1 Introduction

Evaluating and revising concepts lies at the heart of conceptual engineering as a philosophical method. Conceptual engineering rests on the possibility that certain concepts, at some point in time, turn out to be defective in some given respect, and that this defectiveness is recognized by at least some concept users. In turn, recognizing this defectiveness would be a first step towards revising the concepts at issue.

Importantly, all this could potentially happen with any concept in our representational repertoire; no portion of this repertoire is, in principle, exempted from evaluation and immune from revision. As Herman Cappelen puts it, there are no “safe spaces” in conceptual engineering (Cappelen 2018, p. 73). Considering this possibility, one may wonder: should the conceptual engineer be in principle ready to question all concepts? This is what Cappelen calls “representational skepticism” (Cappelen 2018, pp. 5–6): a refusal to “uncritically take over” the representational devices one is presented with in one’s cognitive life.

At least at first blush, representational skepticism seems to go well with conceptual engineering. Traditional forms of skepticism are often associated with a thinker doubting the justification for certain claims (for instance, knowledge claims) based on the relevance of alternatives (such as brain-in-a-vat scenarios) that could undermine our justification for those claims (for instance, the claim “I know I have hands”). Relevant alternatives that could undermine confidence in the good quality of our concepts play a role in the motivation for conceptual engineering too. For Cappelen, the thinker who realizes the general fact that “our representational devices can be defective in ways W1…Wn” is called to investigate whether these defects obtain before starting an inquiry (Cappelen 2018, p. 40). Similarly, the thinker who realizes that, in general, “there will typically be indefinitely many alternative meanings that would be better meanings [for a word W]” is called to “make sure our words have as good meanings as possible” (Cappelen 2020, p. 134). This sounds very similar to skeptical doubt about knowledge claims stemming from one’s attempt to rule out relevant “bad” alternatives. The only difference is that one tries to rule out relevant undesirable alternatives so that one can deem one’s conceptual representations of sufficiently good quality (in a number of respects and for certain purposes).Footnote 1

Entertaining doubts based on such general considerations might imply a particularly strong disposition to question conceptual representations every step of the way. Yet, how sustainable would this attitude be? Is representational skepticism really a good option? Are there any better alternatives? The aim of this paper is to clarify the implications of representational skepticism and to suggest an alternative, better approach, which is however still compatible with the demanding attitude a conceptual engineer would be expected to hold towards our concepts.

I start, in Sect. 2, by presenting representational skepticism and its opposite position, representational complacency, as Cappelen outlines them. Partly concurring with Cappelen, I stress how unsustainable each position (and even a hybrid between the two) is. In Sect. 3, I introduce an alternative view that I call critical concept conservatism. Section 4 proposes three arguments in favour of critical concept conservatism, while Sect. 5 is devoted to a final statement of the view. In Sect. 6, aspects that pertain to the position’s critical character are clarified.

2 Representational Skepticism and Complacency

In his 2018 book Fixing Language, Herman Cappelen relates conceptual engineering to what he identifies as an “underlying attitude” towards concepts and representational devices at large. He calls this attitude “representational skepticism”, and he characterizes representational skeptics as follows:

Representational skeptics do not uncritically take over the representational devices handed to them. […] When a representational skeptic starts reflecting on an issue, the first question she asks herself is whether the language used to articulate the key questions is good enough. (Cappelen 2018, pp. 5–6; see also p. 153).

Cappelen thinks of himself as a representational skeptic, but he also admits that practicing this kind of skepticism does not make one’s life easier. First, he notes, “[c]ommunication with others is hard, because we [representational skeptics] refuse to take their language for granted” (Cappelen 2018, p. 6). Moreover, representational skepticism quickly transmits at the higher orders, since “we representational skeptics need a language in order to engage in critical reflection about representations. But what about that language? Shouldn’t we be critical about that as well? The answer is ‘yes’” (Cappelen 2018, p. 6; see also p. 48). If the representational skeptic cannot avoid higher-order skepticism, then this project, “in its most extreme form […] will seem internally inconsistent. […]”. As Cappelen explains: “If the representational skeptic is disposed to be skeptical about the way questions are framed, how can she ever start inquiry? She would have to be skeptical of the very question she is asking about representational devices” (Cappelen 2018, p. 6). This, of course, would generate a vicious regress preventing the starting of any intellectual pursuit.

One may wonder why skepticism should carry over to higher-order representations. Can’t one simply stop at some point? To get a firmer grip on the problem pointed out by Cappelen, let us go back to the idea that the conceptual engineer doubts the representations they are presented with based on general considerations regarding “relevant alternatives”. This may imply doubting a certain conceptual representation unless suitably strong, positive reasons are given to deem it adequate. This creates the conditions for the problem to arise: if the conceptual engineer is to be consistent, they should apply the same general considerations at every level. If so, then they should doubt any representation at any level, on pain of inconsistency or lack of principled reasons for disregarding said general relevant-alternatives considerations. Thus, the combination of general relevant-alternatives considerations, demand for consistency, and rising of orders generates a problem, or at the very least a tension, which calls for a resolution.

One may try to escape the problem by avoiding any questioning, criticism, or doubt. This would lead to what Cappelen calls “representational complacency”, consisting in one’s “uncritically taking over” the representational devices one is handed (Cappelen 2018, p. 5). Representational complacency is not a viable option either, though, as it clearly precludes any possibility for progress in the realm of conceptual representations.

Cappelen describes representational skepticism and representational complacency as the two ends of a continuum. He also recognizes that one could be skeptical about some concepts and complacent about other concepts (Cappelen 2018, p. 6). Yet, if the conceptual engineer is susceptible to general considerations about relevant alternatives, it is difficult to see how the two positions could be consistently held – i.e., how a conceptual engineer could be skeptical in a certain area of discourse, and not skeptical in another area of discourse. Of course, a particular person who happens to be a conceptual engineer could, at some point in their life, simply care about a certain area of discourse (and be skeptical about it), and not care about other areas (thus being complacent about them). Yet, this might be due to that person’s (or that person’s community’s) particular circumstances, goals, or interests. Given that, at this stage, we are solely interested in the joint theoretical tenability of skepticism and complacency (as opposed to features like applicability, or viability for practical purposes), these aspects will not be considered. To assess whether the two positions could be consistently held across areas of discourse, we need a certain level of idealization. We should presume that the “ideal” conceptual engineer equally cares about all areas of discourse.Footnote 2 In this ideal situation, since the conceptual engineer believes that concepts could generally be defective, or have better alternatives, they should, for the sake of consistency, be equally skeptical about all these areas. If they were to be complacent in one specific area, this would seem unprincipled – and declaring “I don’t care about this area of discourse” would seem too easy a way out in such idealized environment. The result of these considerations is that, on a closer look, there seems to be no “hybrid” position between representational skepticism and representational complacency (that is, a partly skeptical, partly complacent position). Barring particular circumstances, goals, and interests that lead subjects to care about some areas and not about others, once representational skepticism is endorsed in one area of discourse, it threatens to spill over across all other areas of discourse within the same representational level. This would occur in addition to representational skepticism threatening to climb orders – as already pointed out by Cappelen. The problem, or at the very least the tension that it engenders therefore calls for our philosophical attention even more.

3 Is There a Third Way? Enter Critical Concept Conservatism

On the one hand, as we have seen, being a representational skeptic implies doubting or questioning any conceptual representation at any level, making it impossible for the subject to even start inquiry. On the other hand, representational complacency implies uncritically accepting any conceptual representation, making conceptual improvement impossible. Neither attitude is ideal, and a hybrid position does not seem viable either. Is there a third way?

In this section, I will outline a position that does not give rise to the difficulties described in the previous sections, but still upholds, in a relevant way, the idea of the conceptual engineer as someone who holds (or should hold) a demanding attitude when it comes to concepts. This position could be read as a refinement on Cappelen’s own representational skepticism, rather than an alternative to it. I am fine with this interpretation. What matters for the purposes of this paper is that it is overall a better position than Cappelen’s officially endorsed view in a number of respects.

Cappelen’s statement of representational skepticism seems to recommend doubting any conceptual representation before addressing any question or starting any inquiry. The reasons for such widespread doubting seem to be very general. As already noted, Cappelen mentions the possibility that our terms and concepts be defective in ways that we presumably cannot rule out, and the fact that “no domain of thought or speech is safe” (Cappelen 2018, p. 41). He also mentions that, when it comes to words’ meanings (which are the main focus of his conceptual engineering account) “we have no good reason to think that the meaning that we ended up with is the best meaning we could have” (Cappelen 2020, p. 134). There are “typically indefinitely many alternative meanings M*1…M*n that would be better meanings for [a word] W” (Cappelen 2018, p. 41). These could be read as very general reasons for doubting any concept at the outset of any inquiry. Cappelen’s sweeping representational skepticism could be formulated as the following meta-conceptual claim: call it “Skeptical Meta-Conceptual Claim” or “SMC”. For any concept K, SMC reads as follows:

[SMC] Concept K is to be questioned (for reasons R1…Rn).

where reasons R1…Rn may include the general points mentioned above. SMC seems to capture Cappelen’s idea that doubting conceptual representations is the “first thing” one should do when approaching a certain subject.

By contrast, the position I wish to outline is articulated on two separate levels, and combines two different elements. The first is a “conservative” approach, whereby the thinker “takes over” the representational devices handed to them, assuming that, if they have the concept in question, then it is rational for them (due to a range of reasons to be explored) to retain that concept. This could be captured in terms of a “Conservative Meta-Conceptual Claim” or “CMC”. For any concept K, CMC reads as follows:

[CMC] The fact of having concept K makes it rational to retainFootnote 3 K (for reasons R1… Rn).

I will delve into some very general reasons R1… Rn for CMC in a later section. For now, note that CMC is clearly in contrast with representational skepticism’s immediate doubt. CMC could, however, still lead to representational complacency. A second element is needed, to be articulated on a different level. This second element may be related to a thinker’s critical attitude expressing their sensitivity to the potential ways in which concepts may fail us, so to speak. This could be captured in terms of a meta-meta-conceptual claim that considers the possibility of doubting the rationality of retaining a certain concept. Call this the “Critical Meta-Meta-Conceptual Claim” or “CMMC”.

[CMMC] Were appropriate groundsFootnote 4 to hold (at a context and a time), it may be questioned whether it is rational to retain concept K.

Examples of appropriate grounds for doubting a concept are easily found in the conceptual engineering literature. I will say more on this towards the end (see Sect. 6). Right now, I am just interested in outlining the overall position.

I propose that the conjunction of CMC and CMMC results in what one may call a critical concept conservatism, whereby an inquirer accepts, albeit with a critical attitude, the representational devices offered to them by the extant representational repertoire. Note that CMC and CMMC are presented as two separate claims, holding at different levels. Another way of articulating critical concept conservatism is by means of a general statement, which may read as follows: “For any concept K, the fact of having K makes it rational to retain K (for reasons R1… Rn), unless there are grounds, at a context and time, to question (whether it is rational to retain) K”. From this statement, one cannot immediately glean that the conservative component (“having K makes it rational to retain K”) and the critical component (“unless there are grounds to question…”) hold at different levels and for different reasons. This difference can be better appreciated if the two claims are considered separately; this will help us to better tease out each claim’s motivation, as well as to compare the proposal more clearly with Cappelen’s view. The general statement will become more important later on.

The mutual relation between the conservative claim (CMC) and the critical claim (CMMC) is central for critical concept conservatism. To understand this relation, recall, first, the following assumption: each claim would presumably be supported by a range of reasons – which I will cover in later sections. For now, let us assume that some such reasons exist, and that they offer a certain degree of support to each claim. The two claims are related as follows. On the one hand, if the reasons for the conservative claim regarding a concept K are – at a certain context and time – stronger than the reasons for the critical claim, then it will be rational for one to retain concept K. If, on the other hand, the reasons for the critical claim regarding a concept K were to become – at a certain context and time – stronger than the reasons for the conservative claim, then it will become appropriate for one to question whether it is rational to retain concept K.

The most apparent difference from Cappelen’s representational skepticism is that critical concept conservatism does not demand questioning every concept in every context whatsoever. Whether it is rational to retain or to question a concept K depends on the strength of the reasons that militate in favour of retaining it or of questioning it – something to be evaluated in specific contexts. This means that the critical component need not always dominate over the conservative one. Rather, the conservative component generally counterbalances the critical element,Footnote 5 preventing the latter from blocking or severely hindering any attempt at inquiry, as well as from spreading across one’s conceptual repertoire and over at the higher orders. In other words, the conservative component avoids what were described in Sect. 2 as the most displeasing consequences of representational skepticism. Yet, at the same time, the critical component prevents the position from sliding into a form of representational complacency, thus making it suitable for a thinker who holds a demanding attitude towards concepts – as a conceptual engineer does (or should do).Footnote 6

To complete this preliminary presentation of the view, let us see how it applies to a concrete example. Consider the concept planet. Up until discovery of the Kuiper belt in 1992, which led to the observation of a multitude of celestial bodies very similar to planet Pluto (see Brown et al. 2005), astronomers were not questioning this concept. Perhaps they were aware of some of its semantic defects – such as its being indeterminate, or poorly defined. These flaws would have been enough for representational skepticism to enjoin questioning the concept, even if it was deemed “good enough” for the purposes of the astronomical community. That’s because “better alternatives” could have been employed.

Critical concept conservatism can describe the situation differently. Prior to discovery of the Kuiper belt, the astronomical community’s reasons for being conservative towards planet were overall stronger than their reasons for questioning the concept. After such discovery, however, the situation changed. They had a stronger reason for questioning and revising the concept planet than they had for retaining it (the reason being, roughly, that including the Kuiper belt objects into the planet-category would indicate a failure of the concept planet to sharply delimit what is, and is not, to fall into its extension). This makes sense of their keeping the concept pre-discovery and of their wanting to improve the concept post-discovery. By contrast, for the representational skeptic, their pre-discovery conservative attitude would have been misplaced all along. Yet, this judgment would have been too harsh; it would have called for doubt where there was no need for doubt, thus potentially undermining several intellectual pursuits.

This section was devoted to presenting critical concept conservatism in broad outline, as well as to comparing-and-contrasting it with Cappelen’s representational skepticism. The next section further develops and qualifies the proposal.

3.1 Being Conservative and Being Critical: Qualifications and Connections

The position’s core components could use some further qualifications and refinements. Let us start from the conservatist component. The label “conservatism” is not chosen randomly. It is indeed inspired by belief conservatism, a position known from classic debates in epistemology. According to belief conservatism, the fact of holding a belief makes one epistemically warranted to hold it, unless there are grounds to doubt it (see Quine 1951, Sklar 1975, Chisholm 1980, Foley 1982, Harman 1986, Lycan 1988, Kvanvig 1989). Belief conservatism can be invoked, among other things, as a way of resisting well-known forms of skepticism. Consider the belief that there is an external world. The Cartesian skeptic tells us that we are not justified in holding that belief, since we cannot rule out a skeptical hypothesis (e.g. an evil demon hypothesis, a brain-in-a-vat hypothesis). The conservatist can argue that we do have such justification, since we have this belief already (indeed, we’ve had it for a long time), so there is a presumption in its favour (see Kvanvig 1989, pp. 147–149; McCain 2008, pp. 189–190)Footnote 7.

Analogously, critical concept conservatism could be seen as contrasting representational skepticism, which enjoins doubting any concept whatsoever. Against this position, the critical concept conservatist may draw inspiration from belief conservatism and argue that, since we have a concept, this fact (modulo appropriate qualifications) makes it rational for us to hold it, unless there are grounds to doubt it. By requesting grounds for questioning the concept at issue, the critical conservatist blocks the road to indiscriminately doubting any concept from the get-go.

Let us now move to the critical component, which is expressed by the “were appropriate grounds to hold…” clause in CMMC (and in the “unless there are grounds…” clause in the general statement). Note that there is an intellectually “passive” and an intellectually “active” way of accepting these clauses. On the one hand, one who is “actively” critical displays (to an appropriate level) an ability and a readiness to spot conceptual flaws and to question a conceptual representation, should this become necessary. This requires a certain level of “intellectual vigilance” and first-person involvement in finding grounds for questioning a concept. Being an active critical conservative does not mean questioning every conceptual representation just because a concept could be flawed, or could be improved, like the skeptic would do. It just means being ready to question a conceptual representation, were there grounds for doing so. One has therefore to develop a sensitivity to potential grounds for questioning, together with a capacity to understand when certain conceptual flaws could be tolerated, and when they should prompt a critical reaction. Even if the critical concept conservatist were to attribute a flaw to a concept, this may not yet constitute grounds for doubt, if they were to think that such a flaw can be tolerated. On the other hand, someone who is more “passively” critical simply endorses the “were appropriate grounds to hold…” clause but need not hold a first-person investment into finding grounds for concept-questioning. Of course, one could be a bona fide critical concept conservatist even if one held a more passive attitude, as long as one indeed doubted a concept, were one presented with grounds to do so. The profile of an active critical concept conservatist just seems to best fit with the profile of a conceptual engineer, who is supposed to care about spotting conceptual defects and improving conceptual representations. Meanwhile, the profile of a passive critical conservatist would seem to better fit with a layperson, who is not actively invested in a project of conceptual improvement.

Compare now critical concept conservatism with belief conservatism, which also includes the “unless there are grounds to doubt” condition. To my knowledge, none of its versions explicitly mentions whether this condition should be accepted with a vigilant, active attitude or with a more passive attitude towards potential grounds for doubt. Belief conservatism seems, therefore, less specific than critical concept conservatism, even if it does include a critical element spelled out in terms of the “unless there are grounds to doubt” clause.

Let us now consider how critical concept conservatism can potentially bear connections to other accounts dealing with concept preservation. The first of such accounts is about trust in our concepts. Suppose it is rational for one to retain a concept one has. Then, it seems one will also be entitled to trust one’s use of that concept (provided one is a competent concept user). We can therefore draw a link between critical concept conservatism and the idea of trusting our own concept use. Self-trust is highlighted by David Plunkett and Tristram McPherson as a response to a global form of skepticism about normative concepts. Plunkett and McPherson write: “we are entitled to normative-conceptual self-trust: that is, […] we are entitled to treat our own normative concepts as acceptable starting points for the conceptual ethics of normativity” (Plunkett and McPherson 2021, p. 223). Indeed, critical conservatism may be used to justify conceptual self-trust about concepts in general – not just normative concepts, which are Plunkett’s and McPherson’s focus.

A second account has to do with concept preservation as a condition for conceptual reform. Think of Neurath’s image of a sailor rebuilding their boat plank-by-plank on the open sea. This image illustrates treating some concepts as settled, while we revise other parts of our conceptual scheme. Critical concept conservatism may supply the rationale for keeping concepts settled: the conceptual engineer has – at the relevant context and time – sufficient reasons for holding said concepts fixed, and has insufficient reasons for questioning them. (Of course, in a future situation, they may acquire strong enough reasons to question the concepts that were previously kept settled. This may lead them to revising one or more of those concepts, while holding yet other concepts still.) This way of putting things seems to aptly spell out, in a non-figurative fashion, the image of Neurath’s boat. If so, it provides (at least the beginning of) a response to a point in this regard made recently by Matthew Shields (2021), who criticizes the Neurathian image for being too vague.Footnote 8 The relation established by critical concept conservatism between one’s reasons for the meta-conceptual claim (CMC) and one’s reasons for the meta-meta-conceptual claim (CMMC) goes at least some way towards clarifying the Neurathian image, thus helping dispel Shields’s worries.

Given these considerations, an exploration of the argumentative foundation for critical concept conservatism is called for, not just to ensure that the view is adequately motivated, but also because this position may in turn provide the foundation for other related accounts. The next section focusses on some of the main argumentative options available.

4 Why a (Critical) Conservatism About Concepts?

This section reviews considerations that one might invoke to motivate critical conservatism about concepts. (i) The first is an argument “by comparison” between critical concept conservatism, representational skepticism, and representational complacency. (ii) The second is an argument linking (critical) belief conservatism to critical concept conservatism. (iii) The third is a series of arguments focussing specifically on the conservative component of the view (CMC), presented independently of conservative accounts in the realm of belief.

4.1 The Argument by Comparison

There is a very straightforward argument in favour of critical concept conservatism: it avoids the drawbacks of skepticism and of complacency. On the one hand, unlike representational skepticism, it does not imply systematically questioning every representational device; this enables the start and unfolding of inquiry. On the other hand, it allows for progress, as the critical element plausibly also involves openness to conceptual improvement. These considerations should at the very least make critical concept conservatism attractive by comparison. One may wonder, though, if there are any non-comparative considerations in favour of critical concept conservatism.

4.2 The Argument from (Critical) Belief Conservatism

To start with, it is interesting to explore to what extent critical concept conservatism would be attractive for someone who already endorses (a critical form of) belief conservatism. For suppose there is a case for (critical) conservatism about beliefs. It is commonly accepted that concepts compose beliefs, that is, concepts’ tokens occur in the propositions that make up the content of our beliefs. If so, it would also seem legitimate to think that, if there is a case for (critical) conservatism about beliefs, it extends to whatever composes beliefs. Since concepts compose beliefs, it would follow that there is a case for (critical) conservatism about concepts. Given this reasoning, for someone who already endorses (critical) belief conservatism, it should be straightforward to also endorse (critical) concept conservatism.

Perhaps, though, the road from belief to concept is not that direct. There may be a few complications, arising from the fact that there are several different ways of spelling out concepts and propositions. Suppose concepts have sense and reference. Then certain concepts might contribute the same referent to a proposition, but with different senses. Conservatism may extend from belief to concept as far as reference is concerned, but not as far as sense is concerned; or vice-versa. For example, take the proposition that the morning star is shiny; one might want to simplify one’s conceptual repertoire, and thereby eliminate the concepts the morning star and the evening star. Yet, one may want to keep a concept that refers to that very celestial body (for instance, the concept venus), as well as wanting to still formulate thoughts about that celestial body (for example, the thought that Venus is shiny). One would then be a conservative about the belief and the relevant concept as far as the aspect of reference is concerned, but not when it comes to sense. Vice-versa, consider the proposition that Santa brings gifts. Knowing that the concept santa is referenceless, one may want to be conservative about the belief’s and the concept’s sense, not about the aspect of reference.

Given these considerations, then, if there is a road from (critical) belief conservatism to (critical) concept conservatism, it is probably more winding than one would have initially expected. Still, there are reasonable prospects for the (critical) belief conservatist to accommodate these extra complications.

4.3 The belief-independent Argument: Positive Reasons

Now, what if one does not endorse (critical) belief conservatism? Are there any reasons – be they epistemic, practical, or moral – to be (critically) conservative about concepts independently of beliefs? To address this question, it may be useful to start from a widely shared assumption: concepts are “devices” or “tools”, that is, artifacts that are used to perform a certain function – typically but not exclusively, a representational one. They are used to build more complex representational (or non-representational) contents, which in turn play different roles in communication and in inter-subjective coordination.

We should therefore start with the following general question: why should one want to be conservative about such tools? In what follows, I will articulate very general positive reasons for the conservative component (CMC) of critical concept conservatism. Their aim is to support the conservative side at least until suitably strong considerations for the critical component (CMMC) kick in. These will therefore be pro tanto reasons, liable to be overridden by considerations that push in the direction of doubt and critique.

First of all, a positive reason for accepting (most of) our concepts by default is, quite simply, that they are already there and, usually, they have been there for a fairly long time. They are, to borrow a term from Nelson Goodman (1955), entrenched in our linguistic and reasoning practices. The fact that those concepts are available to us, that they have been available from the past, and that they are successfully used by us, as well as by our fellow speakers, creates a presumption in their favour. This provides pro tanto reasons – whether epistemic, practical, or moral – to retain them, unless there are grounds to doubt them. Thinking back to the planet example, we may say that one pro tanto reason the astronomical community had to keep using that concept was simply that it had been around, and it had served its function to a satisfactory degree, for long enough.

To forestall confusion, let me stress that I am not urging that “a concept’s existence suggest its excellence”. First, it is not a concept’s existence that matters, but rather its existing and being entrenched, i.e., suitably embedded in our thought and speech practices. Secondly, neither existence nor entrenchment suggest a concept’s “excellence”, just its being “fit enough for the job”. Thirdly, to reiterate, entrenchment only provides pro tanto reasons for keeping a concept around; it is not supposed to be a fully conclusive proof, only a positive consideration in favour of conservatism. The fact that the reasons in question are only pro tanto is, to my mind, a feature and not a bug. If critical concept conservatism relies on a delicate balance between a conservative component (CMC) and a critical component (CMMC), one would expect the reasons on both sides to illuminate only part of the set of benefits and inconveniences that come with keeping or questioning a concept. If reasons were too strong on either side, one of the two components would be overly dominant, thus compromising the balanced, contextually sensitive character of the view.

Of course, this appeal to entrenchment may open up a range of issues: is a concept’s being entrenched really sufficient for presuming its adequacy? What about entrenched concepts that are nevertheless extremely detrimental to inquiry, communication, and inter-subjective coordination? Think about, for example, concepts linked to slurs (e.g. the concept of savage, or of slut); or concepts that obscured certain theoretically fruitful distinctions (e.g. the concept of phlogiston). These concepts were, or are, entrenched, but they are either suboptimal or straightforwardly damaging. Their entrenchment may be deemed insufficient even as a pro tanto reason for keeping them around. One way of confronting these issues is to further qualify the idea of entrenchment. To qualify as a pro tanto reason, entrenchment should be combined with a concept’s having an adequately reputable history or “career”, either from an epistemic or from a non-epistemic (e.g. practical, moral) point of view. Single subjects need not be aware of this history’s details, although it might help if they had an inkling that there is a history.

If conservatism requires not just concept entrenchment, but a reputable career too, then “bad but entrenched” concepts like the ones listed above would likely turn out as not worth preserving. Plus, different histories or careers may be relevant for different types of concepts. Take, for instance, theoretical concepts, or concepts that are derivative from more basic concepts, as for example mass, or gross domestic product. It should be possible to tell an empirical story that explains why they are entrenched and have a reputable career, specifying their role and function within our scientific and discursive practices. As to non-theoretical concepts, or basic concepts, such as the concepts good, exist, or I, one could explain their entrenchment and reputable career by invoking non-empirical considerations, for instance by using transcendental arguments, or genealogical considerations. The most difficult case is offered by concepts that are evaluatively or normatively charged in ways that are controversial (the most divisive cases nowadays would probably be gender and race concepts, such as woman or black), for it will be notoriously difficult to establish whether their career is “reputable” in a relevant way. In these cases, concept users may be employing concepts that they really have grounds to doubt. The lesson to be learned, here, is that the critical concept conservatist should accept certain epistemic limitations. Entrenchment may be a good (pro tanto) indicator for concept adequacy up to a certain point, after which one simply runs the risk of deploying a flawed concept. All one is left with, then, is the hope that the concept’s flaws will sooner or later be exposed.

4.4 The Belief-Independent Argument: Negative Reasons

Let us move to further, negative reasons to retain concepts, having to do with the negative consequences of doubting them. These should be read as pro tanto reasons that back up one’s conservative attitude towards a concept (spelled out by CMC). They may strike the reader as “weak”, but, as already noted in Sect. 4.3, they may be as strong as they could get to provide a general underpinning for the conservative component of critical concept conservatism. Such component, as one may recall, is meant to co-exist with the critical component (CMMC), but the two components will come into tension on occasions. The pro-conservatism reasons could therefore be trumped, if stronger reasons in favour of questioning a concept were to arise.

What general (pro tanto) negative reasons could therefore motivate keeping a concept rather than questioning it? First, questioning a concept would require rising at the meta-level; this is generally not practical or efficient, and it could get in the way of truth-oriented inquiry. When we communicate and reason, we rely on concepts so that communication and reasoning flow seamlessly. If we had to question one or more of these concepts, we would have to pause our object-level train of thoughts or our object-level flow of communication, rise at the meta-representational level, and examine the concept’s relevant features. This would cause a disruption in our reasoning or communication, which, absent strong enough reasons for questioning the concept at issue, could be deemed detrimental to their proper unfolding, and could eventually hinder us in the pursuit of epistemic goods such as knowledge or truth. There are, therefore, practical and perhaps even epistemic reasons (in a suitably broad sense of “epistemic”) for holding on to our conceptual resources unless positive, suitably strong grounds arise to doubt them.Footnote 9

One may reply that it is possible to question a concept without rising at the meta-level. For instance, I could question the concept of torture by engaging with object-level contents, stating things like “I think this is torture!”, “What exactly is torture?”, and so on.Footnote 10 I would say that speech acts like these only indirectly and incompletely engage with the concept of torture, by rather directly engaging with the property of being torture. The speaker who questions what torture is, is also simultaneously questioning applications of a concept, and is also, in a sense, pushing for modifications in the application of that concept. Yet, such a speaker is not saying anything about that concept; she is not talking about any of its semantic or non-semantic features. If she does have problems with the concept itself (besides having problems with the property it denotes), she is not making it explicit, or she is communicating it in an elliptical and incomplete way. If she wanted to communicate explicitly about that concept, though, she would have to rise at the meta-representational level, thus disrupting the first-order flow of communication. At the very least, then, explicitly questioning concepts requires going meta, even if implicitly questioning them does not; but then again, implicit questioning is a less optimal form of questioning.

Second, we often don’t precisely know the history, genealogy, or function of our concepts. The fact that we have these concepts, and that others around us successfully use them largely in sync with ourselves, is enough to create a presumption in their favour. In ordinary circumstances, absent strong reasons to do so, nothing or too little would be gained in trying to probe them, by either retracing their causal history, investigating their genealogy, or testing their function. That would take too much time and effort, and it could get in the way of truth-oriented inquiry.Footnote 11 (Note that this would seem to hold even if probing and testing our concepts did not require rising at the meta-representational level.) As a default position, then, it might be practically and epistemically reasonable (in a broad sense of “epistemically”) to accept those concepts, unless of course positive, and suitably strong grounds emerged that warranted our doubts.

Third, concepts often seem connected, forming clusters or networks. This seems intuitively to be the case (pace atomism!), once one considers that defining one concept often requires other concepts, whose definitions in turn need further concepts, and so on. This indicates that at least some concepts are connected to each other, forming several, and perhaps in some cases overlapping networks – in keeping with a molecularist view of concepts. Questioning one concept, then, may imply questioning others, with potentially unpredictable consequences, both practical and epistemic. Revision could give rise to unforeseen domino effectsFootnote 12 that, in normal circumstances and absent special reasons to do so, would excessively disrupt inquiry or communication, as well as the pursuit of epistemic goods like knowledge and truth. These considerations make it rational (practically and epistemically, in a broad sense of the word) to hold on to our actual concepts, unless there are strong enough reasons to doubt them.

Far from being exhaustive, these positive and negative points illustrate the kind of considerations that could motivate being (critically) conservative, independently of belief conservatism and under the concepts-as-tools analogy. The analogy with tools seems fruitful: for tools as well, there seems to be a presumption of adequacy unless grounds emerge for doubting it. Of course, concepts aren’t just any tools. They are thought to principally serve a representational and epistemic function, although they have normative, practical, moral aspects as well. Critical conservatism about concepts should reflect this complexity.

5 A Broad-Spectrum Position

In light of the previous points, critical concept conservatism should not be thought as a purely epistemic position. It should, in this respect, diverge from belief conservatism. The latter is (or is purported to be!) an epistemic position, whereby having a certain belief makes one epistemically justified in having it. This is because beliefs are thought of as representations aiming at justification or truth. By contrast, concepts qua representational tools need not be directly recruited in the achievement of epistemic goods such as justification or truth. They can fulfil a variety of functions, such as categorizing, distinguishing, systematizing, evaluating, promoting or suppressing values. In view of this functional plurality, critical concept conservatism should be conceived as a broad-spectrum position. Whether it is rational to retain (or question) a certain concept need not be just a matter of epistemic rationality; it can be also a matter of practical and moral forms of rationality. The position’s general statement (already mentioned in Sect. 3) should therefore allow for a broad sense of “being rational”, which can be sensitive to (for instance) epistemic, practical, and moral considerations. Accordingly, the view may be stated in the following terms:

Critical Concept Conservatism (CCC)

For any concept, K, the fact of having K makes it rational to retain K (for reasons R1… Rn), unless there are (epistemic, practical, moral) grounds, at a context and time, to question (whether it is rational to retain) K.

This general statement can be adapted to particular cases. Thus, for example, we may say that, for a long time, the fact of having a planet concept – with a certain degree of entrenchment and career – made it rational for the astronomical community to deploy it without questioning it. With the 1992 discovery of the Kuiper belt, a mix of epistemic, “zetetic”, and practical grounds became relevant that pushed the community towards questioning the concept. Turning to a different example, we may also say that, given a certain historical context, the fact of having the concept woman (as in “adult human female”) – with a certain degree of entrenchment and career – has made it epistemically and morally rational for us to use that concept. Sally Haslanger (2000), though, has championed the view that the concept of woman is, at some level, linked to the oppression of women in many areas. If so, that would constitute grounds for doubting the epistemic or non-epistemic good quality of this concept.

6 Doubting our Concepts: Being Critical and Being Skeptical

I have devoted much attention to general reasons in support of the conservative component (CMC) of critical concept conservatism. This was motivated dialectically by the need to distance the position from Cappelen’s representational skepticism, which seems to starkly go against conservatism. In this section, I will briefly focus on the critical component of critical concept conservatism (CMMC), also encapsulated in the clause “unless there are (epistemic, practical, moral) grounds, at a context and time, to question (whether it is rational to retain) K”.

The presumption of a concept’s adequacy (stemming from CMC and the reasons motivating it) could be defeated by a variety of features – many of which already familiar from the conceptual engineering literature. These include, to name but a few, a concept’s inconsistency (see Eklund 2002, Scharp 2013), its vagueness, indeterminacy, open texture (see Williamson 1994 on vagueness, and e.g. Tanswell 2018, who highlights the open texture of set-theoretic concepts), but also lack of extension (see e.g. Appiah 1996 about the lack of extension of the concept race). A concept may be deemed defective if it does not “carve nature at its joints” (Sider 2011), or if it fails to give rise to fruitful explanations (Pérez Carballo 2020). Furthermore, a concept could give rise to an ethically questionable, directly or indirectly harming categorization, over which doubt could be cast (see Goetze 2021, Marques 2020).

Reading this list of conceptual flaws, one may point out that grounds for genuine doubt seem very easy to come by. Nearly any concept, for example, seems vulnerable to a charge of indeterminacy, open-texture, vagueness, or lack of exactness. If legitimate doubt concerning a concept is this easy to come by, then why isn’t that a good motivation for representational skepticism? In response to this observation, I would be inclined to say that all these conceptual flaws provide grounds for being critical (whether actively or passively), rather than skeptical. Given how the two positions have been articulated, there is a difference. The skeptic is disposed to question concepts because of general considerations regarding the presence of flaws and of better alternatives. The critical conservatist has a more nuanced position, which is based on weighing reasons in favour of the conservative claim (CMC) and reasons in favour of the critical claim (CMCC). The conservatist may be aware that most concepts are vulnerable to charges of indeterminacy, open-texture, vagueness, lack of exactness, and so on. Yet, they may want to distinguish cases in which these flaws are to be tolerated (thus leaning on the conservatist side), and cases in which these flaws should lead to questioning and perhaps reform (thus leaning on the critical side). When flaws are tolerated, one is not being fully complacent – because these flaws are at least acknowledged. This seems enough to make the concept user critical, rather than skeptical.

Another observation that seems to point in the direction of skepticism is the following. We have historical examples of whole conceptual schemes proving defective: think of the conceptual systems related to alchemy, Ptolemaic astronomy, or slavery. For all we know, we also may be using a bankrupt conceptual scheme (after all, thousands of people have done so before us). If so, we should ensure that we aren’t. This sort of “pessimistic meta-induction”, though, seems to be a plausible motivation for something closer to representational skepticism, rather than for critical concept conservatism. Once again, I think these pessimistic considerations provide good reasons for being critical rather than skeptical. While the skeptical is disposed to be suspicious no matter what, the critical concept conservatist can admit that most of our conceptual systems could be defective while still being “good enough” at least until their flaws become intolerable, or a better set of concepts comes up. This prevents the skeptical component from always dominating on the conservative one, thus setting apart the view from representational skepticism.

In closing, let us consider that, sometimes, it may be unclear which component should prevail over the other. We have mainly considered occasions (like the planet case) in which the reasons for questioning a concept become stronger than the reasons for retaining it – because of some discovery, or some re-evaluation of that concept. Yet, it could be that the evidence the conceptual engineer garners in support of the critical component (CMMC) is controversial, or underdetermined. What to do then?

The answer, presumably, will vary from case to case. First, one might stick to conservatism. After all, one might reason, if there are grounds to doubt the very defeaters that should prompt concept-questioning, maybe the most reasonable thing to do is to just retain the concept(s) under examination. A second option is that of retaining the relevant concepts, while at the same time remaining on the lookout, so to speak, for new elements that might bolster the defeating evidence so far gathered. This attitude would clearly embody the spirit of critical conservatism, yet it might be difficult to put it into practice. One might too easily slip into just retaining the concept in question, while forgetting about the need to look for new evidence. Or, vice-versa, one might get too involved in the task of looking for new defeating evidence, suspending one’s use of the concept. A third option is that of trying to invoke considerations that are as much as possible independent of the current evidence, and that could eventually tip the balance in favour of questioning and possibly revising a certain concept as opposed to retaining it. These might be normative considerations, for instance considerations that pertain to the goods and values that would accrue upon questioning and possibly revising the concept(s) under examination. This exercise is acknowledged by Alexis Burgess and David Plunkett (2013) as part of the method they call “conceptual ethics”, that is, the investigation of which concepts we should use, and why.

There is no ready-made recipe for how to overcome contestability or underdetermination in the case of defeating evidence for concepts. Whether something counts as evidence that a concept is defective in a certain respect is, partly, an empirical hypothesis. As such, it is fallible and revisable. The three strategies just sketched provide guidelines for approaches that acknowledge the empirical, fallible and revisable character of claims about concept defectiveness.

7 Concluding Remarks

Conceptual engineering has been linked by Herman Cappelen to “representational skepticism”, that is, the refusal to uncritically take over the conceptual representations one is handed. As it turns out, however, representational skepticism is not a sustainable position. Representational complacency, the opposite of skepticism, is perhaps even less advisable, plus, there seems to be no rationally sustainable partly skeptical, partly complacent, “hybrid” position.

This paper was devoted to outlining an alternative to all these options, called critical concept conservatism. Critical concept conservatism states that having a concept K makes it rational (in a broad sense of the term) for one to retain K, unless there are (epistemic, practical, moral) grounds to question K. It could be viewed as the conjunction of two claims, one stating that having a concept makes it rational to retain it (CMC), and the other stating that, were appropriate grounds to hold, it may be questioned whether it is rational to retain that concept (CMMC).

Unlike the representational skeptic, the critical concept conservatist does not doubt concepts from the very start of inquiry. Rather, they accept and retain those concepts, unless there are grounds for questioning them. In the ideal case, the critical concept conservatist is also actively vigilant for conceptual flaws, although they also have the option to “passively” accept the “unless there are grounds to question…” condition. Furthermore, precisely because they endorse the latter condition, the critical concept conservatist diverges from the representationally complacent inquirer, in that they are open to the possibility of discovering conceptual flaws.

Critical concept conservatism avoids the drawbacks of representational skepticism and complacency. It seems to straightforwardly follow from (critical) belief conservatism – with a few manageable adjustments. Furthermore, its conservative component (CMC) seems to be independently bolstered by a number of positive and negative considerations. On the positive side, there are pro tanto reasons to retain a concept that is adequately entrenched and has a reputable (epistemic or non-epistemic) career. On the negative side, it seems inconvenient, in normal circumstances and absent strong reasons to do so, to question an entrenched concept (by rising at the meta-level, inquiring into its history or function, or querying neighbouring concepts), as that would potentially jeopardize communication and inquiry. All these are pro tanto reasons for keeping a concept around, which could be overturned were adequate grounds for questioning a concept to arise.

As to the critical side of critical concept conservatism, we know from the conceptual engineering literature which features could defeat the presumption regarding a concept’s adequacy. These problematic features may seem too easy to come by, or they may spark associations with a “pessimistic meta-induction”. These aspects do not, however, favour skepticism over simply maintaining a critical attitude. Finally, sometimes the evidence supposedly backing a critical attitude may be controversial or underdetermined. Since claims as to a concept’s defectiveness, and evidence thereof, are (at least partly) empirical, and thus fallible, this is a limitation the critical concept conservatist seems to have to live with.