2. Conceptual Engineering as a Way of Improving the Epistemic Environment

Conceptual engineering means purposefully assessing and improving concepts—that much nearly everyone agrees on. The controversial part is what the “concepts” are. I take a broadly psychological approach here (for examples and discussion, see Machery, Reference Machery2017, chapter 7; Machery, Reference Machery2021; Fischer, Reference Fischer2020; Isaac, Reference Isaac2023; Koch, Reference Koch2021). Concepts, on that approach, involve a mix of often rudimentary and incompletely articulated theories, possibly inconsistent assumptions, imagery, prototypes, and exemplars that are associated with a term. We typically employ concepts unreflectively, but we can also reflect on them and try to make them explicit, as philosophers engaging in conceptual analysis often aim to. Activating a concept gives rise to a range of cognitive activity—thoughts, feelings, and imagery. As Edouard Machery puts it: “Concepts are the tracks our minds prefer to travel on” (Machery, Reference Machery2017, 222). This way of thinking about concepts contrasts with thinking of them as semantic meanings that might well not be in the head and that might be determined by inscrutable external circumstances (e.g., Cappelen, Reference Cappelen2018). I take the psychological approach to concepts because concepts in this sense have the role in our cognitive lives that is relevant for engineering concepts to improve cognitive dispositions—and that is the kind of conceptual engineering that I am here concerned with.

Let us look at some examples of how we can engineer such psychologically construed concepts to improve cognitive dispositions. A common case for eliminating slurs is that they reinforce false associations, such as the association between an ethnicity and some negative trait or simply negative affect. For example, using the word “Boche” to think about Germans reinforces the association between being German and being cruel and evokes unjustified negative affect. (This is an example that Machery (Reference Machery2017, 223) uses, drawing on Robert Brandom.) The hope, then, is that eradicating slurs from public discourse and thereby making people less likely to employ these slurs in their private thinking can make the false associations less salient and thereby reduce the occurrence and the strength of false beliefs about the relevant groups.

Another example comes from Georgi Gardiner, who suggests that we can engineer the concepts in the vicinity of “love”, which “affect an individual’s perceptions of their own attachments, which can—owing to interpretative feedback loops—affect those attachments” (Gardiner, Reference Gardiner, Fairweather and Montemayor2023, 304). For example, suppose that you find yourself having feelings of affection for an authority figure. What terms in the vicinity of “love” are available to you, how salient each is, and what content is associated with the terms shape how you interpret this affection; this in turn shapes how your thoughts, feelings, and expectations develop further. If you interpret your feelings as (possibly) “true love,” you might go on to think of the situation as a very significant opportunity that must be pursued even at great cost. If you make sense of the affection as “limerence,” you might see it as something pathological and worrisome that must be overcome. And if you think of it as “puppy love,” you might judge that it is something trivial that will soon pass on its own. Presumably, these directions of thought are not all equally good, (broadly) epistemically speaking—some of them fit reality better or are otherwise more apt than others. So, engineering which terms are easily available for interpreting such experiences, and what content we associate with those terms, is a possible way of improving people’s belief formation in this domain.

For yet another example: properly instilling the concept of sexual harassment in people’s minds arguably supports appropriate mental reactions to certain behaviors. For example, classifying someone’s behavior as sexual harassment naturally leads to the thought that the victim has been treated unfairly and the perpetrator should be punished. If these are true thoughts, then having and using the concept helps us form important true beliefs that we might otherwise fail to form (see also Fricker, Reference Fricker2007, chapter 7). Once the concept is instilled, we might then purposefully broaden it, for example, to cover not just workplace harassment but also street harassment (Crouch, Reference Crouch2009), to make people spontaneously sensitive to and critical of a wider range of problematic behaviors that need attention. Of course, some are critical of such conceptual expansion, arguing that this makes it difficult to tell when a strong critical reaction is needed and when there should be a milder reaction or even no reaction. The cognitive effects of a given conceptual change are frequently a matter of dispute, and even more importantly, there can be reasonable disagreement about the desirability of these effects.

Although concepts are here understood as things “in the head,” in the individual mind, it makes sense to think of such conceptual engineering as a way of improving the epistemic environment. This is because in order to shape the concepts in individual minds, we need to shape the linguistic environment that in turn shapes the concepts. Shaping the linguistic environment typically means changing how influential speakers use words; the linguistic practices that are prevalent in mainstream media (or alternative media, depending on whom we are targeting) and in everyday discourse; what the dictionary says; and the various linguistic policing practices: how people encourage or discourage ways of using language. For example, in order to change what terms in the vicinity of “love” are available and easily accessible for someone, and what associations or imagery these terms evoke, we need to change how, and how frequently, those around that person use the word “love” and other relevant words, what the dictionary entry for “love” says, and so on.

Calls to assess and improve the linguistic environment align naturally with epistemic environmentalism—the view that we should focus much more than so far on assessing and improving the epistemic environment. For example, Neil Levy writes: “Just as we urgently need to repair and to manage our natural environment … we must repair our epistemic environment” (Levy, Reference Levy2022, 110). For Levy, this centrally means better calibrating the cues for expertise with actual expertise—ensuring “that people who lack expertise can’t easily give themselves an unearned appearance of expertise” (ibid., xv). Shane Ryan takes the comparison with the natural environment quite literally and argues that “dishonest testimony by experts and certain institutional testifiers should be liable to the sanction of inclusion on a register of epistemic polluters” (Ryan, Reference Ryan2018, 97). Alessandra Tanesini likewise argues that we often rely on “physical and social scaffolds that create an environment or niche conducive to knowledge” and that epistemologists should subject these scaffolds to assessment (Tanesini, Reference Tanesini2022, 367). A common idea motivating this environmental turn is that humans’ cognitive abilities are very limited, and we depend on others in epistemic matters much more than has been recognized by mainstream epistemology. Mainstream epistemology has focused on prescribing norms of rational belief formation and inquiry to idealized agents; but we should think more about how to improve real, limited humans’ belief formation in the real world. And this we can most effectively do by improving their epistemic environment—for example, by making sure that those who appear as experts in that environment do tend to be experts.

An important aspect of our epistemic environment is the language used around us. So far, the emphasis in the discussion on improving epistemic environments has been on testimony; for example, echo chambers or epistemic bubbles (Nguyen, Reference Nguyen2020), where one is repeatedly exposed to certain testimony and deprived of other testimony. The linguistic environment, however, also shapes our beliefs, and not too dissimilarly. For example, on the one hand, one can be exposed to authoritative speakers often asserting that Germans are cruel and thereby acquire the corresponding belief. On the other hand, one can be exposed to people often using the slur “Boche” and therefore adopt the corresponding concept, along with an association between being German and being cruel. Likewise, how people use the word “love” around us shapes how we think of romantic relationships, in addition to the “love experts” in our lives who explicitly advise us on these matters. Furthermore, how the term “sexual harassment” is used influences whether we even notice certain behaviors, and how much we are inclined to condemn them—again, regardless of whether we are explicitly taught by others that these behaviors are bad and just how bad they are. If we are to take the environmental turn, then it is surely worthwhile to consider not just what explicit testimony is salient and authoritative to us but also the ways in which our linguistic environment implicitly instructs us to think, see, and feel.

3. Worries about Epistemic Autonomy

But should we take the environmental turn? An important set of worries, in this regard, concerns epistemic autonomy. To understand when and how environmentalism can threaten epistemic autonomy, we first need to get a grip on what “epistemic autonomy,” in the relevant sense, is. It is by now a familiar point (expressed, e.g., by Zagzebski, Reference Zagzebski2012; Bullock, Reference Bullock2018, 441–442; Matheson, Reference Matheson2023, 84–89) that epistemic autonomy, in the sense in which it is important and valuable, should be understood as self-governance rather than self-reliance. Footnote ¹ So, “epistemic autonomy” in the relevant sense does not mean achieving valuable epistemic results by oneself—for example, refusing to consider expert testimony as evidence and figuring everything out on one’s own. It means, instead, exercising rational control over one’s beliefs: subjecting one’s beliefs to scrutiny and governing them conscientiously, instead of uncritically accepting whatever beliefs one finds oneself having and instead of having one’s beliefs governed by others.

We can distinguish between a motivational and an ability component within such autonomy. The motivational component includes wanting and trying to exercise rational control over one’s beliefs, and the ability component includes the internal and external factors that allow one to do so at all or to do it successfully. Linda Zagzebski suggests that the basic norm of such autonomous self-reflection is conscientiousness: “the property of exercising my faculties in the best way I can to make the outputs of those faculties fit their objects—to make my beliefs true, my desires of the desirable, my emotions appropriate to their intentional objects” (Zagzebski, Reference Zagzebski2012, 230). I leave it more open what it is to govern beliefs and other mental states well—whether it is about making faculties fit their objects or something else, and what it is for a particular faculty to fit its object. But I take on board the more general idea of “the self in its role of manager of itself” (Zagzebski, Reference Zagzebski2012, 234) as central to autonomy. As Zagzebski (ibid., 235–236), among others (e.g., Matheson, Reference Matheson2023, 86–88), emphasizes, epistemic autonomy as self-governance is compatible with deferring to experts in a conscientious, reflective manner, and more generally with relying on others—for example, improving one’s reasoning skills in a logic class or developing one’s beliefs through debate with others. Epistemic autonomy as self-governance is thinking for oneself but not necessarily by oneself (e.g., King, Reference King2021, 88).Footnote ²

The idea that real-life epistemic progress requires a focus on improving the epistemic environment seems to be in tension with the idea that we should all be the managers of our own minds and let others be the managers of theirs. So, there seems to be a tension between epistemic environmentalism and respecting epistemic autonomy. To make the worry clearer, I will distinguish between three ways in which we might want to respect autonomy in the sense of self-governance—namely, as a right, an ideal, and a duty—and accordingly, three ways in which epistemic environmentalism might fail to respect autonomy. We will see that not just any focus on improving the epistemic environment disrespects autonomy: the worries arise when we go about improving the environment in certain problematic ways or do it at the expense of developing individual motivation and capacity for self-governance. So, these worries do not ultimately pose a devastating objection to environmentalism as such; but environmentalists should bear these pitfalls in mind and make sure to avoid them.

Let us begin with epistemic autonomy as a right: the right not to have one’s beliefs governed by others. As Zagzebski puts it: “From the outside, autonomy is violated when another person imposes her will on me” (Zagzebski, Reference Zagzebski2012, 234). Having one’s epistemic autonomy as a right respected does not require internal motivation or ability to govern one’s own beliefs well—it only requires freedom from excessive external purposeful shaping. We do not have a right to an environment that does not influence our intellectual lives; that would be an absurd demand. But we arguably do have a right to an environment that is not excessively designed by others to shape our intellectual lives in accordance with their will—even when those others want to improve our beliefs and in fact do that. Such a right to freedom from excessive control by others—including benevolent and competent control—is a central issue in the literature on epistemic paternalism (e.g., Bullock, Reference Bullock2018).

For example, nudges are often suspected of violating the right to govern one’s own choices or beliefs. Nudges are non-coercive interventions, like placing healthy food at eye level in the cafeteria (Thaler & Sunstein, Reference Thaler and Sunstein2008, 1–4) or publicizing unflattering images of a charismatic politician, to convince people that he committed the crime that he was rightly convicted of (Grundmann, Reference Grundmann2023, 214). It is widely held that people exercise excessive control over other people’s minds when they bypass the latter’s rationality by means of clever nudges, designed to exploit human cognitive biases. Levy (Reference Levy2022, chapter 6) has tried to dispel the worry that designing a good epistemic environment (where, for example, expert opinions are salient and quacks have a hard time finding a platform) involves manipulative nudging. The rough idea is that we typically assume that what is made salient to us by others, in human-designed environments like cafeterias or social media, is recommended by those others as choice- or belief-worthy. And so, making choice- or belief-worthy options salient bypasses rationality no more than ordinary responsible recommending does. That idea is controversial (see, e.g., Grundmann, Reference Grundmann2023). But even if there is some truth to it (I think there is), one might argue that respecting autonomy as a right requires more than just not bypassing rationality, in Levy’s sense. Even if we read the environmental nudges as recommendations, and thus as higher-order evidence, curating such higher-order evidence without consultation and transparency might still amount to excessive control. Perhaps respecting autonomy as a right requires that one has some say, or is somehow represented, in designing the implicit recommendations in one’s environment (Iizuka & Kobayashi, Reference Iizuka and Kobayashi2023). For present purposes (teasing apart the distinct worries about autonomy), we need not specify when exactly environmental design crosses the line and amounts to excessive control over others’ minds; but this is no doubt an important issue for environmentalism.

Conceptual engineers also sometimes go about improving the epistemic environment in a way that threatens other people’s right to freedom from excessive control. This is what I have argued in earlier work (Kitsik, Reference Kitsik2023): certain implementation strategies for conceptual engineering threaten people’s right to govern their own inferential and attentional patterns. Queloz and Bieber (Reference Queloz and Bieber2022) raise related worries about conceptual engineering as a potential threat to freedom of thought. Queloz and Bieber are especially concerned about institutionalized conceptual engineering, from the perspective of democratic and liberal principles. I discussed how conceptual engineers can also problematically shape other people’s concepts and thereby their mental lives by other means—for example, by lobbying for certain editorial policies or by changing social norms about language use with the help of influential people and social sanctions (Kitsik, Reference Kitsik2023). The targets of such practices, I argued, can reasonably think of these practices as violating their sovereignty over their own minds—especially when such shaping of beliefs is chosen over addressing people directly by rational persuasion. So, to the extent that conceptual engineers purposefully shape the linguistic environment in order to improve other people’s cognitive lives, without engaging them directly by explicit rational persuasion, we have reason to worry about violating their epistemic autonomy as a right.

There are, however, two further autonomy-related worries that arise for epistemic environmentalism in general and engineering the linguistic environment in particular. These worries have not yet been distinguished from the previous one in the literature on conceptual engineering, and they are also often not distinguished clearly from the previous worry in the literature on epistemic environmentalism and paternalism. One of these worries concerns epistemic autonomy as an ideal. My distinction between epistemic autonomy as a right and as an ideal is broadly the same as Zagzebski’s: “[A]utonomy can be understood as a right—a claim to be permitted to govern oneself, or it can refer to the successful exercise of that right—governing oneself correctly” (Zagzebski, Reference Zagzebski2012, 234). Epistemic autonomy as an ideal, then, means governing oneself correctly—or as I would rather put this, it means having the internal motivation and ability to govern one’s beliefs well. The corresponding worry for environmentalism, then, is that when we focus too much on creating an environment that helps people to acquire true beliefs, knowledge, or other epistemic goods regardless of their own motivation and ability, we might neglect the important task of developing this internal motivation and ability. Further, to the extent that people fail to develop it, they miss out on a central aspect of the good life: conscientiously governing one’s own beliefs, one’s own intellectual life.

Autonomy as an ideal is threatened by the sort of environmentalism that endorses accepting the explicit and implicit testimony of our social and physical environment in a relatively uncritical manner. Such environmentalism, which I take Levy (Reference Levy2022) to represent,Footnote ³ focuses on making the environment provide better testimony, rather than on improving people’s ability to conscientiously select their information sources or to think through the issues for themselves. But having an active mind of one’s own, being a self-directed seeker of truth and understanding, is plausibly an important aspect of the good life. One might, of course, question this aspect’s importance, compared to that of true beliefs and knowledge, as Robin McKenna does: “It is not clear that we have any reason to think intellectual autonomy is more important than having true beliefs or knowledge about things that matter to us” (McKenna, Reference McKenna2023, 96). One might also say that we can and should do both: improve the epistemic environment and people’s own capacities and motivation for self-governance. Perhaps that kind of moderate environmentalism is no threat to epistemic autonomy as an ideal. The worry is just that a too one-sided environmentalism results in diminished capacity and motivation for self-governance. This is one way to understand Daniella Meehan’s (Reference Meehan, Axtell and Bernal2020, 252) worry about Harry, whose flatmate creates a better epistemic environment for him by offering a discount on the subscription for a reliable newspaper and leaving neutral news programs playing on the television. Meehan argues that this is like placing a bubble wrap around a vase: underneath it, the vase (or Harry’s epistemic character) remains fragile (ibid., 255). And this is a problem, one might say, because such superficial interventions fail to empower people to live a good life along an important dimension: epistemic self-governance. So, we should not focus single-mindedly on such interventions.Footnote ⁴

As mentioned earlier, worries about conceptual engineering as a threat to the right to autonomy have already been raised (Kitsik, Reference Kitsik2023). But conceptual engineering can also sit uncomfortably with epistemic autonomy as an ideal: conceptual engineers can focus too much on designing an epistemically cushioning linguistic environment and neglect the task of encouraging and teaching people to scrutinize their own conceptual repertoire and choices. Plausibly, an important part of living well is questioning the received wisdom of our culture, including the ways of thinking encoded in our language. One reason to question the wisdom of our language is that it might in fact not be so wise. In Friedrich Nietzsche’s words, our concepts are “the inheritance from our most remote, most foolish as well as most intelligent ancestors” (Nietzsche, Reference Nietzsche, Kaufmann, Hollingdale and Kaufmann1968/1901, 221). But we might want to question the tracks on which our concepts put our minds not only because these might not be the best tracks but also because we want to manage our own minds—to be the agents in our own intellectual lives. Such agency need not involve devising new and better concepts from the ground up by oneself. It can involve, for example, being more discerning and reflective in accepting others’ explicit and implicit advice on how to use language.

Yet another autonomy-related worry, often not clearly distinguished from the first two—indeed, it is usually not recognized as autonomy-related at all—is the worry about epistemic autonomy as a duty. Whereas the ideal of epistemic autonomy is motivated by the thought that managing one’s own mind contributes to personal flourishing, and is thus something rewarding and enjoyable, epistemic autonomy as a duty is rather conceived as a burden that everyone must carry for themselves, to be a good citizen of one’s community. William K. Clifford did not use the term “autonomy” in “The Ethics of Belief”; but the main theme of that essay is the duty of each person to govern one’s own beliefs well. We must each “guard the purity of [our] belief with a very fanatism of jealous care,” Clifford writes, and this is not only the job of the experts or authorities: “It is not only the leader of men, statesman, philosopher, or poet, that owes this bounden duty to mankind” (Clifford, Reference Clifford1999/1877, 74–75). Recall Meehan’s (Reference Meehan, Axtell and Bernal2020) case with Harry and his flatmate. This can also be understood as a worry about epistemic autonomy as a duty: the flatmate is doing too much of the work that is properly Harry’s, and the interventions thus fail to contribute to Harry becoming a good epistemic citizen (possibly even turning him into a worse epistemic citizen). This version of the worry is not about Harry’s own flourishing; it is a worry about the community that he is part of. Plausibly, a well-functioning society requires everyone to assume a great deal of responsibility for managing their own beliefs.

The duty to govern one’s own beliefs does not require abstaining from reliance on testimony. Clifford stressed, however, that we must rely on others conscientiously, considering whether our informant is sincere and knowledgeable (Clifford, Reference Clifford1999/1877, 79). We also cannot always consciously reflect on these factors, of course; a more plausible idea in the vicinity is that we should all be responsible epistemic “consumers.” For example, Jonathan Matheson (Reference Matheson2023, 197–200) discusses the “consumer virtues” of intellectual discernment, charity, being properly attuned to important information, and testimonial justice. Levy’s (Reference Levy2022) environmentalism has been criticized for downplaying the individual’s responsibility as an epistemic consumer (Dutilh Novaes, Reference Dutilh Novaes2023, 801; Murphy-Hollies & Caporuscio, Reference Murphy-Hollies and Caporuscio2023, 809). In response, Levy concedes that in his view, “It’s not far from the truth … to say that beliefs happen to people” (Levy, Reference Levy2023, 856). It is, however, hard to accept that, for example, people who buy into conspiracy theories or harmful false gender stereotypes, or deny climate change, are just blameless victims of their epistemic environment. They should try and do better, and we should not approach engineering the epistemic environment in a way that fails to recognize that.

The environmentalist might reply that of course, people ought to take good care of their own beliefs. But they very often do not; because individuals’ epistemic failures can be detrimental to others as well as themselves, we need to design the environment (e.g., make expert voices salient) to a greater extent than we would in a world of model epistemic citizens. This makes good sense; but my point is that the environmentalist should be careful not to lower the expectations for individuals’ efforts too much, in light of their non-compliance, and that we should still strive for a situation where everyone is motivated and able to do their own part of the epistemic work.

Disregarding the duty of epistemic self-governance is another pitfall that conceptual engineers must also avoid, as they try to improve the linguistic environment. An engineer who undertakes to design a supportive linguistic environment that puts people’s thoughts on the right tracks, without much effort on their own part, must bear in mind that people should themselves be responsible, reflective concept users—that this is their duty to their community. For example, people should put some work into making sense of their experiences, rather than settling on the first concept that pops into mind and sliding along on those tracks. Joe Goldberg, from Caroline Kepnes’ You (2014) (and the TV series), whose uncritically internalized bad ideas about love have led him to a pattern of stalking women and killing everyone standing in the way, is not blameless victim of his linguistic environment. We cannot scrutinize every concept we use; but surely, we should put more effort into scrutinizing those that guide our behavior in such important ways. Furthermore, we have a responsibility to be discerning about whom we look to, in matters of language. For example, when one’s male friends and role models insist that “sexual harassment” should be used in a very narrow way, one should pause to question whether those informants are sincere and competent regarding the matter. Conceptual engineers should not proceed as if people had no such self-governance duties of their own as concept users. People’s non-compliance with these duties should not lead us to do all the work for them; we should, instead, think about how to make them more able and willing to do this work.

4. Participatory Conceptual Engineering

The worries about self-governance as a right, an ideal, and a duty arise when the linguistic environment is created “from above,” by some people—the experts and those in (formal and informal) power—for others. We might then well worry that some people exercise too much control over other people’s mental lives; that we neglect the task of empowering people to flourish through self-governance; and that we fail to acknowledge ordinary language users’ duty to reflect on their conceptual choices and thereby govern their own beliefs. So, what would conceptual engineering that respects epistemic autonomy as a right, an ideal, and a duty look like? I offer just one vision here. The key is a move from other-regarding to self-regarding conceptual engineering. While certain kinds of other-regarding conceptual engineering might also respect autonomy in some or even all the relevant ways, I am not concerned with those possibilities here. The vision of conceptual engineering offered here—I call it “participatory conceptual engineering”—means open, more or less rational discussion on how to shape the shared linguistic environment, where the participants are primarily aiming to govern their own beliefs by governing their own concepts, which are shaped by that shared linguistic environment.

I am thus also not concerned with self-regarding conceptual engineering that aims at engineering only one’s own concepts, without affecting the wider linguistic environment. Catarina Dutilh Novaes suggests that “explication exemplifies the ideal of epistemic autonomy, most famously captured by Kant’s Enlightenment motto: ‘Sapere Aude!’” (Dutilh Novaes, Reference Dutilh Novaes2020, 1018). And explication is often understood as a solitary project: a thinker defining how a term is to be understood in the context of their inquiry. But breaking apart from the linguistic environment like this is an importantly limited way of pursuing autonomy. First, one misses out on the relevant knowledge and perspectives of others, which one would encounter in a public debate on language. Second, as Eugen Fischer (Reference Fischer2020) and Edouard Machery (Reference Machery2021) have both argued, we have little direct control over our own concepts: often, the folk concept still silently guides our thinking in the background, post-explication. Machery and Fischer argue that this happens with concepts explicated for scientific and philosophical purposes. If this is indeed the case even in such contexts, where we are especially reflective about how we use language, then the project of revising concepts only for oneself and then consistently using these concepts unreflectively in one’s everyday cognition is surely a rather hopeless enterprise. Inevitably, we must often allow our minds to be guided by the concepts that we have internalized from our linguistic environment. But another form of responsible agency remains available and allows us to indirectly manage what happens in our minds when we are not directly in control. We can participate in shaping the shared linguistic environment that shapes our concepts, which then in turn shape our cognitive dispositions.

It is not an uncommon idea that we can best practice epistemic autonomy—govern our beliefs—through interactions and debate with other people. For example, Catherine Elgin argues that through encounters with others, we find out whether our commitments “can be sustained from other points of view” (Elgin, Reference Elgin, Matheson and Lougheed2022, 68) and that this is essential for epistemic autonomy as self-governance. A contrast between the present proposal and Elgin’s view is that according to Elgin (ibid., 62–65), we cannot really govern beliefs through exchanging perspectives and reasons with others. We can only govern our acceptances: our reflective intellectual endorsements. Elgin’s reasoning is that what she calls “belief”—feeling that p is so—is involuntary, and therefore, we cannot control it by self-discipline. I suggest, however, that we can to some extent govern, through participating in shaping our linguistic environment, also the sorts of beliefs or belief-adjacent phenomena that do not involve reflective, voluntary acceptance of a proposition. For example, we can to some extent govern our unreflective associations (by eliminating slurs from public discourse and thereby from individual cognition), our phenomenal reactions of condemnation (by engineering concepts like that of sexual harassment), our hopes and expectations for relationships (by engineering the concepts in the vicinity of love), and so on.

We need not get stuck in the debate on defining “belief.” The important point is that governing the relevant belief-adjacent phenomena is also plausibly our right, ideal, and duty. It is no more alright for others to exercise control over our unreflective associations and reactions (e.g., by engineering the linguistic environment) than it is for them to govern our reflective endorsements (e.g., by engineering the testimonial environment or planting belief-inducing imagery). Further, the ideal of self-governance as an aspect of the good life plausibly includes taking conscientious, rational control not just over our reflective endorsements but also over how we spontaneously see the world and the tracks that our thoughts tend to run on. And in so far as individuals can govern the unreflective belief-adjacent states, they plausibly also have the duty to do so.

Participatory conceptual engineering, then, means collective, open rational deliberation on how to improve the linguistic environment so that it would improve the concept users’ beliefs and belief-like states. Participation in engineering one’s linguistic environment does not have to take the form of writing an opinion piece or even a social media post (though it can take that form). One can participate in the public conversation in more low-key ways. For example, at a pub night with one’s male friends, when others say that sexual harassment can only happen at the workplace, toward an employee by their direct boss, one can ask for and engage with the reasons for using language in this way. It does not require great expertise or intellectual capacity to participate in a reflective and open-minded manner in some of the many conversations about language that happen around us all the time. Naturally, one cannot participate in all such conversations; those where one has relevant knowledge and those that affect one personally will have to take priority.

Participatory conceptual engineering respects epistemic autonomy as a right. The primary aim of the participants of such rational debate on concepts is not to govern other people’s beliefs; it is, first and foremost, a self-regarding enterprise, where people each govern their own concepts and thereby beliefs, by participating in shaping the shared linguistic environment. Further, the emphasis in this kind of conceptual engineering is on explicit, respectful rational debate, not on getting one’s preferred linguistic policies enforced, using the influence of celebrities or using social sanctions to change how people use language.

Participatory conceptual engineering is also a way of approaching the ideal of self-governance—of exercising rational control over one’s beliefs. Such participation allows people to gain some rational control over their unreflective associations, judgments, and feelings that are not under their direct voluntary control. When we encourage and enable participatory conceptual engineering, we are thus helping people to live a good life, along the dimension of self-governance. Furthermore, participatory conceptual engineering helps us to honor self-governance as each person’s duty. Participatory conceptual engineers do not take over other people’s job of governing their beliefs. They are doing their own duty of governing their own beliefs—often the unreflective, apparently involuntary ones that they cannot govern in a more direct way. They are also calling for others to do the same by joining them in the debate and participating in improving the linguistic environment. So, there is no worry about an improper division of labor—of ignoring the responsibility of those whose concepts are being engineered for them. There is no strict division of labor between those who design and popularize better concepts and those who passively absorb these concepts from their linguistic environment. Participatory engineers recognize that all people ultimately have the duty to govern their own beliefs and should therefore be reflective and discerning in their linguistic choices, and even participate in developing the shared language.

One might wonder: does participating in improving the linguistic environment really amount to a way of exercising control over our beliefs? Is this not a too uncertain und unstable way of influencing our beliefs—considering how little we each can in fact do, with the power of our arguments, to change the linguistic environment, and how much is up to others? This problem only arises, however, if we think that the exercise of rational control over one’s beliefs must not require hospitable external circumstances—that when we do need the help of other competent and benevolent agents for shaping our beliefs, then we are not really exercising control. We may well reject this individualist assumption about what it means to exercise rational control over one’s beliefs. However, our limited ability to shape the linguistic environment does mean that the duty to govern our beliefs by participating in shaping the linguistic environment must be understood in terms of reasonable efforts to make constructive contributions, given our circumstances.

The idea that it is generally desirable for the engineering of non-specialist language to proceed by means of explicit, rational discussion among all language users is not particularly controversial or novel. But I have hopefully made clearer the three distinct autonomy-related motivations for such a participatory approach. The natural question now is: how are we to move from the current situation to that desirable one? As things stand, people often try to subject others’ minds to their own will—by linguistic interventions or otherwise, sometimes benevolently and sometimes self-interestedly—and are not particularly interested in taking good care of their own beliefs. We want to achieve a rather different state of affairs: one where people are able and willing to manage their own minds through participating in shaping the shared linguistic environment in a rational manner. I have made clearer the reasons for heading to that destination; but I can only give very coarse-grained advice for reaching it. On the one hand, we need to put more effort, already in early education, into developing the will and the skills to engage respectfully with and learn from people who disagree with us about important matters. On the other hand, we need to make the infrastructures of public debate more egalitarian, while not allowing them to be overrun by those who have, after all, not developed the ability to engage respectfully and learn from others. The details are beyond the scope of this paper; but the emerging discussion on cultivating “receptive publics” (Habgood-Coote, Ashton, and El Kassar, Reference Habgood-Coote, Ashton and El Kassar2024) might help with tackling some of the relevant challenges.

Another natural question is: should we not defer to the experts on a given concept, instead of trying to contribute anything to the discussion ourselves? I am not recommending, of course, that everyone participate in developing the concepts of quantum physics—the proposal concerns concepts commonly used in everyday life in spontaneous cognition. But even when it comes to such concepts, one might argue, some have expertise that others lack. For example, people with disabilities or those who have thought deeply about the issue might be the experts on the concept of disability, and those who have experienced or done research on microaggressions might be the experts on the concept of microaggression.

In response: first, it is not entirely clear that, for example, able-bodied people who have not thought much about disability in the past would be participating beyond the boundaries of their expertise in the discussion of this concept. One issue here is how we understand “expertise.” According to Harry Collins, for example, “ordinary language-speaking, literacy and the like exhibit a high degree of expertise even though everyone has them” (Collins, Reference Collins2013, 256)—expertise need not be esoteric. Many people have insight into some effects of implementing a conceptual proposal—for example, that they personally find a neologism confusing or otherwise difficult to get on board with. Able-bodied people who are not steeped in research on disability arguably have a perspective not so easily available to the more obvious experts: the former, and not the latter, have likely experienced the anxiety of not knowing the currently accepted ways of speaking and the fear of inadvertently causing offense. For another example, someone potentially (or actually) accused of microaggressions has a relevant perspective which is less easily available to someone who is more likely to experience microaggressions.

Second, even where comparative non-experts’ active participation lowers the quality of the discussion, the importance of autonomy—of allowing, empowering, and requiring people to indirectly govern their own beliefs through participation in shaping the environment that shapes their beliefs—might outweigh this concern. Active participation, where one not only listens but also asks probing questions and formulates one’s own arguments, puts one in a better position to understand why certain concepts are better than others.Footnote ⁵ Someone who governs their linguistic behavior in the light of such understanding lives a better life, along the dimension of self-governance, and arguably complies with the duty of self-governance more fully, than someone who merely conscientiously selects whom to defer to about a given linguistic controversy.

5. Participatory Environmentalism

A broader lesson from the above discussion of participatory conceptual engineering is that there is an important and largely neglected way in which individuals can and should govern their own minds: participating in shaping the epistemic environment together with others through rational deliberation. We have seen that participation in improving the linguistic environment allows us to gain some rational control over beliefs (or belief-like states) that we cannot control more directly, by mere self-discipline. If people are to approximate the ideal and do the duty of self-governance also for such beliefs, then we need to think more about how individuals can participate in rationally shaping their epistemic environment. In other words, a certain kind of epistemic environmentalism—a focus on assessing and improving the epistemic environment—is needed. It is not the familiar kind of environmentalism that envisions some people, the experts (in a narrow sense) and those in formal or informal power, curating the epistemic environment for others; but it is nevertheless an environmentalism. We may call it “participatory environmentalism.” Far from being in tension with respect for epistemic autonomy, this kind of environmentalism is motivated by respect for epistemic autonomy (in the sense of self-governance) as a right, an ideal, and a duty.

One might wonder: how much of a departure from ordinary, expertise-based environmentalism is this, really? Maybe participatory environmentalism just calls for all experts to contribute to shaping the epistemic environment—even those who might themselves not realize that they have some relevant non-esoteric expertise? Indeed, Levy recognizes that “self-silencing by those who recognize their inferiority may lead to ‘hidden profiles’: information relevant to deliberation going unshared” (Levy, Reference Levy2022, 53). My call for broad participation in shaping the epistemic environment is motivated by respect for autonomy, however; so, it is not motivated by the need to include all the experts in the discussion. Further, in so far as participatory environmentalism is motivated by respect for autonomy, it allows and encourages participation in joint rational shaping of the shared epistemic environment even where such broad participation makes the process less effective from the perspective of values like truth and knowledge. But those who are not convinced by the autonomy-related reasons might also favor participatory environmentalism on the grounds that broad participation is required to include all the experts in designing the epistemic environment.

Somewhat similarly to the present proposal, Iizuka and Kobayashi (Reference Iizuka and Kobayashi2023) have defended public participation in improving the epistemic environment. They argue, commenting on Levy’s (Reference Levy2022) case for environmentalism, that in addition to better calibrating indicators of expertise with actual expertise, we also need to increase public participation in science, in order to restore trust in science and to ensure the transparency of nudging. Such transparency would protect autonomous choice as a right: “we must take seriously the importance of public engagement in controlling the system of nudges to respect our rational and autonomous choice” (Iizuka and Kobayashi, Reference Iizuka and Kobayashi2023, 835). I propose that public participation in shaping the epistemic environment is even more significant than Iizuka and Kobayashi suggest. Participation not only ensures the right not to have one’s beliefs subjected to others’ will (which, as I understand it, democratization of the choice/belief architecture should help with, according to Iizuka and Kobayashi); it also allows people to positively govern their beliefs themselves. Participatory environmentalism thereby allows each participant to approximate the ideal and do the duty of self-governance.

Iizuka and Kobayashi’s proposal also still retains the idea that the environment is ultimately to be designed “from above,” by those with special knowledge and capacities; the public should just be strategically included and engaged with in the process. Participatory conceptual engineering, as discussed above, is envisioned without such top-down control. On this approach, the targets of the engineering are themselves the engineers. It is perhaps not viable to apply this model of truly participatory epistemic environmentalism to governing the beliefs about climate change, for example. It is more applicable to what can be described as the “social imaginary”—like the stereotypes associated with women and men or certain ethnicities, or how we expect people to live and be, or what is considered good or bad behavior. (Self-governance as a right, an ideal, and a duty is plausibly especially pertinent to forming beliefs on such issues, and not so pertinent to forming beliefs about issues like climate change; why, exactly, will remain unanswered here.)

Certain kinds of top-down, expert-led epistemic environmentalism may also be needed, then, alongside participatory environmentalism. Fortunately, some top-down environmentalisms are plausibly compatible with respecting self-governance as a right and can support epistemic autonomy as an ideal and a duty. For example, we can engineer the environment top-down only in so far as this is vital to repair widespread misbelief that affects behavior in important ways (McKenna, Reference McKenna2023, 77). Or information can be presented in ways that help people to properly take in the evidence—thus supporting rational self-governance (McKenna, Reference McKenna2023, 99). Experts can also design the environment specifically to develop motivation and ability for rational self-governance. For example, Mia Karabegovic and Hugo Mercier (Reference Karabegovic and Mercier2024) discuss how to engineer an environment conducive to displays of intellectual humility—a virtue relevant for rational self-governance. So, I do not argue that respecting epistemic autonomy requires limiting ourselves to participatory environmentalism only. I have discussed this as one way for environmentalists to avoid the autonomy-related pitfalls—an option especially suitable for certain environmentalist undertakings, such as engineering the shared linguistic environment.

A further clarification concerns what I mean by participating in shaping one’s epistemic environment. Stephen Gadsby (Reference Gadsby2023) discusses something in the vicinity, yet distinct: choosing what people to surround oneself with or what other information sources to consult and how. Gadsby further proposes that how we shape our environment, in this sense, can sometimes be properly assessed for rationality. For example, Lyndon Johnson and George Bush, who distanced themselves from those who disagreed with them about political matters, “structured their environments in ways that led to bad beliefs and that kind of behavior seems irrational” (Gadsby, Reference Gadsby2023, 784). In order to understand how my proposal is distinct from Gadsby’s, it is helpful to distinguish between two ways of shaping one’s epistemic environment: changing one’s location in the epistemic landscape and changing (or contributing to changing) the shared landscape itself. We can and should also exercise self-governance by shaping our epistemic environment in Gadsby’s sense: by reflectively choosing what people to be around or what channels to watch and what forums to frequent—moving closer to sources that are likely to help us epistemically and navigating away from those that hinder us (e.g., joining or exiting Twitter/X or revising whom one follows on Twitter/X). All this can be aptly described as moving our own position on the epistemic landscape, without changing that landscape. Participatory conceptual engineering, however, exemplifies that we can also change the shared landscape itself, in collaboration with the others occupying that landscape. For another example: instead of just following or unfollowing people on social media platforms, we could improve the way those platforms work for us, from the epistemic perspective, together with other users. (This would, of course, require giving much more control over the platform to the users than is currently the norm.)

6. Concluding Remarks

I have argued that epistemic environmentalism, when pursued in certain ways, can present a triple threat to epistemic autonomy—as a right, an ideal, and a duty. One way to avoid those pitfalls is to opt for participatory environmentalism: a new focus on individual participation in improving the epistemic environment. This not only avoids violating people’s right not to have their beliefs governed by others but also helps people to approach the ideal of governing their own beliefs and thereby live a good life along this dimension. Furthermore, participating in improving the shared epistemic environment is a way in which people can more fully honor their duty of governing their own beliefs. I have illustrated this by showing how we can acquire some rational control over our unreflective beliefs by participating in shaping our linguistic environment and thereby our concepts, which in turn shape these unreflective beliefs. A new focus on improving the epistemic environment need not mean any disrespect to epistemic autonomy, then. On the contrary, such a shift of focus is required to facilitate governing beliefs that we cannot control in a more direct way.

It is by now a familiar point that we can better manage our own minds in interaction with others—learning from them and with their help. I have called attention to an underrecognized way in which we can and should govern ourselves together with others: by participating in improving the shared epistemic environment that shapes our beliefs. One important mode of such participation is contributing one’s perspective and insight into the ongoing conversations, or starting new conversations, on improving our shared language and thereby the tracks our minds tend to travel on.