw the book cover with the name.¹

You are standing next to them, following the conversation. You know that the author David Mitchell is not the comedian David Mitchell; they just happen to have the same name. Nevertheless, you don’t say anything. At the end of the day, Martin falsely believes comedian David Mitchell wrote Cloud Atlas. You could have prevented that from happening. You could have easily intervened and ensured that Martin did not form a false belief. There is a clear intuition that you did something wrong—at least you did not fulfil some expectations as a friend. But there is also a sense in which you committed an epistemic wrong. You are to some degree responsible for Martin’s false belief in virtue of being an epistemic bystander—you have all the competences, knowledge and opportunity to prevent another person from forming a false belief, but you do not intervene.²

I aim to explain the mechanism that makes you an epistemic bystander and to understand the epistemic wrong of being a bystander. The plan is the following: I start with a rough first pass on epistemic bystanding. I then proceed with my analysis with a closer look at the phenomenon and discuss why this is genuinely epistemic and not merely a linguistic phenomenon in part 2. In part 3, I propose an account of the mechanism of epistemic bystanding by building on Ishani Maitra’s (2012) notion of licensing. I discuss licensing in epistemic bystanding and show how it can tell us more about the distinctive epistemic wrong in epistemic bystanding. Finally, I sketch why interventions required to avoid bystanding are not a form of epistemic paternalism.

2 Epistemic Bystander—A First Pass

I introduced the idea of an epistemic bystander with a conversation at a dinner party. Call this case DINNER PARTY. It is a simple situation of a person having all the competences, knowledge and opportunity to prevent another person from forming a false belief, but not acting accordingly. Clearly, in this scenario, you could easily intervene and tell both Linh and Martin that they mix up different David Mitchells. You know that there are two different relevant David Mitchells, and you know who is the fiction author and who is the comedian. Perhaps you heard the comedian comment on sharing a name with the fiction author in an episode of Would I Lie To You and searched the internet out of interest afterwards. You certainly also have the opportunity and competence to make an assertion at a dinner party between friends. I can just stipulate that you are fully aware of the conversation and that there is no pressure on you to keep quiet. I will bracket whether more realistic cases of epistemic bystanding are exactly like this for now, but come back to that later. All I need at this point to illustrate the general phenomenon is a situation that fits the bill.

DINNER PARTY strikes me as a case in which you are doing something wrong by not intervening. It is to some degree on you that Martin forms a false belief. And, given that false beliefs seem to be bad most of the time,³ you are therefore responsible for something bad happening to Martin. You are responsible for a harm being done to Martin. At this point, I will leave it open what exactly this harm is. For now, it is enough to provide an intuitive sense of the wrong of your action in DINNER PARTY. And this is easily reasoned for: Martin has every right to complain about your behaviour as soon as he finds out that you knew that Linh’s assertion was wrong. Linh did not know better, but you knew all along. Why did you not tell him? This is something that Martin will identify as a wrong towards him.⁴ You were not a good friend to Martin. You did not uphold the norms set within the conversation and now you lost some of Martin’s trust. Your bystanding in DINNER PARTY comes with disappointment and resentment on Martin’s end as soon as he notices that you knew better.

One might immediately worry here that I am overgeneralising. Sometimes staying quiet might be beneficial! Suppose Maggie tells Carlos that she will not see Carlos again in the next 2 weeks. But I know of a surprise birthday party for Carlos that Maggie will be a part of tomorrow. Intuitively, I should not intervene here. I am an epistemic bystander. But am I blameworthy? I am doing the right thing after all, so why should I be blamed? Carlos most likely will not be angry with me afterwards, so it might look like a case of epistemic bystanding without a wrong. However, I do not think this is the only way to interpret this case. An alternative is to suggest that there still has been some harm involved, but it is excused or compensated in the circumstances. This is akin to how a violation of a norm of assertion might be excused in some circumstances (cf. Hawthorne & Stanley 2008, Williamson, 2000, 2005). One reason to go for the excuse reading of the birthday case is that Carlos will still have a reactive attitude towards my bystanding. He will acknowledge that I did not say anything for a good reason, but my bystanding nevertheless will have a different status than an ordinary conversation. Carlos will feel tricked. Tricked for a good purpose, but tricked nevertheless. Even bystanding for a reason leaves a bitter taste.

As a first pass, epistemic bystanding appears to be a behaviour in which one’s failure to intervene leads to someone else forming a false belief. And in being an epistemic bystander, one violates a norm and thereby harms the person forming the false belief. I aim to develop this first pass in two steps that provide the desiderata for an account of epistemic bystanding:

• 1. I want to find a definition that picks out the right cases of epistemic bystanding. DINNER PARTY fits and is easy to grasp, but cannot stand alone as a paradigmatic example. Hence, my first goal is to find a definition with the right extension.

• 2. I look at the mechanism of epistemic bystanding. What does an epistemic bystander do exactly? What behaviour is essential for an epistemic bystander? Here, the aim is to identify the action with which the bystander is harming someone else. I want to explain the distinctive wrong of epistemic bystanding in virtue of what a bystander does when they do not speak up in an epistemically risky situation.

At the end, I will be able to identify cases of epistemic bystanding, explain which action of the bystander exactly is blameworthy, and in what sense the action is blameworthy.

3 What Is an Epistemic Bystander Exactly?

Let me begin by revisiting DINNER PARTY and my label of ‘epistemic bystanding’ as a starting point for a better definition of an epistemic bystander. This case on its own will not do to establish a genuine epistemic phenomenon. It is a case centred on a conversation.⁵ Therefore, one might suspect that this is just a philosophy of language question. Perhaps what is happening here can be fully captured by the tools of analysing how assertions work. Both of the two big picture frameworks of analysing assertions in conversations seem promising: Lewis’s (1979) scorekeeping in a language game and Stalnaker’s (1999) analysis of assertions as (proposed) changes to a common ground. Let’s assume Stalnaker’s common ground framework. The same kind of argument also applies to Lewis.

Let me look at Linh’s assertion that comedian David Mitchell is a fiction author and wrote Cloud Atlas. In Stalnaker’s framework, this can be modelled roughly as a proposal of adding the proposition ‘Comedian David Mitchell is a fiction author and wrote Cloud Atlas’ to the common ground. Martin accepts this proposal—perhaps just with a small nod. You know that the proposition that Linh wants to add to the common ground is false. But you keep quiet and do not explicitly reject it. Because you do not reject Linh’s assertion and Martin indicates acceptance, the proposition is added to the common ground of this conversation. The context sets change accordingly—now ruling out all the possible worlds in which comedian David Mitchell did not write Cloud Atlas. Unfortunately, that includes our actual world. You know that, and your acceptance of the proposition was just for the sake of the conversation. But Martin ends up with a false belief. He ends up with a false belief because you did not reject Linh’s assertion.

This reconstruction of the case in terms of Stalnaker’s notion of assertion and common ground looks promising. However, the phenomenon of epistemic bystanding is wider than purely conversational cases.⁶ You can be an epistemic bystander when no person says a single word. These cases indicate that the epistemic phenomenon cannot be fully captured with the tools used to analyse conversations. Consider the following case:

LIGHTING Aimee and you look at a table, aware of each other’s presence. Aimee judges that it is red. You, however, know that the lighting conditions make it impossible to tell whether the table is red, or is just under a red light. You do not tell Aimee, and she ends up with a false belief. You could have easily intervened, but you did not.

In LIGHTING, no conversation takes place. Aimee just forms a belief through visual perception. Unfortunately, she is in an epistemically bad situation in which her perception is unreliable. Moreover, she is unaware of this environmental⁷ condition. You, however, know perfectly well that one cannot trust one’s eyes in this room. Perhaps you got fooled before and now know better. You could warn Aimee, but you do not. Intuitively, you are committing an epistemic wrong in LIGHTING just like in the conversation about David Mitchell. You have all the competences, knowledge and opportunity to prevent another person from forming a false⁸ belief, but you do not intervene. You are an epistemic bystander, without any conversation taking place. This is bad news for an explanation in terms of assertions and common grounds. And similar cases are very easy to find.

Cases like these can also bring pragmatic considerations into play when we consider our intuitions. If I know that Aimee cares a lot about finding a red table, then being a bystander seems more blameworthy, than in cases in which I do not know anything about Aimee’s interests in tables. Here, in addition to the blameworthiness of being an epistemic bystander, I am blameworthy for not aiding Aimee in her practical goals. However, LIGHTING points to an epistemic bystanding that is blameworthy regardless of practical goals that suffer due to my bystanding. Hence, the story is told in a minimalistic way. Nevertheless, my access to Aimee’s goals and intentions will affect whether I have the required competences, knowledge and opportunity to prevent her from forming a false belief.

Let me consider one final case:

MÜLLER-LYER Peter looks for the first time at Müller-Lyer arrows. He has no information about these arrows and judges that one line is shorter than the other. You are standing next to Peter and he is aware of you next to him. You know of the Müller-Lyer illusion – you know the lines are exactly the same length. Moreover, you know that Peter has never heard of the illusion. Nevertheless, you do not speak up and Peter ends up with a false belief. You could have easily intervened but did not.

Just like in LIGHTING in MÜLLER-LYER, no conversation takes place (Fig. 1). Nevertheless, Peter forms a false belief that you could have easily prevented by intervening. You have the required knowledge, being acquainted with the Müller-Lyer illusion. You also have all the necessary competences and opportunity to tell Peter not to trust his eyes. Clearly, you are an epistemic bystander in a situation that cannot be captured by a theory of assertion. And these cases are not restricted to testimony and perception either. The same structure works for any source of knowledge. A person can form a belief based on memory, but you know that they are in a specific environment that leads to false, apparent memories. Or a person can reason from one belief to another belief, while you know that they are under the effect of some anaesthetic and are currently unable to reason properly. Granted, in the last case, it is not completely clear whether they would respond to an intervention of yours. Nevertheless, the structure of these cases is clearly not reliant on any conversation taking place.

Fig. 1

Müller-Lyer arrows (Taken from https://commons.wikimedia.org/wiki/File:M%C3%BCller-Lyer_illusion.svg (October 4th 2023))

Moreover, the focus on non-linguistic cases also distinguishes the problem of epistemic bystanding from Lackey’s (2020a) work on the duty to object. In that work, she explores whether there is a duty for me to respond to your asserting something that I take to be false. In particular, she argues that such a duty not only exists, but differs from other common duties, such as the epistemic duty to believe in accordance with the evidence one has, or the moral duty to keep promises. The former is only intrapersonal. It applies only to me and my beliefs. The latter is interpersonal, but does not concern beliefs. The duty to object has the special nature of being both interpersonal and concerned with beliefs and belief formation. This unique nature motivates the inquiry into the duty to object. However, that duty to object is only one (although paradigmatic) form of an interpersonal epistemic duty.

I see my account of epistemic bystanding as a further step towards understanding such duties in responding to certain epistemic actions of other agents more generally. Understanding epistemic bystanding helps to get a better idea how duties to intervene in risky belief-forming processes function. As such, my discussion concerns a subset of cases that involve what Lackey (2020b) calls general interpersonal epistemic duties. I suggest that we understand epistemic bystanding along the following condition:

Epistemic Bystander_{First Pass}: An epistemic agent A is an epistemic bystander iff A has all the required opportunity, competences, and knowledge⁹ to prevent another epistemic agent B from forming a false belief, but fails to do so.

DINNER PARTY, LIGHTING and MÜLLER-LYER all fit this condition. In each case, an epistemic agent has all the required competences, knowledge and opportunity to prevent another person from forming a false belief but fails to intervene. I am going to unpack the parts of the definition a little further. My aim here is to make clear when someone is an epistemic bystander by focusing on the different requirements that an epistemic agent has to satisfy before a norm demands the agent to intervene in another agent’s belief formation.

The easiest part to spell out is the opportunity condition. This condition captures that A is free to intervene in B’s belief formation. No outside factors prevent A from providing B with the information that they should not form the belief in question. Moreover, no factors in A’s psychology make such an intervention impossible. The opportunity condition thereby rules out cases in which agents are, for instance, physically unable to talk, coerced not to intervene and afraid of speaking up. ‘Opportunity’ is just a grab bag for anything that could prevent a person from informing someone else about the unreliability of a token belief formation.

I have presented ‘opportunity’ as binary. Either you have the opportunity or you do not. Perhaps this is not quite right. In real life, opportunities are more fine-grained. For instance, a situation in which there is pressure not to speak up because of risky consequences might not qualify as having no opportunity at all to speak up. But it also cannot qualify as having an opportunity in the full sense, because speaking up is hindered by outside influence. Marginalised and oppressed groups are often in situations in which speaking up to prevent someone from forming a false belief about their group comes with risks of being talked down or attacked. Are they epistemic bystanders? Or do they lack the required opportunity? I take it that these cases are not always clear-cut. However, there are at least some considerations that can help as indicators for situations in which marginalised or oppressed agents are not epistemic bystanders, even though they do not intervene. I provide three examples:

First, one can look at the potential threats to speaking up. Threats can vary and range from physical violence to exclusion from communal activities and loss of social status. Those threats might not always be made explicit, but their presence changes the actions of the oppressed. For instance, they might lead to testimonial smothering (Dotson, 2011), a form of self-censorship to avoid harmful consequences for speaking up.

Potential threats for speaking up are often difficult to recognise for people outside of the oppressed group—partially because outsiders lack the relevant experience, but also partially because of a wilful ignorance that might be established in a community (see for instance Mills (2007) on white ignorance). Hence, one needs to be cautious in identifying epistemic bystanders. One ought not to ignore excusing factors that prevent agents from intervening in faulty belief formation.

Second, one can look more closely into whether the situation involves a new belief formation at all. For instance, in a racist society, some racist beliefs will be already widely held. Refraining from trying to change an already-held belief does not make an agent an epistemic bystander. Bystanding, as I am interested in, is only concerned with the formation of beliefs. Relatedly, even if a person is about to form a new false belief individually, if that belief is already widely held in the community, it seems perfectly reasonable not to intervene. One can only be an epistemic bystander if one recognises a new, false belief being formed. In a racist community, it will be difficult for the oppressed (and the non-racist generally) to identify whether someone forms a new, racist belief or whether they already have that belief. In such a community, it will often be reasonable to assume that another person already has the racist belief in question. As soon as this assumption is justified, refraining from intervention does not turn an agent into an epistemic bystander. Here, knowledge of the larger community provides a legitimate reason not to intervene.

Third, and related, one might have knowledge of the larger community that they will not be responsive to an intervention. Suppose a Black woman growing up in a community learns how wilfully ignorant the larger epistemic community around her is when it comes to issues of racism (see the aforementioned white ignorance proposed by Mills (2007)). The unresponsiveness to intervention can even indicate that the racist is not aware of the oppressed group as potential knowers and testifiers.¹⁰ If that is the case, then there does not seem to be an opportunity for intervention.

It is relatively easy to characterise the opportunity condition with cases in which it is clearly satisfied and cases in which it is clearly not satisfied. But for some in-between cases, I do not have a substantial answer. These have to be looked at case by case. Hence, I retreat to merely defining epistemic bystander with reference to the ‘required […] opportunity’ and examples for considerations that can tell us whether an intervention is called for. This is the best I can offer for now and should still be sufficient to explain what kind of opportunity I have in mind.

The notion of competence captures all cognitive abilities required to prevent someone from forming a false belief. An agent needs to be able to identify a situation as a candidate for intervention. This is an exercise of mind-reading abilities in combination with contextual cues. The epistemic bystander is only a bystander if they recognise that another epistemic agent is about to form a false belief—or at least is about to form a belief in an epistemically risky way. One cannot be an epistemic bystander without any awareness of any belief formation in any other person taking place. Think back to DINNER PARTY. If Linh tells Martin that comedian David Mitchell wrote Cloud Atlas and I was at that moment fully focused on a piece of lemon tart and did not listen to anything, I would not be blamed for not intervening—I did not even know there was anything to intervene. Of course, I might still be blamed for not listening, but that is a different issue. To be an epistemic bystander, I have to be aware of the opportunity to intervene. That goes for bystanding more generally, even outside of the epistemic realm. Hence, bystanders regularly pretend not to notice issues in order to avoid blame for inaction.¹¹

Moreover, the competence condition also captures that the intervention has a reasonable chance of success.¹² If someone is stuck in their ways and will ignore the intervention, then it does not seem that an intervention is called for. This also explains why, for instance, in cases of very strong political partisanship intervention might not be required, because it cannot succeed in preventing false or risky belief formation. Competence is here considered in an externalist fashion. What matters is whether someone has the competence, not whether they think they have it. However, a belief about one’s own competence might excuse me from being a bystander. If I think there is no point in intervening, but there actually is, then I have still done something blameworthy, even though it is understandable why I acted as I did. Similarly, if I take myself to have the competence, I might intervene, even if I will not be a bystander in any case.¹³

The competence condition is important to limit the cases of epistemic bystanding. An agent is not always and all the time required to correct another agent’s belief formations. Just being in the vicinity of a risky belief formation does not make me a blameworthy bystander.

Finally, the knowledge condition captures the epistemic properties an epistemic bystander has to satisfy. An epistemic bystander has to be in an epistemic position such that their intervention prevents the formation of a false or risky belief. But that cannot be the end of the story. Not every intervention that prevents the formation of a false belief is epistemically praiseworthy. One can prevent belief formation for all kinds of reasons. I can cut your connection to the Internet and thereby prevent you from forming beliefs based on Twitter sources. I did not care about that at all. I just cut the cable out of spite. Perhaps I thereby prevent you from forming false beliefs if you would have looked at unreliable sources otherwise. But this would just be a case of luck. Nothing in my action of cutting the cable was epistemically praiseworthy.

Moreover, if I am in a bad epistemic position myself, I cannot be blamed for not intervening in the belief formation of another person. If I do not know that there exists more than one David Mitchell, I cannot be blamed for not rejecting Linh’s assertion at the dinner party. On the contrary, it would be very inappropriate for me to intervene without any good reason—not even a suspicion that Linh’s assertion might be false. Hence, an epistemic bystander can only be a person who is in a better epistemic position than the person who is about to form a false belief. In my definition, I use ‘knowledge’ as a condition to account for this better epistemic position. Knowledge here is meant to capture knowledge of something that makes the token belief formation epistemically risky. Either because the bystander knows that the belief would be false or because the bystander knows that there is a significant risk of the belief turning out to be false. This is deliberately broad and captures DINNER PARTY (in which the bystander knows that there are two David Mitchells), the lighting case (in which the bystander knows that the environment makes vision unreliable) and the Müller-Lyer case (in which the bystander knows the lines have the same length and that humans fall victim to a specific sort of illusion).

There are further questions about whether a definition in terms of, say, justified belief or true belief would be better than one in terms of knowledge.¹⁴ Similarly, one could, for instance, propose to give a similar definition based on a lack of an attempt to prevent the belief formation instead. I will bracket these issues in order to have more room to discuss the mechanisms of bystanding. For my purpose, I have now sufficiently discussed what cases I want to pick out when I talk about epistemic bystanding. In the next part, I propose a way to explain the mechanism that makes bystanding blameworthy.

4 The Mechanism of Epistemic Bystanding

To understand the wrong in epistemic bystanding, it is not enough to just find a definition that picks out the right cases. I need an explanation of the mechanisms of epistemic bystanding. What exactly am I culpable of when I do not prevent someone else from forming a false belief, even if I could easily do so? They would have formed false beliefs without me being there anyway, so why does my mere presence matter at all?

One answer has been pointed out by Jennifer Lackey (2020b), who proposes general interpersonal epistemic duties. These duties seem very similar to my proposal for identifying epistemic bystander.

Interpersonal Epistemic Duties: If it is in our power to prevent something epistemically bad from happening through very little effort on our part, we ought, epistemically, to do it. (Lackey, 2020b, p. 278)

However, while this characterisation proposes duties that explain why bystanding is bad, I do not think it fully captures the wrong of epistemic bystanding. Bystanding is an instance of an interpersonal epistemic duty in effect, but it’s not merely the failure to prevent something epistemically bad that constitutes the wrong. Rather, I suggest that bystanding itself can constitute illegitimate support for the epistemically bad thing that occurs. My epistemic bystanding at the dinner is not merely followed by Martin’s false belief, but it actively supports the false belief. To illustrate how this can be the case, I borrow the concept of licensing introduced by Ishani Maitra (2012) in a discussion on hate speech. Licensing a speech act involves a form of bystanding in ordinary linguistic practice that I can build on to understand epistemic bystanding.

Maitra is interested in understanding speech acts that can subordinate without the speaker being in any obvious position of authority. A paradigmatic case is the following:

SUBWAY RIDER An Arab woman is on a subway car crowded with people. An older white man walks up to her, and says, “F***in’ terrorist, go home. We don’t need your kind here.” He continues speaking in this manner to the woman, who doesn’t respond. He speaks loudly enough that everyone else in the subway car hears his words clearly. All other conversations cease. Many of the passengers turn to look at the speaker, but no one interferes. (Maitra, 2012, pp. 100-101)

Maitra wants to explain how this racist speech act can subordinate, without the speaker being in any special position at all. The speaker is just a rider of the subway like all the other passengers. Intuitively, they are equals. Nevertheless, she argues, his speech act can subordinate because the other subway riders grant authority by virtue of not speaking up. They license the speaker to perform racist speech acts. Licensing a speaker and their racist speech act is just a special case of licensing more generally. Maitra illustrates this with the example of an impromptu hike organiser:

HIKE ORGANISER A bunch of friends want to go on a hike on the coming weekend. They begin to discuss the logistics of the hike: where to go, for how long, what to bring, how to get there, and so on. Some in the group express mild preferences for one or another of the available options, but no one expresses strong preferences. The discussion goes on, and on. In fact, it continues for so long that one of the group, Andy, begins to be concerned that nothing will get organised. He decides to take over, and begins to make decisions. He assigns each of the other group members a specific task: one is to pick a location, another is to buy enough food for the group, a third is to find some tents, and so on. No one objects. Everyone completes their tasks, and the hike takes place as Andy planned. (Maitra, 2012, p. 106)

In this case, Andy starts out with no authority at all. But by making decisions with the group following along, he is conferred authority. The mere fact that no one objects and they all complete their tasks is enough for Andy to have authority. Andy’s instructions are licensed speech. Maitra emphasises that it does not matter that no one explicitly agrees. All that licensing requires is that no one objects. Even in case they disagree, but do not voice their disagreement, the speech is licensed (Maitra, 2012, p. 107). And in all cases of licensing, one takes on board some responsibility for whatever is licensed. The quiet subway rider is to some degree responsible for the hate speech they witnessed but did not object to. Of course, as Maitra also mentions, at times, such inaction can be excused (Maitra, 2012, p. 116), but there is nevertheless a default obligation not to license hate speech.

Licensing is a good fit with Stalnaker’s (1999) model of understanding conversations I mentioned earlier. Maitra explicitly builds her account of licensing speech acts on a simplified version of Stalnaker’s proposal (Maitra, 2012, fn.34). But as I showed earlier, Stalnaker is no help for cases that do not involve conversations. Nevertheless, the concept of licensing picks out something important to understand epistemic bystanding. Just like keeping quiet can grant authority to speech, keeping quiet can grant authority to belief formation. At least in some cases, an epistemic bystander licenses an agent to use a particular belief formation process.

Let me illustrate this with Aimee and the LIGHTING case:

LIGHTING Aimee and you look at a table, aware of each other’s presence. Aimee judges that it is red. You, however, know that the lighting conditions make it impossible to tell whether the table is red, or is just under a red light. You do not tell Aimee, and she ends up with a false belief. You could have easily intervened, but you did not.

Aimee starts to form a belief by visual perception. Usually, visual perception is reliable. Under normal conditions—no weird lighting, strange glasses, vision-modifying drugs and so on—visual perception generates true beliefs at a high ratio. If Aimee were to look at a tree under normal conditions, her visual perception would produce the true belief that there is a tree. Hence, one trusts visual perception by default. As long as one has no good reason to be sceptical of the reliability of one’s eyes in a particular situation, one will not question beliefs formed by visual perception. In LIGHTING, Aimee has no reason not to trust her eyes. But you have. You know about the unfortunate lighting condition. Nevertheless, you do not speak up. And by not bringing your reason into the picture you license Aimee’s default belief-forming process. Aimee accepts her vision-based belief in part because you do not speak against doing so. If you had spoken up, she would not have formed the false belief that the table is red. And Aimee is aware of your general possibility to speak up. The fact that you did not plays a role in her confidence in believing that the table is red. It’s not just that you did not stop the false belief from being formed. You are at the same time part of the reason why Aimee takes the belief to be justified. Lackey is right that you do not fulfil your interpersonal epistemic duties, but she does not capture that your wrong goes beyond that. You are licensing Aimee’s epistemically bad belief-forming process. Licensing captures the concrete epistemic wrong that Aimee suffers in a case of epistemic bystanding, and it gives us a descriptive term for what exactly you are doing—instead of merely what you fail to do—when you are an epistemic bystander.

Licensing in Maitra’s case is characterised as a form of granting authority. I have similarly described licensing a belief-forming process as granting it authority to emphasise the analogy of licensing speech to licensing belief-forming processes. However, it is unclear whether it makes any sense to grant authority to a process at all. I do not think this is a big worry. Licensing a belief-forming process can also be understood in terms of a broad notion of trust. Aimee trusts her visual perception that the table is red in part because you do not speak up. If you had spoken up, she would not have trusted her visual perception. Hence, you are at least partially responsible for her false belief.

Theories of trust often distinguish between trust proper and mere reliance (e.g. Baier (1986) or more recently Hawley (2019)).¹⁵ Trusting one’s eyes seems to be mere reliance, even if we tend to speak differently. Trust proper seems to be normatively laden in ways that trusting one’s eyes is not. One cannot blame eyes for being misleading. And neither are eyes praised for being trustworthy when they generate true beliefs. On the other hand, at the dinner party, Martin trusts Linh. He does not merely rely on Linh’s testimony. Licensing seems to lead to both cases of misplaced reliance and cases of misplaced trust in the belief-forming process. Trust in those cases that include other epistemic agents as part of the belief-forming process.

The analogy to Maitra’s notion of licensing is not perfect. Maitra applies the notion to instances in which a person initially has no authority and gains authority via licensing. Licensing a belief-forming process functions differently. Perception in LIGHTING already has some authority, some default trust the person puts into visual perception. But nevertheless, licensing provides a boost to the authority the belief-forming process has. Aimee gives additional weight to the process because of your inaction. She expects you to warn her if something epistemically fishy going on. If there is no warning, she will assume everything to be going fine. She assumes to be in an epistemically friendly environment, even though she is not. And you are licensing that assumption.

There is one further worry about applying the notion of licensing to epistemic bystanding. The notion is developed with conversations in mind in which all participants are aware of agents that could speak up. A person is licensed in their speech because the other interlocutors in the conversation stay quiet. Is there something similar for licensing in epistemic bystanding? Does epistemic licensing require that all relevant agents are aware of each other, as they are in a conversation?

Looking back at my cases of epistemic bystanding I have included a condition of awareness. In my cases, the person about to form a belief is aware of you and your opportunity to speak up. Here, it seems straightforward that inaction can impact the belief-forming process. Aimee is aware of you and therefore expects you to warn her of an unusual, unfriendly epistemic situation.¹⁶ You do not say anything, so she continues her belief formation in an unquestioning matter. Licensing seems to be applicable here without much issue. But what if the person is not aware of your presence and opportunity to intervene? If Peter in MÜLLER-LYER were not aware of anyone next to him, he could not expect any warnings. Hence, the lack thereof should not license his belief formation. The silence of other people who could intervene that one is not aware of cannot boost one’s confidence in a belief. What boosts the confidence is the awareness that someone could easily intervene, but does not.¹⁷ Just like in Maitra’s original licensing cases, authority was granted or boosted by virtue of an awareness of people not intervening even though they easily could.

Importantly, awareness itself might not be enough if it does not come with any sort of expectation that someone would intervene in risky belief formation. Aimee could be aware of you next to her, but if she also believes that you want her to fail epistemically your presence would not license her belief formation. Licensing can be defeated. Nevertheless, licensing captures cases of epistemic bystanding in which an agent fails to intervene well. It explains why not intervening seems especially bad in some cases, and it also explains why not intervening in somewhat similar cases—those without the individuals being aware of one another—does not seem equally bad. Perhaps one also fails to satisfy an epistemic duty in cases without awareness, but there seems to be a clear difference in the badness. When I am far away from Peter, who looks at the Illusion, and I do not intervene, I might do something wrong, but it is difficult to see why I would boost Peter’s confidence in any belief. Only with the awareness condition do I take up responsibility for the additional confidence that other people have about their belief formation. I create a sense of security in them. If that belief formation is risky or leads to false beliefs, that is in part on me. Without me, they would not hold that belief to the same degree. In some scenarios, they might not even hold the belief at all without my non-intervention boosting their confidence. This also fits with our ordinary practice. One often blames other people one is aware of for not providing a warning about inaccurate information or epistemically risky environments. There seems to be good evidence for licensing as an especially blameworthy practice of epistemic bystanding.

I have now shown a way in which one can conceptualise licensing in cases of epistemic bystanding. I can now build a final definition of an epistemic bystander that includes the mechanism licensing as a general requirement:

Epistemic Bystander_Final: An agent A is an epistemic bystander iff A has all the required opportunity, competences, and knowledge to prevent another epistemic agent B from forming a false (or epistemically risky) belief, but instead licenses B’s token belief-forming process. Licensing requires that B is aware of A and A’s opportunity to intervene.

This final definition applies to all cases discussed: DINNER PARTY, LIGHTING and MÜLLER-LYER. In all these cases, the epistemic bystander has the required opportunity, competences and knowledge to intervene in B’s risky belief-forming process. However, instead, the bystander licenses the risky belief formation. Thinking of epistemic bystanders in terms of licensing allows us to identify the epistemic wrong of bystanding. It is not merely that an interpersonal epistemic duty was violated. The wrong is in the bystander’s responsibility for boosting another agent’s confidence in their belief formation. This is the primary wrong of epistemic bystanding. They provide a false sense of security to another person about to form a belief in that environment.

One might worry here that epistemic bystanding understood as including licensing is too extensive. This is a well-known worry from discussions of general duties to aid. For instance, Patricia Smith (1990) states that a general duty to aid is in danger of becoming overly demanding. ‘That is, if we recognize one general positive right, we put ourselves on a slippery slope to oppressive obligations to aid everyone everywhere’ (Smith, 1990, p. 20). The same slippery slope presents a worry for the wrong in epistemic bystanding. If it is wrong to not prevent Martin from forming a false belief at the dinner party, it might seem that we ought to prevent people from forming false beliefs everywhere at all times. To make the absurdity even clearer, Lackey (2020a, 2020b) asks with regard to general interpersonal epistemic duties whether I am obligated to answer every single false assertion posted on Twitter. This obviously cannot be right and would be a hopeless endeavour. But Lackey provides a good answer herself. She suggests that interpersonal epistemic duties are imperfect duties. Some ‘[…] discretion and latitude are allowed in their fulfillment’ (Lackey, 2020b, p. 291). I am in an even better position than Lackey. Epistemic bystanding in the sense I have proposed is not as ubiquitous as failing to satisfy some general interpersonal epistemic duty. Bystanding understood as a form of licensing only occurs when an awareness condition is satisfied. This helps to keep the cases of bystanding limited and stops the slippery slope worry. Epistemic bystanding in the sense discussed constitutes only a subset of all cases in which an interpersonal epistemic duty is called upon. The cases—such as Lackey’s Twitter case—that pull duties towards becoming overly demanding are not a problem for me, because they fail regarding the awareness requirement for licensing. If someone is not aware of me and my potential intervention, then I cannot license their belief formation. Hence, I cannot be an epistemic bystander in those cases.

I have now provided an account of the mechanism behind epistemic bystanding. This mechanism—a form of licensing—accounts for the epistemic wrong committed in bystanding. I want to end with a final worry. If someone does not want to be an epistemic bystander, they have to intervene in a belief-forming process. Does this make them automatically epistemically paternalistic?

5 Is This Epistemic Paternalism?

Epistemic paternalism is usually understood as the practice of (i) interfering with someone’s inquiry, (ii) without their consent (iii) and for their own epistemic good (cf. Ahlstrom-Vij (2013), Bullock (2018) and Jackson (2020)). Jackson (2022) worries that this definition might be too broad and rules out cases that we intuitively do not take to be paternalistic - writing a book that influences someone’s belief formation, for instance. She hints that we could rule out these cases by being more careful about the consent condition. This seems the right way to go.

With regard to intervening in a belief-forming process in order to not be an epistemic bystander, condition (ii) does not even need to be considered. Condition (i) is already enough to distinguish the different forms of interventions. When I intervene and thereby avoid being a bystander, I do not interfere with an inquiry in the same sense that is at issue in the paternalism debate. I merely support the inquiry by adding evidence. This is made obvious when we look at cases usually discussed in the epistemic paternalism literature. They centre around withholding or manipulating evidence (Jackson, 2020), with Goldman’s (1991) prime example being about withholding evidence from jurors by judging it as inadmissible. What is wrong in cases of epistemic paternalism is that the epistemic agency is taken away non-consensually. Hence, Goldman describes that someone is guilty of epistemic paternalism ‘[…] whenever they interpose their own judgment rather than allow the audience to exercise theirs […]’ (Goldman, 1991, p. 119). Importantly, that is not what is going on when one intervenes in the cases I discussed—for instance when I tell Martin that there are two distinct David Mitchells, one who wrote Cloud Atlas, and one who is a comedian. I do not decide for Martin what to believe. I just provide additional evidence with my testimony. I am not guilty of being epistemically paternalistic.

6 Conclusion

Throughout the paper, I have illustrated and examined the phenomenon of epistemic bystanding. Moreover, I have now provided a path to understand the distinctive wrong in epistemic bystanding. I license another person’s risky belief formation when I am an epistemic bystander. This can be understood as a form of licensing.

I take an understanding of epistemic bystanding to be a first step towards reducing its frequency. Similar to attempts of understanding bystander effects that lead to inaction when help is needed (starting with Darley & Latané (1968)), we need to understand effects that increase epistemic bystanding next. If we understand factors that cause or impact epistemic bystanding, we can reduce its occurrence. In doing so, we will end up with fewer false beliefs. In a sense, the project of dealing with epistemic bystanding is therefore in the same realm of veritistic evaluations of social practices. Understanding epistemic bystanding will help us form better practices. Each of us can benefit from improved epistemic positions. I suspect that the opportunity condition will end up being the most relevant. Discussions of silencing (cf. Dotson (2011)) and epistemic injustice (cf. Fricker (2007)) already give us insight into mechanisms that prevent people from speaking up and cause an overall mean deficit of true beliefs in members of a society. The path forward is a reduction of the risks involved in intervening in other people’s risky belief formation, which will also reduce epistemic bystanding. How exactly such a reduction ought to function will be up to future research.