Educational Falsehoods

3.1. Three Mechanisms

There are at least three mechanisms connecting intentional misrepresentations to educational or epistemic gains. The first of these, which is familiar from the practice of scientific modelling, is simplification. Filtering out features of the target (the thing that is being represented) can help highlight some of its other features.¹⁰ This is particularly applicable when the target is complex. A representation that strives to represent all features of a complex target might overwhelm us with information, or distract us from properties of the target that are relevant to the matter at hand. Accordingly, a simplified representation can make important properties of the target more salient than they would otherwise be.¹¹

Consider how this might apply to climate communication. Imagine a simple visual model of global temperatures over time. At any given time, this model represents the Earth as one color, corresponding to the average global temperature. This representation clearly simplifies its target. It removes, among many other things, information about regional differences in temperature. And, consequently, it misrepresents regional temperatures, with some regions represented as colder or hotter than they really are. But this misrepresentation yields a countervailing epistemic benefit: it helps highlight the size and rate of increase in global average temperatures. Regional differences—and notably the fact that some regions have grown colder while others have grown hotter—may otherwise make these properties less visible.

The second process whereby falsehoods can facilitate epistemic access is exaggeration. We can make a feature of the target more visible, not just by filtering out some of its other features, but also by exaggerating the feature in question—that is, by distorting it in a way that makes it more striking or visible.

Exaggeration is commonly deployed in the context of scientific consensus statements. Consensus statements aim to convey that the scientific community agrees on a particular proposition. Such statements are widely thought to play an important role in convincing the public that particular scientific findings have strong evidential support (Bayes et al. 2020: 2). This is, in part, because disagreement among scientists tends to reduce their credibility in the eyes of the public.¹² Hence, consensus statements have long been a central part of science communication, particularly in the context of climate communication (e.g., “97 percent of scientists agree that anthropogenic climate change is real”) (Bayes et al. 2020: 2).

But getting the public to recognize the existence of a broad scientific consensus on climate change is challenging. According to Karen Kovaka (2021: 2368), this is partly because climate skeptic media tend to amplify what dissent there is, and partly because—independently of this amplification—laypeople often lend excessive credibility to scientists who dissent from the mainstream.¹³ The public therefore typically vastly underestimates the level of scientific consensus over climate change (Duffy 2022).

To counteract this phenomenon, climate communicators often exaggerate the extent of this consensus through the practice of “masking disagreement”. On this approach, members of scientific panels who disagree with a particular claim agree to refrain from expressing their disagreement, and to let this claim stand as the group’s position, so that the group can speak with one voice.¹⁴ Take, for example, a scientific panel that disagrees on the likely extent of climate-induced sea level rises. A member whose estimate (Y) is far lower or higher than their colleagues’ estimates (which cluster closely around X) might decide to let the much more widespread estimate stand as the group’s position. This allows the group to state a single “consensus” position.¹⁵ Strictly speaking, this needn’t involve asserting the falsehood that everyone agrees on X. For example, the panel might simply say “It is the panel’s position that climate change will cause sea levels to rise by approximately X” without explicitly saying that every panel member agrees with the panel’s position. Nonetheless, the practice is plausibly designed to implicate a misrepresentation. The panel could readily declare their actual level of disagreement. By masking their disagreement instead, they intentionally make it seem as though there is a higher level of consensus than is really the case.

This exaggeration, in turn, can help secure important epistemic benefits.¹⁶ Given the common perception that the existence of disagreement is a sign of bad science, presenting an artificially united front may help the public appreciate, fittingly, that climate scientists are credible sources on matters surrounding climate change (Beatty 2006: 54). And, more specifically, it may encourage the true belief that the object of the apparent consensus (e.g., that climate change will cause sea levels to rise by approximately X) is strongly supported by the evidence.¹⁷

The final epistemic mechanism I wish to outline here concerns familiarity. This mechanism builds on the previous two. Simplifying or exaggerating falsehoods matter, not simply because they allow us to filter out distracting features, or to enlarge features we wish to highlight, but also because, in doing so, they can help present information in a more familiar—and therefore, more cognitively accessible—way.

There are two notable ways this can happen. To begin, falsehoods can help package information in a familiar format. A key example of this relates to narrative. Suppose we wish to explain how climate research operates. We can present this information in different formats. We might, for instance, enumerate facts relating to the process of climate research. But we might instead choose to communicate this process in narrative (or story) form. Opting for the latter option invariably involves some measure of falsification. A story has a standard structure: prototypical stories involve one or more protagonists, who encounter a meaningful problem, and who then confront that problem, leading to some form of resolution (Fraser 2021). Making information fit this structure almost always involves simplifications and exaggerations. For example, identifying clear protagonists might require focusing disproportionately on the contributions of one or two scientists or scientific groups. Moreover, to dramatize the problem, communicators might disproportionately emphasize the obstruction of climate research by a few states or corporations, while omitting details of the challenging peer review process (Flottum & Gerjstad 2017).

Such simplifications and distortions help improve the audience’s understanding of the climate research process partly for reasons already mentioned: they help filter out material that may have overwhelmed or distracted us; and they enlarge important pieces of information we may otherwise have overlooked. But the crucial point here is that this is not the only reason they improve epistemic access. The further reason is that the story format resulting from these simplifications and exaggerations is cognitively familiar. As social psychologists have argued, we therefore often find it easier to process and remember information when it is presented in narrative form (Mandler & Johnson 1977; Fraser 2021).¹⁸

So far, the point has been that falsehoods can help present informational content about the target in a familiar format. But falsehoods can also contribute to making the target familiar—and so, more epistemically accessible—by likening it to some other, more familiar, object. The paradigmatic example of this is metaphor. Metaphor involves characterizing one thing (the target) in terms of another (the source). Hence, metaphors help us understand unfamiliar things by drawing on what we know about more familiar things. As Therese Asplund (2011) has shown, for instance, climate change is often publicly explained using metaphors that liken it to a greenhouse or to war.

How does this relate to falsehoods? Metaphorical representations of a target characteristically require some measure of simplification and distortion. For one thing, metaphors are always partial, in that the source helps understand some parts of the target, while hiding or concealing other parts. For instance, the greenhouse gas metaphor for climate change highlights some features of climate change (e.g., the fact that heat from the Sun is trapped, and that this engenders temperature increases). But it simultaneously conceals other aspects of the process (e.g., oceans’ role in thermal transfers) (Asplund 2011: 3–4; Chen 2012: 109–110). As for the parts of the target that are meant to be represented by the source, metaphors often exaggerate the similarity between the two, and thus yield a partially distorted representation of the target. Although the greenhouse metaphor highlights the causal connection between trapped sunlight and global temperature increases, it also invites us to see this connection as much more rapid than it actually is (Chen 2012).

Let us take stock. I have introduced three mechanisms through which the communication of falsehoods can facilitate epistemic access to a target. Falsehoods can simplify a target, by removing some of its features to make other features more visible. They can exaggerate, and thereby highlight, features of the target. And they can help present the target in a more familiar way, either by packaging information about it in a familiar format, or by likening it to something familiar.

Now, even though these processes are conceptually distinct, in practice they are often entangled with one another. Indeed, we have already seen that presenting a target in a more familiar light generally requires simplifying some of its features, and exaggerating others. To further exemplify these processes, and to illustrate their possible entanglement, I therefore wish to conclude this section by offering a final example in which all three combine to produce an epistemically felicitous outcome.

In the 1980s, atmospheric scientists became increasingly concerned about the growing depletion of the ozone layer, which protects the Earth from dangerous ultraviolet radiation. These concerns intensified in 1985 with the discovery of an especially thin area above the Antarctic, which came to be known as the ozone “hole.” The discovery prompted a rapid international response, culminating in the accelerated phaseout of depletion-accelerating chemicals (Ungar 1998; Grevsmuhl 2018).

In his fascinating historical exploration of this episode, Sebastian Grevsmuhl observes that the depleted area was not initially referred to or represented as a “hole.” Because, strictly speaking, what had been observed was a thinning or decrease of the ozone above the Antarctic (but not a complete lack), references to a “hole” were considered somewhat misleading and “metaphorical.” Accordingly, “one referee of [Richard Stolarski’s groundbreaking scientific article on ozone depletion] objected to the use of the ‘hole’ metaphor in the title of the paper,” so that the more accurate term “decrease” was eventually preferred by its authors (2014: 46).¹⁹ This was mirrored in visual representations. The key findings were initially represented in a simple graph showing ozone units above the Antarctic decreasing over time. This graph, Grevsmuhl notes, represented the findings “very efficiently,” but it “clearly could not make the case for a ‘hole’ in the ozone layer” (2018: 78).²⁰

Yet, though it was considered less accurate, the idea of an ozone “hole” soon gained currency in scientific discourse and, Grevsmuhl argues, ultimately played a key role in successfully educating the public about the real threat posed by ozone depletion. Importantly, for our purposes, a number of intentional misrepresentations featured in the construction of this idea. To see this, consider the famous NASA satellite picture of the ozone hole (Figure 1).

Figure 1: NASA Satellite Image of the Ozone Hole (Grevsmuhl 2018).

This visual representation was importantly simplified. As Grevsmuhl explains, the image “homogeni[zes] a substantial number of satellite measurements by correlating a certain data interval with a specific color” (2018: 78). Put differently, the picture color-codes measurements of the ozone layer in a way that lumps together large measurement intervals. Thus, it filters out substantial amounts of information about the different thickness of different parts of the ozone.

This simplification in turn leads to an exaggeration. It results in—and indeed, was designed to result in—an artificially clear contrast between the comparatively depleted area situated over Antarctica, and the rest of the ozone layer. By exaggerating this discontinuity between the two areas, the image “convey[s] the erroneous impression of a discrete hole in the atmosphere over the South Pole” (Mazur & Lee 1993: 711).²¹

Finally, artificially creating the appearance of an actual hole—and metaphorically referring to the depleted area as such—matters because it helps present the problem in terms of something familiar. To most people, the idea of thinning layers of ozone in the stratosphere is unfamiliar, and somewhat abstract. By contrast, the image of a hole in a shield—and even of a hole in a shield protecting us from lethal rays—is a familiar one. As Sheldon Ungar (1998: 523) explains, this image “meshes nicely with abiding and resonant cultural motifs. These Hollywood affinities range from the shields on Starship Enterprise to Star Wars.” So, the construction of the ozone “hole” helped convey, in familiar terms, the severity and urgency of the problem posed by ozone depletion.

The ozone case thus exemplifies how the three falsehood-involving epistemic mechanisms I have outlined can come together fruitfully. Misrepresentations of a target can simplify it, exaggerate it, present it in a familiar light—and thus, improve epistemic access to it. In the rest of §3, I wish to introduce several worries with, and corresponding qualifications of, this epistemic defense of falsehoods.

3.2. Must the Falsehoods be Recognized as Falsehoods?

I have argued that deploying falsehoods (understood as intentional misrepresentations) about a target can facilitate epistemic access to this target. But my argument in §3.1 might seem to imply that the epistemic value of falsehoods depends on their being transparently false to, and thus recognized as false by, the audience.

For one thing, my discussion of simplification is directly inspired by the practice of scientific modelling. But scientists generally know, and know that other scientists know, that scientific models are simplified. Philosophical discussions of scientific modelling are often explicit about this. Catherine Elgin, whose influential analysis of epistemically felicitous falsehoods I am indebted to, is clear that scientists are generally aware that the misrepresentations embedded in their models do not “purport to be true.” As a result, scientists typically do not believe these falsehoods (2017: 3, 23). One might think that this shared background knowledge, in the scientific community, is a key reason why the simplifications and other misrepresentations that feature in scientific models are capable of generating epistemic benefits. What is more, some of my examples in §3.1 were underspecified, such that we could readily imagine the public knowing that the misrepresentations in question are misrepresentations. For instance, the public know, given their background knowledge, that the greenhouse metaphor for climate change is a metaphor. Hence, they know that it does not purport to represent climate change with full accuracy.

More generally, then, one might worry that the positive epistemic functions associated with falsehoods exclusively apply to intentional misrepresentations that are communicated transparently, such that they are recognized as false by the audience. As explained in §2, this conclusion is unacceptable for my purposes. I aim to challenge an absolute or near-absolute commitment to scientific honesty. But falsehoods that are communicated transparently are plausibly neither dishonest nor deceptive. Accordingly, I need to show that the epistemic benefits discussed above can also be generated by misrepresentations that are communicated non-transparently, and whose falsity is therefore not recognized by the audience.

To argue for this claim, one might be tempted to appeal to ordinary pedagogical practice. Secondary school science teachers often teach theories that are strictly speaking false. As Sindhuja Bhakthavatsalam observes:

Today, we consider Newtonian mechanics to be false in light of Einstein’s theories of gravitation and quantum mechanics; we consider Rutherford’s and Bohr’s planetary models of the atom to be false in light of quantum mechanical models; … But these continue to be taught in K-12 science. In fact, it is not until late in high school or in many cases only in college that accepted-as-true—and often more sophisticated—theories are taught[.] (2019: 6)

This, Bhakthavatsalam goes on to argue, is not an accident. Teaching false theories, even when more sophisticated theories are available, can yield important educational benefits, not least by improving students’ understanding of the world (2019: 11–20). For example, though Newtonian mechanics may be false, accepting Newton’s laws of motion can help students better grasp how two bodies attract one another, and make accurate predictions about their movement, in immensely many cases.²² Now, if teaching false theories is educationally valuable, one might conclude, by analogy, that non-transparently communicating scientific falsehoods can be as well.

On closer inspection, however, this first argument runs into difficulties. Notice, first of all, that many schoolteachers teach false theories as false.²³ For example, they might preface the teaching of Newtonian mechanics by noting that it has been superseded by Einstein’s theory of general relativity, which rejects some of Newtonian mechanics’ theoretical claims (e.g., about the nature of space). Second, even when teachers do not preface false theories in this way, they often intend to reveal that these theories are false in due course—for example, when explaining, subsequently, why a new theory must be introduced.²⁴ Finally, and partly as a result of these first two practices, students might come to expect that the theories taught in school are not always strictly true. Consequently, even when a teacher does not explain that a theory is false, before or after it is taught, the educational context might make it reasonable for students to doubt that this theory is strictly true.

The point is that the falsehoods deployed by teachers are often either intended to be transparent to students (when the teacher explicitly discloses their falsity) or at least foreseeably transparent to students (when, for contextual reasons, students can reasonably expect that theories they are taught are not strictly true).²⁵ Accordingly, the pervasiveness of falsehoods in ordinary teaching contexts does not obviously show that non-transparent falsehoods (and, relatedly, dishonesty) can be epistemically valuable.

Even so, there is another, more direct, reason for thinking that non-transparent falsehoods can be epistemically valuable. Indeed, we can appreciate why this is the case by attending more closely to the mechanisms outlined in §3.1. Upon closer examination, none of these mechanisms depends on whether the falsehoods are recognized as false. Rather, they depend on whether, as a result of removing, or exaggerating, or otherwise misrepresenting (parts of) the target, some of the target’s features appear more visible or more familiar than they otherwise would.

To illustrate, suppose for the sake of argument that I do not know that the picture of the ozone hole described in §3.1 homogenizes many of the ozone layers, and that, consequently, it exaggerates the boundary between parts of the ozone situated above the Antarctic, and the rest of the ozone. Even if this is the case, these simplifications, and the resulting exaggeration, will still make it easier for me to see that an area of the stratospheric ozone is especially depleted, and to identify that area. Relatedly, even if I do not know that references to an ozone “hole” are metaphorical, this metaphor can still improve my epistemic access to an important fact—namely, that there is a critical weakness in the protective screen safeguarding us from radiation.²⁶

An important qualification is nonetheless needed here. Even if non-transparent falsehoods can yield epistemic benefits, they also come at an epistemic cost. The person who believes that there really is a hole in the ozone (as opposed to simply a substantially thinner area) represents the world less accurately than someone who knows that the “hole” is a metaphor, which highlights, but also exaggerates, a real phenomenon. Similarly, the person who believes that the consensus on sea-level rises is greater than it really is believes something false, which constitutes an epistemic cost, even if that false belief also facilitates true beliefs relating to climate change.

The implication is that, when deciding whether to communicate a (non-transparent) educational falsehood, we must weigh its epistemic benefits against its epistemic costs. In some cases, the benefits outweigh the costs. For instance, making it seem as though there is a literal hole in the ozone may be overall epistemically justified if this exaggeration successfully highlights the existence of a substantially depleted area. In other cases, however, misrepresentations conceal more than they reveal (Moser & Dilling 2004: 37). My defense of communicating falsehoods does not extend to these cases. Thus it is qualified, first of all, in that it applies exclusively to falsehoods whose epistemic benefits are reasonably expected to outweigh their epistemic costs.

This first qualification assumes that it is possible for science communicators to weigh these epistemic benefits and costs against each other. This may seem controversial. Yet I believe this assumption is warranted. To see why, consider that similar assessments are commonplace in both scientific research and pedagogical practice. It is widely acknowledged that scientific research pursues, not just any truths, but “significant truths”—for, as Philip Kitcher influentially observes, “there are vast numbers of true statements [e.g., about the number of blades of grass in London] it would be utterly pointless to ascertain” (2001: 65, emphasis added; see also Elgin 2017: 10–11, 82–83). Bhakthavatsalam applies this insight to science teaching: “it is not the case,” she insists, “that we teach any and all truths—we want to weed out the trivial, the obvious, the tautological, and so on” (2019: 6; see also Slater 2008: 528). So, both scientific researchers and teachers recognize, in their respective practices, that some true propositions are more significant than others, such that it is more important to believe them than others.

How do practitioners assess the significance of different true propositions? Practical considerations undoubtedly play a role. We research and teach about greenhouse gases or the ozone hole, in part, because there is a lot at stake (Kitcher 2001: 65; Slater 2008: 528). But judgments of significance are also guided, at least in part, by considerations of comparative epistemic significance. There are different ways of making such epistemic assessments. One common approach appeals to explanatory power. Elgin (2017: 58, 82–84), for example, observes that some propositions are “central” to our understanding of particular subject matters—by which she means, notably, that they help explain a wide range of phenomena within this subject matter—whilst others are comparatively peripheral to our understanding—such that comparatively few features of this subject matter depend, for their explanation, on this proposition.²⁷ For instance, one might think that the true proposition that the ozone layer is substantially depleted is needed to explain various biological or environmental phenomena (e.g., increased risks of skin cancer, decreases in phytoplankton production) (EPA 2023), but that the true proposition that ozone depletion consists in thinning layers rather than a literal hole is not.

I am not suggesting that such assessments of comparative epistemic significance are easy to make. Nor do I mean to suggest that judgments of overall epistemic significance can be wholly divorced from practical considerations (for reasons to be discussed in §3.4). But the important point is that they are routinely performed by scientific researchers and teachers. And so, from the fact that my defense of educational falsehoods requires science communicators to make such judgments, we should not immediately conclude that it is overly demanding.

3.3. Are Falsehoods Necessary?

But even with this first qualification, the fact that non-transparent educational falsehoods come at an epistemic cost might raise another question: namely, are these falsehoods really necessary? Put differently, would it be possible to help the audience achieve the same epistemic benefits without incurring these epistemic costs?

Take the case of consensus statements. Instead of masking scientific disagreement about p (“climate change will cause sea-levels to rise by approximately X”), one might provide accurate information about the extent of disagreement, while explaining that disagreement is not a sign of bad science, and that, despite residual disagreement, p is well supported by the evidence (see, for a similar suggestion, Pamuk 2021a: 84–85; Moore & MacKenzie 2020: 1–8; and Kovaka 2021: 2368–2369). Such an approach might seem to achieve the best of both worlds, epistemically speaking. It retains the epistemic benefits associated with masking disagreement (by helping the audience see that climate scientists and their conclusions are credible) while avoiding its epistemic cost (the concealment of disagreement).

This question of necessity matters. If there is an alternative way of achieving the epistemic benefits associated with (non-transparent) educational falsehoods, which avoids their epistemic costs, then, other things being equal, we should forego communicating such falsehoods.²⁸

Yet communicating non-transparent falsehoods is in fact sometimes epistemically necessary. Whether or not we can achieve certain epistemic benefits without deploying such falsehoods depends significantly on context. More specifically, it depends on at least four contextual factors. First, how complex is the matter at hand? The less complex the matter is, the more epistemically accessible it is—and so, the less necessary it is to misrepresent it so as to enhance its simplicity, exaggerate its features, or render it more familiar. Second, how much background knowledge does the public have?²⁹ Even if a matter is complex, the public might already know a significant amount about it, thus making it easier to convey information concerning it without distortion. Third, how pressing is the political issue on which the scientific matter bears? Insofar as we have time to thoroughly explain complex matters, and dispel public misconceptions, falsification seems less necessary. Finally, how cooperative is the broader communicative environment? Holding complexity, background knowledge, and time equal, it also matters what other communicators are doing. Trying to accurately explain scientific matters is more likely to succeed when the media, and other influential communicators, are cooperative, in that they are willing and able to help the public improve their scientific understanding.

What does this mean for the necessity of communicating falsehoods? The upshot is that educational falsehoods are more likely to be necessary, and thus more likely to be permissible, insofar as we are communicating about something complex, pressing, about which the public is uninformed or misinformed, and in a context where other influential communicators actively foster misunderstanding. Accordingly, my defense of (non-transparent) educational falsehoods will be restricted to these conditions. I agree that, the more we move away from these conditions, the harder it is to make the case that educational falsehoods are epistemically needed.³⁰

Though important, this qualification does not render my defense trivial by any means. This is because the non-ideal conditions just described are common in real-world science communication.³¹ This is notably the case with communication about climate change and climate research. The issue is extremely pressing; it raises technical or esoteric issues; these issues are often poorly understood, or indeed gravely misunderstood, by significant portions of the public; and rampant disinformation helps sustain these misunderstandings.

To make this more concrete, consider again the proposal that, instead of masking climate-related disagreement, we should communicate honestly and transparently about it. Explaining why disagreement is not necessarily a sign of bad science is complex. Epistemologists themselves continue to debate why and when disagreement should reduce our confidence in a proposition (Christensen 2009). Moreover, this complexity is compounded by public misconceptions about the significance of scientific disagreement. For instance, Kovaka argues that science is often taught in a way that lionizes scientists, such as Galileo, who dissent from the mainstream. This, she suggests, encourages skepticism toward mainstream positions whenever they face dissent (Kovaka 2021: 2368). Given the urgency of climate change, there is little time to correct these misconceptions. But even if there were, climate deniers actively obstruct attempts at doing so. They strategically exploit misconceptions about scientific disagreement by elevating dissenting voices, thus further unsettling trust in climate scientists and their conclusions. Put together, all of this makes it extremely risky, epistemically speaking, to be fully honest and transparent about climate-related disagreement (on this point, see also John 2018: 81). In such circumstances, it may therefore be necessary to mask disagreement, and thus exaggerate the degree of consensus, to help the public form the true belief that climate scientists are credible, and that the positions held by panels of climate scientists are well supported by the evidence.

3.4. Beyond Education?

My purpose so far has been to show that non-transparent falsehoods are sometimes needed to perform a valuable educational function. At this point, however, one might worry that the falsehoods I am defending require scientists to go beyond a strictly educational role. Central to my argument, in §3.3, was the idea that the scientific issue at hand is “pressing” or “urgent.” Relatedly, many of my examples refer to the way falsehoods can help highlight “dangers” or “critical” problems. But the judgment that something is pressing, or urgent, or dangerous, or critical, is, at least in part, a value judgment—in particular, it depends on ideas about what is morally or politically desirable. And if scientists are making such practical judgments when deciding whether or not to deploy falsehoods, then one might think that they are no longer acting primarily as educators, who seek to inform their audiences. Rather, they are acting as “issue advocates,” who seek to mobilize their audiences to promote ends that they consider important. This, in turn, might seem like an inappropriate role for scientists to perform.³²

The judgment that a scientific issue is pressing or urgent does indeed require appealing to nonepistemic (e.g., moral or political) values. From this, however, it does not necessarily follow that scientists who make such judgements are acting as issue advocates. To appreciate this, it is important to note that science is inescapably laden with nonepistemic values. That is, philosophers of science have widely argued that such values usually feature, and should feature, at numerous stages of scientific practice—for example, when deciding which questions to pursue, which methodology to use, which variables to represent in models, how to interpret events when recording them as data, whether to accept a hypothesis based on existing data, and so on.³³

The significance of nonepistemic values carries over to science communication—including science communication that aims to educate or inform the audience. A key function of educating or informing people is to empower them to engage in rational, autonomous decision-making (Brighouse 2009: 35–36; Elgin 2020: 64–65). Now, as Stephen John (2018: 83; 2019: 69–70) and S. Andrew Schroeder (2022: 41–48) have both observed, properly informing people—improving their epistemic position in a way that facilitates their capacity for rational, autonomous decision-making—requires attending to their values.³⁴ This is because what pieces of information are needed for autonomous decision-making depends significantly on one’s values. To cite one of Schroeder’s examples, taken from the medical context, being informed that a particular medicine contains gelatin matters for a strict vegetarian’s capacity to make an autonomous decision to take, or not take, the medicine; but this same piece of information might not matter, or matter nearly as much, for a non-vegetarian (2022: 43). The broader point is that, insofar as education aims to facilitate autonomy—as we will further discuss in §4.2—it must consider nonepistemic values.

This, however, does not mean that education simply collapses into issue advocacy. For it matters whose values are being used. Scientists who act as issue advocates often seek to mobilize others to promote goals that they consider to be important.³⁵ In contrast, on the picture outlined above, the values relevant to educating one’s audience, and thus empowering them to engage in informed decision-making, are in the first instance the audience’s values (Schroeder 2022: 54–56).³⁶ We can apply this to my running examples. When judging that climate change or ozone depletion constitutes a pressing issue, scientists needn’t be appealing to their own values, or seeking to promote their own goals. Their judgment might instead be that these issues are pressing in light of the audience’s own values and concerns. The thought might be, for instance, that given their audience’s own concerns for their health and security, or for the environment, or for their descendants’ well-being, they have a strong interest in being made aware of the reality of climate change or ozone depletion. In this way, we can account for the value judgements in question even if the scientist deploying falsehoods is primarily aiming to educate or inform, rather than advocate.³⁷

4.1. Are Educational Falsehoods Deceptive?

Deception involves intentionally causing beliefs in one’s audience that the speaker knows to be false (Carson 2010: 50). As discussed, many epistemically beneficial falsehoods are not deceptive. In many cases, scientists intentionally deploy misrepresentations in a context or manner that makes it plain to the audience that what they are communicating is a misrepresentation. But my focus, once more, is on educational falsehoods that are communicated non-transparently. Here, the speaker intentionally does not reveal, and may actively conceal, the falsity of what they are communicating. The case of masking disagreement is a paradigmatic example: the scientific panel exaggerates the level of consensus, while concealing that this constitutes an exaggeration.

Does this constitute deception? I believe it often does. To see this, consider that, in the non-ideal circumstances under consideration (§3.3), the epistemic benefits of the falsehood in question very often depend on the audience believing it. For example, in conditions where audience members misunderstand the epistemic significance of disagreement, a scientific panel hoping to impart truths about the reliability of climate science and the severity of climate change may need the audience to believe that there is less disagreement than there really is. If, as a result, they exaggerate the level of consensus to promote this last belief, then they are deceiving the audience: it is part of their intention that the audience form a belief they know to be false.

Or take the case of the ozone hole. We can, of course, imagine a version of this example where scientists refer to a “hole” in a transparently metaphorical sense, and do not intend for their audience to believe, counterfactually, that a literal hole exists. In this case, no deception occurs. Yet we can also imagine a version of this case where, due to the non-ideal communicative context, scientific communicators arguably need the audience to believe the misrepresentation. If, for example, the public knows little about atmospheric science, they may not realize the significance or imminence of the dangers associated with a thinning ozone unless they believe there is an actual hole going through the ozone layer. Suppose that, consequently, scientists hoping to impart truths about these ozone-related dangers tell the audience that there is a literal hole in the ozone, and conceal the fact that this is strictly speaking inaccurate. In this case, it seems plausible to think that the speakers are engaged in deception. Indeed, in this latter specification of the case, it is part of their plan that the audience believe the falsehood. Hence, in failing to be transparent about the inaccuracy of what they are communicating, they intend to induce a false belief.³⁸

One might object that this characterization is misleading. Even though (non-transparent) educational falsehoods intentionally cause false beliefs, their ultimate aim is to impart true beliefs, among other epistemic benefits, to the audience. Accordingly, one might deny—as Shlomo Cohen (2018: 494–496) has recently done—that deploying such falsehoods is deceptive.

I do not wish to rely on this response. The first reason is that, even if the aim of non-transparent educational falsehoods is to impart true beliefs and other epistemic benefits, they achieve this aim by creating false beliefs. Consequently, it seems more accurate to say that these falsehoods deceive as a means to an epistemically good end, rather than to say that they do not deceive at all.

To put this first point slightly differently, consider by contrast what Rebecca Brown and Michael de Barra have termed “altruistic lies.” An altruistic lie involves making “an untruthful statement to B with the assumption that B will not believe [this statement], and will instead come to believe the truth” (2023: 92). It is plausible to think that this is not deceptive, since it does not involve an intention to cause any false belief. But, to reiterate, the non-transparent educational falsehoods I am focusing on are not like this: they often aim to induce a false belief as a means to an epistemic gain. To claim that they are not deceptive at all is thus to deny this key (and prima facie morally important) difference between them and altruistic lies.

The second reason for not relying on Cohen’s response is dialectical. Even if Cohen’s response were correct, it remains controversial. To bolster my defense of educational falsehoods, I therefore wish to show that their permissibility does not depend on accepting this response. That is, educational falsehoods can be permissible even if we assume that they are deceptive.

I will establish this conclusion by scrutinizing three moral concerns associated with deception: its cost to personal autonomy (§4.2), its potential erosion of public trust (§4.3), and its impact on democratic accountability (§4.4).

4.2. Deception and Personal Autonomy

Deception is thought to be problematic partly due to its impact on the personal autonomy of those who are deceived. The threat it poses for autonomy is threefold.

The first threat concerns autonomy in the belief-forming process. The ability to form one’s beliefs autonomously—that is, to “govern” one’s belief-forming process—depends on the ability to reason effectively. That ability, in turn, depends partly on how good the evidence at one’s disposal is, and to what extent one has the opportunity to scrutinize this evidence.³⁹ Deception might seem problematic in both respects. By putting forward falsehoods, and presenting those falsehoods as true, deception risks impoverishing the quality of one’s evidence, and making it harder effectively to scrutinize this evidence. The upshot, one might worry, is that beliefs induced by deception are less likely to be formed autonomously.

But the issue is not just that deception can lead us to form beliefs non-autonomously. In addition, the beliefs induced by deception can impair our subsequent practical decision-making. This is because successful deception leads to false beliefs, and false beliefs can make it more difficult accurately to determine which courses of actions will promote one’s values or goals. So, deception can impair one’s ability to direct oneself towards one’s values or goals.⁴⁰

The final threat deception poses for autonomy concerns the relationship between the deceiver and the deceived. Often, deception turns those who are deceived into a mere means to the deceiver’s ends. In other words, deception often involves using those who are deceived as an instrument or tool, rather than treating them as a self-directing autonomous agent (O’Neill 1989: 111; de Melo-Martín & Intemann 2018: 43).

Consider again, for example, Fauci’s false claim about the ineffectiveness of masks. First, Fauci’s intervention impoverishes the quality of our evidence about masks, and conceals the fact that it is doing so. Hence, it arguably impairs our ability to reason effectively, for ourselves, about their usefulness. We therefore form our beliefs about masks less autonomously than we may otherwise have done. In turn, the false belief encouraged by Fauci’s deception (e.g., that masks are ineffective at protecting those who wear them) diminishes our ability to direct our lives in accordance with our ends. Suppose that what I most value is minimizing my personal exposure to COVID-19. Erroneously believing that masks are ineffective makes it more difficult for me to select the correct means to this end. Finally, on one interpretation of this case, the audience is being treated as a mere means. Suppose Fauci’s principal goal is to maximize public health, whereas some members of the public aim to maximize their own health. These two ends can diverge. Public health could best be served by having all non-frontline workers refrain from wearing masks (to make the limited supply available for frontline workers). But the best for me could be to wear a mask, while other non-frontline workers refrain from doing so. In this case, the deceptive interference does not simply make it difficult for me to pursue my ends. It also uses me as a tool to promote an end that is not my own.

Put together, these three costs to autonomy help explain why the Fauci intervention seems so problematic—and why, more generally, many philosophers consider deception to be morally wrong. One might therefore worry that, if (non-transparent) educational falsehoods are deceptive, then, notwithstanding their epistemic benefits, deploying them is impermissible.

There are several things to say in response. First, even if educational falsehoods impose these autonomy-related costs, it does not necessarily follow that communicating such falsehoods is always wrong. After all, there could conceivably be cases where the benefits of deception are so great that they outweigh its moral cost to personal autonomy.⁴¹

Admittedly, however, this first response is limited. It establishes that educational falsehoods could be permissible despite their cost to autonomy. Yet this is compatible with thinking that, given how unpalatable these autonomy-related costs are, the circumstances under which educational falsehoods are permissible remain vanishingly rare.

But there is a second and more important response. Not all forms of deception are equally problematic for autonomy. And educational falsehoods, in particular, seem less problematic for autonomy than standard cases of deception. Consider the second autonomy-related cost of deception, namely that deception makes it harder for the deceived to direct themselves in accordance with their values. Educational falsehoods can actually have the opposite effect. By improving the audience’s epistemic position, an educational falsehood can, as mentioned earlier, help audience members direct themselves according to their own values.

Take, for example, the case where audience members mistakenly view scientific disagreement as indicative of bad science. As discussed in §3, this background misperception can prevent us from registering significant truths (e.g., that the reality and severity of climate change is very strongly supported by the evidence). Failing to recognize these truths can make it more difficult to identify the course of action that will best promote our values (e.g., reducing one’s carbon footprint if one values the environment; pursuing geoengineering if one values industrial solutions to problems; etc.). The implication is that, insofar as educational falsehoods (such as exaggerations of the climate consensus) help us appreciate truths that are relevant to our values, they can also empower us to direct ourselves according to these values.

For the same reason, educational falsehoods do not necessarily treat the audience as mere means. To reiterate: the educational falsehoods I am defending can enable the audience to achieve their own goals or values by placing them in an overall better epistemic position. Whether one cares more about the environment or the promotion of industry, knowing that anthropogenic climate change is real is relevant to determining what course of action one should adopt to satisfy one’s concerns. Thus, insofar as educational falsehoods empower people to pursue their own values, it seems inaccurate to say that they treat people as mere means or instruments.⁴²

Can we say something similar about the first autonomy-related cost, namely undermining autonomy in the belief-formation process? One might try to alleviate this concern by pointing to the audience’s pre-existing epistemic predicament. As explained in §3.3, my defense of educational falsehoods focuses predominantly on situations where people are uninformed about the relevant issue, and where the communicative environment is rife with misinformation. In such a situation, people’s ability to reason about the relevant issue—and thus, their ability to engage in effective doxastic self-government—is already compromised to some degree. So, one might argue that, relative to the status quo, educational falsehoods come at little cost to autonomy in the belief-formation process.

But this response only goes so far. At most, it shows that the process through which (non-transparent) educational falsehoods induce beliefs may not be substantially worse, from the perspective of autonomy, than the process through which audience members would otherwise form their beliefs. Yet this should not distract us from the fact that educational falsehoods attempt to steer the audience’s beliefs in a way that bypasses their rational scrutiny. Like Fauci’s intervention, educational falsehoods achieve their positive effect (in this case, their epistemic benefits) by deploying falsehoods, and presenting them as true.

My conclusion is therefore not that (non-transparent) educational falsehoods come at no cost to personal autonomy. I believe they do: they realize their epistemic benefits in a way that bypasses the audience’s rational scrutiny, and thus, their autonomous control over their belief-formation process.⁴³ But the crucial point remains that, in other respects, educational falsehoods seem less inimical to autonomy than standard cases of deception. Indeed, they may even enhance their audience’s autonomy in meaningful ways. The epistemic benefits they induce can empower the audience better to pursue their own values. And, partly for this reason, they treat the audience not merely as instruments for others’ ends.

The upshot is that, even if educational falsehoods are deceptive, considerations of autonomy are unlikely to constitute a decisive objection to their deployment. In some cases, educational falsehoods may increase the autonomy of their target more than they diminish it. And even when they do not, their cost to autonomy is substantially less than that of standard cases of deception—and so, it could plausibly be overridden by the value of their epistemic benefits, and of the downstream consequences of those benefits.

4.3. Deception and Trust

Nevertheless, one might worry that the deceptive character of (non-transparent) educational falsehoods makes them problematic for a different reason: simply put, it risks undermining trust in science.

The reason for this worry is straightforward. With all deception, there is a risk of detection. And, as Pamuk observes, the detection of deceit seems likely to undermine trust in the deceivers. Hence, resorting to educational falsehoods in science communication risks eroding the public’s willingness to believe scientists. For Pamuk (2021b), this constitutes a decisive reason to steer clear of “noble lies.” Instead, scientists should be honest and transparent, notably about “the uncertainty and limitations of their knowledge.” Prasad (2020) likewise concludes that, given the risks deception poses for trust, “the safest path is to always and only present the truth.”⁴⁴

The risk of eroding trust is a real problem. What is worse, there are reasons to think that this problem is especially severe in the non-ideal circumstances I am focusing on. In uncooperative communicative environments, scientific claims are likely to be closely scrutinized by ill-motivated actors who are intent on delegitimizing them. This scrutiny, in turn, increases the odds that educational falsehoods will be exposed. So, the same conditions that make educational falsehoods epistemically necessary (§3.3) may also make such falsehoods more dangerous for trust.

There are several things we can say to soften the force of this objection. The first is that educational falsehoods seem less vulnerable to this worry than many other kinds of falsehoods. The extent to which finding out that one has been deceived warrants distrusting the deceiver depends importantly on the rationale for the deception—and, relatedly, on the extent to which one is likely to find this rationale acceptable. Indeed, finding out that you deceived me to steal my life savings plausibly gives me a far stronger reason to distrust you than finding out that you deceived me to surprise me for my birthday (a rationale that, let us assume, I wholeheartedly endorse). The former act of deception thus seems, on the face of it, to present greater risks for trust than the latter.⁴⁵

It therefore matters that educational falsehoods are intended to improve their targets’ epistemic situation—and that, as we saw in §4.2, such falsehoods consequently do not treat people as a mere means to another’s ends, but rather aim to empower them to pursue their own ends. It is not clear that being deceived for these purposes provides a warrant, or at least a strong warrant, for distrust. After all, being deceived for these purposes is something one might plausibly and rationally assent to.⁴⁶ Accordingly, finding out, after the fact, that you deceived me for these reasons seems less likely to destroy my trust in you—including my trust in you as an epistemically reliable source—than finding out that you deceived me simply for your own gain, or in a way that did not improve my epistemic situation. Returning to the objection at hand, the point is that, even if it is true that educational falsehoods risk being exposed, exposure may be considerably less detrimental to trust than it would be with most other forms of deception.

Yet one might think that this first response, too, is insufficiently sensitive to the challenges posed by uncooperative communicative environments. Even if the rationale underpinning educational falsehoods were acceptable to the public (such that appreciating it would mitigate the trust-related costs of exposure), the response assumes that scientists can make this rationale clear to the public. But doing so may be very difficult in non-ideal circumstances, particularly if ill-motivated actors actively promote misinformation about this rationale, for example by ascribing sinister motives to scientists who use educational falsehoods. This, to be clear, does not wholly negate the first response—it is not strictly speaking impossible successfully to communicate the rationale underpinning educational falsehoods. But it does mean that the extent to which this response can alleviate the trust-related worries at hand remains limited.

Still, these remaining worries about trust needn’t constitute a decisive objection to educational falsehoods. One reason for this—and this is my second response—is that they do not necessarily constitute a comparative problem for educational falsehoods. The objection at hand emphasizes that, in non-ideal and uncooperative communicative environments, deploying educational falsehoods can erode trust. But, in these same circumstances, a strict commitment to telling the truth can also undermine trust in scientists and their findings.⁴⁷

Consider again the case of scientific disagreement. Pamuk recommends that, instead of masking their disagreement, scientists should publicize their dissenting opinions, in order to initiate a process of public deliberation on the topic. Yet, as we saw in §3.3, this recommendation can sometimes do more to damage trust in science than to sustain it. Where people falsely believe that disagreement is a mark of bad science, publicizing disagreement could lead them to lose trust in scientists.⁴⁸

Of course, ideally, we would dispel these misconceptions about scientific disagreement. This is precisely why Pamuk (2021a: 88–90) recommends that scientific advice be integrated within a robust process of democratic deliberation on scientific findings. But dispelling such misconceptions takes time, and, as Pamuk notes, it may require the media to play a supportive role. The problem is that, in non-ideal conditions, these conditions commonly do not obtain. Time may be in short supply if, say, scientists are communicating about vaccine safety amidst a deadly pandemic. Moreover, in uncooperative communicative environments, prominent media sources may actively fuel the misconception that disagreement means bad science, and opportunistically highlight scientific disagreement to erode trust in epistemic institutions (Kovaka 2021: 2368). This last worry is far from hypothetical. As Russell Muirhead and Nancy Rosenblum (2020: chs. 5–6) have shown, contemporary disinformation often uses dissent to create a climate of disorientation, where the public no longer know whom to trust. And while it is possible to counteract such disinformation, doing so is typically a long-term task, which may not always be achievable in particularly pressing cases.⁴⁹

The upshot of this second point is that in the kinds of non-ideal circumstances I am focusing on, scientists often face a trust-related dilemma, where both educational falsehoods and telling the truth risk eroding trust.⁵⁰ So, even if educational falsehoods do come with trust-related risks (notwithstanding their distinctive rationale), these risks needn’t always give us a reason to favor the communication of truth over educational falsehood.

An important qualification is needed here. The different claims considered in this section—about the likely effects on trust of different categories of falsehoods and of truth-telling—are at bottom empirical claims. Ideally, therefore, we would simply appeal to empirical evidence to determine whether or not using educational falsehoods in uncooperative communicative environments always has an overall negative effect on trust, relative to truth-telling; and, if it does, how significant this effect is likely to be. However, existing evidence tends to be insufficiently fine-grained to settle these questions. While there is evidence, on the one hand, that perceptions of dishonesty are often detrimental to trust, and on the other, that telling the truth can also damage scientific trust in some circumstances,⁵¹ existing evidence tends not to distinguish sufficiently finely between different kinds of communicative contexts (e.g., cooperative and uncooperative), and different types of falsehoods (e.g., educational and non-educational).

Having said this, even if existing evidence cannot definitively settle the specific questions outlined above, it can still give us grounds for skepticism towards the most pessimistic hypotheses concerning the impact of falsehoods on trust. (Non-transparent) educational falsehoods are not new. For example, scientists, as we have seen, already practice masking disagreement. And they do so, often, in real-world contexts involving hostile media environments. So if such falsehoods were highly destructive of trust in science, all things considered, we might expect this trust to be low already. Yet trust in science remains, on the whole, quite high—including in areas of science, like climate science, that are highly politicized (Gundersen et al. 2022: 6–7; Cologna et al. 2023: 3–4; Wong 2024). Now, there is only so much we can infer from this evidence, for the reasons stated in the previous paragraph. But it does provide tentative grounds for thinking that, in real-world settings, including in settings involving uncooperative communicators, the use of educational falsehoods may not be as destructive of trust as it is sometimes feared to be.

4.4. Deception and Democratic Accountability

Over the course of my defense, I have invoked and relied upon several key qualifications. The permissibility of communicating non-transparent educational falsehoods depends, firstly, on the precise social context in which we find ourselves: educational falsehoods are more justifiable, as we have seen, in contexts involving pressing and complex issues, an uninformed public, and an uncooperative communicative environment (§3.3). And their permissibility depends, secondly, on the type of falsehood we are communicating: the falsehood in question must be such that the epistemic benefits of believing it are likely to outweigh its epistemic costs (§3.2).

One might therefore worry that my defense places excessive faith in scientists and other scientific communicators. It requires them to show restraint, and to exercise keen judgment (relating to descriptive and evaluative matters, as discussed in §3.4), when assessing when to deploy falsehoods, and what falsehoods to deploy. This, one might think, is problematic. Given my focus on non-ideal circumstances, we should not adopt an overly idealized vision of scientists. Real-world scientists are imperfect, and thus, accidentally or deliberately, they may fail to communicate the right kinds of educational falsehoods, in the right kinds of conditions (Bok 1978: 24–30; Director 2023: 956–962).

What this concern shows is that those tasked with communicating falsehoods must somehow be held accountable to the public. In other words, oversight mechanisms are needed to ensure that, as far as possible, science communicators deploy educational falsehoods in ways that conform to the conditions I have outlined. This seems problematic for my defense, because holding scientists and other scientific communicators accountable is likely to be very difficult.

What can we say about this worry? One might firstly deny that the problem at hand is really about educational falsehoods. Any attempt at integrating scientific expertise within democracy faces the problem of how scientists and other experts can be held accountable. Given the informational asymmetry between experts and laypeople, it is extremely difficult to assess whether expert communications are true or false, or whether they result from good epistemic practice (see, e.g., Anderson 2011; Pamuk 2021a). Thus, one might think that the problem stems not from anything specific to educational falsehoods, but rather from the informational asymmetry between scientists and laypeople.

But this initial response may be too quick. A skeptic might insist that the difficulty of holding scientists accountable is amplified when they deceptively communicate falsehoods. One reason for thinking so concerns the epistemic consequences of the communication. Communicating falsehoods about something is likely to cause false beliefs about it, and one might worry that the resulting ignorance in turn makes it difficult to hold communicators accountable.

Yet this first concern seems inapplicable to the educational falsehoods I am defending. Even if these produce false beliefs, they thereby put the audience in an overall better epistemic position than they would otherwise have been. Accordingly, if ignorance erodes accountability, this is conceivably an argument for educational falsehoods, rather than against them.

Still, there is a second reason for thinking that the deceptive communication of educational falsehoods makes it more difficult to hold scientific communicators accountable. The problem here is not so much about the epistemic consequences of the communication. Rather, it is about the lack of transparency regarding the reasons that underpin this communication. The falsehoods I am defending are not and often cannot be fully transparent: a scientific communicator cannot disclose that they are communicating a falsehood in order to produce epistemic benefits because, as discussed in §3.3, the epistemic benefits of the falsehoods I am defending may well depend on the audience believing those falsehoods. The worry is that this opacity makes it more difficult for the public to scrutinize scientific communications. If scientists are not transparent about the fact that some of what they are saying is false, this plausibly makes it harder to determine whether the falsehoods they communicate conform to the requirements I have outlined.⁵²

This second concern is more worrying than the first. But, on closer inspection, it does not constitute a decisive objection to educational falsehoods. The issue of transparency, and its relationship to accountability, is in fact a familiar one in democratic theory. In particular, deliberative democrats have influentially argued that it may sometimes be useful for public speakers to debate the merits of a policy in secret, and later express support for (or opposition to) that policy, without necessarily disclosing the reasons that led them to support (or oppose) it. Doing so is useful, in part, because it allows speakers to change their minds in response to these reasons, when doing so would otherwise be politically costly (e.g., Gutmann & Thompson 1996: ch. 3; Chambers 2004: Kogelmann 2021).⁵³

Crucially, deliberative democrats widely agree that this lack of transparency about the reasons underpinning one’s communication can be reconciled with the demands of democratic accountability. This is because it is possible to devise institutional mechanisms that reduce the likelihood that non-transparency will lead to communicators behaving unaccountably. Numerous institutional proposals have been advanced (see Kogelmann 2021 for an overview). Some focus on the internal composition of the deliberating group. For instance, Simone Chambers (2004: 408) recommends that the composition of the secret deliberative body should mirror, as far as possible, that of the broader population, such that it remains responsive to diverse values, concerns, and interests. Other proposals involve external oversight measures. Here, one possibility would be to charge a committee of policy advisers, or even a small assembly of randomly selected citizens, with the task of assessing whether the group in question is making responsible use of its secrecy.⁵⁴

Although these proposals were not designed with educational falsehoods in mind, they may conceivably apply to this particular case of non-transparent communication. For example, we might ensure that scientific panels are internally diverse, so that, when they decide whether to publicly communicate falsehoods, they are more likely to do so in a way that is responsive to an appropriate range of values, concerns and perspectives (Moore & MacKenzie 2020: 3–4). In addition, or alternatively, we might ask a committee composed of scientifically literate policy advisers, or randomly selected citizens who have undergone issue-specific training, to monitor prominent scientific panels such as the IPCC.⁵⁵ The committee would examine the panel’s findings and communications, and question the panel where needed, in order to assess whether their communications distort the findings—and where they do, whether it is plausible to think that these falsehoods serve an appropriate purpose. In cases where they blatantly do not, the committee could then blow the whistle.

To be clear, these proposals are tentative. There may well be better institutional mechanisms for promoting accountability, and it is beyond the scope of my present argument to identify the best such mechanism. But this should not detract from my central point: namely, that the issue of non-transparency—and more specifically, of its potential impact on accountability—is not unique to educational falsehoods. On the contrary, it is a familiar issue, shared by other communicative practices, which are nonetheless widely considered to be permissible and compatible with respecting the value of democratic accountability. The implication, and what the above proposals are meant to highlight, is this: just as we can devise strategies to make non-transparent deliberative practices accountable, so too we can devise strategies to hold communicators of non-transparent educational falsehoods accountable.