Conceptual engineering and conceptual innovation

ABSTRACT

Conceptual engineering, motivated one natural way, involves the search for new concepts. But to what extent does conceptual engineering as practiced involve such conceptual innovation, and to what extent can it do so? In this paper, I first argue that conceptual engineering as practiced surprisingly does not appear to involve conceptual innovation, and then I discuss problems regarding the extent to which it can involve conceptual innovation.

KEYWORDS:

• Conceptual engineering
• Quinean bootstrapping
• concept acquisition
• alien languages
• concepts

1. Introduction

In this paper, I discuss conceptual engineering. ‘Conceptual engineering’ is a broad label. I suspect that one of the reasons for the recent popularity of talk of conceptual engineering is precisely that it can encompass many things. What I will focus on is conceptual engineering as I prefer to see it. I will motivate it, and describe a central challenge that it faces. To some extent, this is different from conceptual engineering as others prefer to see it. But comparison with other conceptions of conceptual engineering will not be a main theme.

Here is my preferred way of describing and motivating conceptual engineering:

The concepts we have and employ are only some of all the concepts there are. There are many other possible concepts, including concepts very different from the ones we have and employ. There is a project of (i) figuring out what other concepts there are, (ii) assessing how good they are (compared to our actual concepts) relative to various aims, and (iii) implementing relevant changes accordingly: employing and focusing on the concepts deemed better relative to relevant aims.

This description and motivation put the focus on new concepts, serving as alternatives to the concepts we actually have and employ.

The project described has three separate parts. Part (i) is one of exploration. Part (ii) concerns evaluation. Part (iii) concerns revision. There is a question of how to use the label ‘conceptual engineering’ in relation to this. Merely engaging in the exploration project is hardly to engage in conceptual engineering, in any reasonable sense of the phrase. But I am thinking that so long as there is an element of (ii) or (iii), conceptual engineering is going on. Some may think that the label is best reserved for (iii), and that ‘conceptual ethics’ is a better label if only (ii) is involved. Fine by me in principle: labels are not that important. But as I will use ‘conceptual engineering’ it need not involve (iii).

Here’s a motivation for taking an interest in exploration. The concepts we actually have and employ are only some of all the possible concepts there are. There is both a theoretical issue regarding the nature of concepts that motivates inquiring about the limits of the space of concepts, and a practical issue – what conceptual tools are there to use, and what might they do for us? The second point brings us to the motivation for taking an interest in the project of evaluation. Given that there are other concepts for us to employ, it would be bizarre to assume that the concepts we find ourselves with are the best for all purposes. Maybe by some measure, the concepts we have must be somewhat good or we would have revised them, but that does not mean that there are not better concepts. And this brings us naturally to revision. Of course, if it is possible for us to start using different concepts, and those concepts are better than the ones we actually use for the purposes for which we employ them, then it would appear to be a good thing if such changes can be implemented.

The three projects are separable. Suppose that, for whatever reason, one is skeptical of the project of revision. For example, some theorists give voice to skepticism about the very possibility of implementation (see Cappelen 2018), thinking that it is not feasible to effect revisions in concepts in the way recommended. Even assuming that there are reasons for such skepticism one may be enthusiastic about exploration and about evaluation, thinking that questions about what concepts there are and how good they are remain significant. Knowing the limits of one’s tools is important even if obtaining better tools is not practically feasible. And even someone skeptical of evaluation can remain enthusiastic about exploration. For example, even someone skeptical of talk of how good a concept is may still take a keen interest in what concepts there are.

Insofar as the implementation problem seems serious, it has to do with problems in ensuring that old expressions come to have new meanings: how can we go about making it the case that a certain expression already in use comes to have a new meaning?¹ Conceptual engineering as motivated is only about what concepts to use, and not which expressions to use to express which concepts. It is consistent with this motivation to have a policy to always use new expressions for new concepts.

Another issue relevant to the implementation problem is the distinction between changing what concepts we employ and changing what concepts we focus on in theorizing. The implementation challenge, as it tends to be presented, concerns the former. The idea is that given plausible ideas about metasemantics, it is not in our control what concepts to employ. When the implementation challenge is discussed, the concern is less often on what concepts to focus on – what concepts to discuss and talk about. In principle, someone inclined to press the implementation challenge might insist that it is also hard or impossible to decide to bring it about that one talks about alternative concepts. But such a claim is more radical. Consider Sally Haslanger’s (e.g. 2000) well-known proposal that, for political purposes, the ordinary concept woman should be replaced by one such that it is necessarily the case that if someone falls under the new concept woman then this person is oppressed. Even someone who takes there to be significant problems regarding bringing it about that we generally employ Haslanger’s proposed replacement concept woman (which is such that it is a given that ‘women’, if there are any, are oppressed) instead of the current concept woman might find it obvious that we can intentionally bring it about that we talk about Haslanger’s replacement concept and discuss, for example, whether it would be better to employ that than to employ the current concept.

The distinction between employing and talking about a concept is relevant also to a very real but seldom highlighted distinction among theorists concerned with conceptual engineering. Many self-styled conceptual engineers – the activists, we might call them – see as an ultimate goal to change what concepts are employed, whether by philosophers or the public at large, and see other aspects of conceptual engineering, such as more theoretical discussions, as subservient to that general goal. Others – the theoreticians as we might call them – are primarily focused on philosophical theorizing, and seek to steer the focus of theorizing away from ordinary concepts and toward alternatives to these concepts. The theoreticians can be interested in concepts as objects of study as much as something to employ. Needless to say, there is a spectrum of positions. I am very much on the theoretician side, in terms of where my own philosophical interests lie. The implementation challenge may, because of what was pointed out in the preceding paragraph, be more of an issue for the activists.

While I will focus on how I prefer to see conceptual engineering, as opposed to how conceptual engineering is generally viewed, let me note that in their introduction to the collection Conceptual Engineering and Conceptual Ethics (2020), Herman Cappelen and David Plunkett quote Nietzsche’s The Will to Power:

Philosophers . . . have trusted in concepts as completely as they have mistrusted the senses: they have not stopped to consider that concepts and words are our inheritance from ages in which thinking was very modest and unclear … What dawns on philosophers last of all: they must no longer accept concepts as a gift, nor merely purify and polish them, but first make and create them, present them and make them convincing. Hitherto one has generally trusted one’s concepts as if they were a wonderful dowry from some sort of wonderland: but they are, after all, the inheritance from our most remote, most foolish as well as most intelligent ancestors … What is needed above all is an absolute skepticism toward all inherited concepts.²

And they comment as follows:

Nietzsche here articulates a radical skepticism about all inherited concepts. Philosophers should question whether the concepts we have are good enough and should engage in conceptual critique. What emerges, thinks Nietzsche, is the following: we should not just improve the concepts we’ve been given, reforming or “polishing” them in minor ways, but also create new ones— concepts not tainted by the “most foolish of our ancestors”.

Even if you think Nietzsche’s claim is more than a bit hyperbolic, you might think some more moderate version of his view is justified. For example: maybe some of the concepts we have inherited are defective, or at least not as good as they could be for our current purposes.³

I think Nietzsche’s advice (minus the hyperbole) and what Cappelen and Plunkett say when commenting on it is right on the mark, and is very much in line with how I motivated conceptual engineering above. The concepts we have are just some of all the concepts there are. It would be a miracle if the turned out to be the best ones for all relevant purposes. Hence there is reason to create new concepts. (I say ‘all relevant purposes’. Cappelen and Plunkett speak of our ‘current’ purposes. But I think that focus is not entirely ideal: there can be improvements also with respect to what purposes we have. We may come to realize that some of our current aims and purposes are abhorrent.)

In light of the Nietzsche passage, call what I have described the Nietzschean motivation for conceptual engineering. Much of my discussion will be about a particular issue that arises for conceptual engineering thus motivated. What is it to come up with new concepts, and to what extent do, and can, we actually do so?

The main points I will make are two. First, conceptual engineering as actually practiced has, perhaps surprisingly, had little or nothing to do with coming up with new concepts. Second, actually engaging in coming up with new concepts – engaging in conceptual innovation as opposed to, for example, just using a different label for a concept we already had, at least latently – is more tricky than it might seem.

Here is a roadmap. In Section 2, I will say a bit more about the use of the label ‘concept’ here. In Section 3, I will relate the Nietzschean motivation to the discussion of conceptual engineering in the literature, and specifically to Max Deutsch’s (2020) criticism of conceptual engineering. Section 4 will turn to David Chalmers’ discussion of what he calls de novo conceptual engineering. That might seem as if it is exactly a matter of coming up with new concepts. But as Sections 4 and 5 both discuss, there are good reasons to doubt that it is so. One main conclusion is drawn already after these sections: insofar as conceptual engineering should be seen as an enterprise of coming up with new concepts, it is arguably a grand failure so far. The discussion up to this point also suggests a broader problem of what it is to come up with new concepts to begin with. This is discussed in Sections 6 and 7. Section 8 turns to the issue of concepts belonging to what may be called alien logical categories.

2. Concepts

The Nietzschean motivation for conceptual engineering is independent of any finer details regarding how one speaks about concepts. But it is still useful to say something about how I use ‘concept’, if only to forestall confusions.

One problem regarding use of ‘concept’ is that it can be ambiguous as between vehicles and contents. Sometimes one may speak of concepts as if they were the thought counterpart of words, and words are representational vehicles with content. Words are not meanings but they have meaning. If concepts are mental words that would go for concepts too. Sometimes concepts are instead meanings or contents. Concepts are then not mental words but rather the sorts of things that are the meanings of words.

I will here use ‘concept’ the latter way. Instead of saying, as I said above, that the concepts we have and employ are only some of all the concepts there are, I could equally well say: the things that what we say and think mean are only some of the things that things said and thought could mean.

The choice regarding how to use the label ‘concept’ is just terminological and does not substantively affect anything I wish to say. I could use ‘concept’ the former way, and then the claim that the concepts we have and employ are only some of all the concepts there are amounts to the claim that we could there are mental words with other contents such that is in principle possible to use those mental words.

I have spoken and will speak as if there are these other concepts we could use. In this way of speaking, concepts exist and have their natures independently of whether they are employed. Given this way of speaking, we can never strictly invent new concepts; at most what we invent are ways of coming to have and employ concepts that existed all along. On another way of speaking about concepts, concepts only exist insofar as they are had and employed by thinkers. There are not these other concepts we could use, but there are facts about what concepts there could be. I will speak about concepts the former, platonistic way. Cappelen and Plunkett (as well as Nietzsche) speak of creating concepts and then appear to speak of concepts the latter way. If one thinks of concepts the former, platonistic way, then talk of creation must be taken with a grain of salt. Strictly, the concepts are not created. Instead, when one ‘creates’ a concept one brings it about that something which existed independently of us comes to be understood and employed by us.

Sometimes theorists speak of ‘conceptual engineering’ in such a way that by definition it does not involve new concepts but only revisions of concepts already in use. Conceptual engineering is about engineering concepts, and to engineer a concept is to modify it. It may seem that these theorists then by definition aren’t engaged in what is motivated by what I have dubbed the Nietzschean motivation for conceptual engineering. However, to some extent difference between the present focus and that of these theorists is just terminological. A theorist insisting that they are concerned only with revising existing concepts takes herself to modify a concept C with some given content in such a way that this very concept comes to have a different content. I can discuss that enterprise in my terms. I can say that the Nietzschean motivation speaks in favor of using concepts with contents different from the contents any of our concepts have hitherto had. Whether we call this creating new concepts or coming up with new content for our old concepts to have is of less importance.

One important distinction relevant to understanding the enterprise of conceptual engineering is that between concepts and conceptions. Conceptions are the theories that speakers associate with the things they talk about. My conception of dogs may be that they are scary, disgusting, etc.; yours may be that they are cute, playful, etc. But two speakers can have very different conceptions associated with dogs and still use the same concept dog. Concepts are, again, the building blocks of thought. Two thinkers can think the same thoughts involving the concept dog despite their having different conceptions associated with the concept. Our conceptions change all the time, and proposing that we should assess and revise our conceptions, and come up with new conceptions, would obviously be to propose that we should keep doing something we are already doing. Insofar as the enterprise of conceptual engineering is novel, it is because of its focus on concepts. On some prominent theories of concepts, concepts are associated with theories. Competence with a concept can be held to consist in dispositions to accept a certain set of claims in which the concept occurs, and this set of claims is then a theory associated with the concept. But so long as this is a circumscribed set of claims in which the concept occurs, this still need not be to identify the concept with the associated conception.

3. Deutsch’s dilemma

My focus in this text is on the Nietzschean motivation for conceptual engineering, and the attendant search for new concepts. I will soon turn to some problems for conceptual engineering motivated this way. But first, let me remark on some concerns about some supposed problems for conceptual engineering which are prominent in the existing literature. Given my stated motivation for conceptual engineering, these problems seem irrelevant to the enterprise. Consider for example what has come to be known as Strawson’s challenge. Here is how Cappelen states this challenge:

Change of extension and intension […] is a change of topic, so [projects involving conceptual revision] are bound to fail. Even if the revisions succeed, they do not provide us with a better way to talk about what we were talking about; they simply change the topic.⁴

Strawson’s challenge may be a serious challenge for anyone who wants to defend the idea that one addresses the same topic before and after conceptual revision. But that idea is no part of the above motivation for conceptual engineering. Indeed, it would be in the spirit of conceptual engineering to say that changing the topic is sometimes exactly the point. Why should we treat the topics we find ourselves with as some sort of wonderful dowry, to use Nietzsche’s phrase? Of course, there may sometimes be value in staying on topic, and in such cases, it may be important to keep track of what staying on topic entails. But this is no principled problem for conceptual engineering. Another central challenge is the above-mentioned implementation challenge: can the proposed changes be implemented? Again, even if the implementation challenge is serious, that only affects the project of revision; it leaves the projects of exploration and of evaluation unscathed.

So both these issues are largely non-issues for conceptual engineering as I discuss it here. However, in a much discussed recent article, Max Deutsch (2020) presents a dilemma for conceptual engineering that may be thought relevant to what I want to push, and discussing Deutsch’s dilemma promises to be instructive. Deutsch says,

I will argue that, given the sorts of things would-be conceptual engineers take to be examples of conceptual engineering, they face a dilemma: either we are ignorant of how conceptual engineering can be implemented, or it is straightforward to implement, but deeply uninteresting, involving no new technique, and ill-suited to solving, or even making genuine progress on, any philosophical problem.⁵

The first horn concerns changing the meanings of existing terms; the second concerns introducing new terms by stipulation. Deutsch’s argument regarding the first horn is a version of as the implementation problem, and I have already explained why I do not see that as highly relevant given the motivation stated. What he says regarding the second horn is at least on the face of it more relevant. So let me focus on that. Deutsch elaborates as follows:

Somewhat surprisingly, it seems that some of what its defenders describe as “conceptual engineering” amounts to no more than the stipulative introduction of new terminology. I say “surprisingly” because there is nothing especially remarkable about stipulative introduction. It is certainly not some new, particularly fertile method of philosophizing, one that might supplement or supplant philosophical conceptual analysis. It takes no special insight or skill to stipulatively introduce a term, for example, and it seems utterly incapable of solving any genuine philosophical problem. Stipulative introduction won’t reveal whether we have free will, all women are subordinated, or knowledge is justified true belief. In fact, it seems that most of its value derives from syntactic convenience: via stipulative introduction, we can replace longer descriptions (‘desk chair with five legs’) with a shorter, single term (‘brollop’).⁶

So on the second horn of the dilemma, the conceptual engineer is taken to just be concerned with ‘the stipulative introduction of new terminology’. And that, Deutsch notes, is something philosophers have been doing all along – and they didn’t need to be told to do it.

Deutsch takes me to be a friend of the strategy targeted by the second horn of the dilemma. He describes my stance as follows:

Eklund’s (2014) motivation for pursuing projects of conceptual engineering: like the theoretical physicist, relative to “folk” concepts of physics, philosophers, according to Eklund, need to think about whether our “actual” or “folk” philosophical concepts are the “best tools” for their “theoretical purposes”.⁷

and remarks,

… did philosophers really need to be told any of this? Is Eklund’s advice, as I am interpreting it, something that philosophers have not been heeding? Have defenders of conceptual engineering reminded philosophers of something they forgot, or perhaps never knew, namely that, sometimes, there is a need for technical terms in philosophy? I think it is obvious that the answer to these questions is “no”.⁸

Some things Deutsch says about the second horn of the dilemma are hard to disagree with. For example, most philosophers hardly need reminding that there is a need for technical terms in philosophy. But I would still want to defend conceptual engineering, or my conception of it, against what Deutsch says about this.

A first response to Deutsch is that while philosophers have indeed been introducing new terms by stipulation, that has tended to be a matter of introducing new terms in the course of investigating specific topics of traditional philosophical concern. So, for example, the new technical term ‘safety’ has been introduced in the course of the debate over knowledge, the notion of sentences being ‘grounded’ has been introduced in the course of discussion of what to say about truth in light of the liar paradox, etc.⁹ But what I would urge, when advocating for conceptual engineering as motivated, is rethinking the focus on the traditional questions. Perhaps the concepts in terms of which they are framed are not the best concepts to employ. Perhaps knowledge* is more important than knowledge, and a more worthy target of philosophical concern. There is a difference between introducing new terms by stipulation as part of an investigation of an already given topic, as with ‘safety’, and doing so as part of determining what is to be the main focus of investigation in the first place.

This relates to a second response I want to make. Deutsch says: ‘Stipulative introduction won’t reveal whether we have free will, all women are subordinated, or knowledge is justified true belief’. Of course he is right about what stipulative introduction of terms won’t reveal. But by stipulative introduction I can introduce an expression ‘free will*’ such that where philosophers have hitherto asked whether we have free will, the question of whether we have free will* is more significant. No answer is revealed by doing this, but the new question asked, do we have free will*?, may be a better question to ask than the old question do we have free will?. Whether we have free will* is more relevant to things we care about, or should care about, than whether we have free will. Deutsch criticizes the method he is discussing for not providing answers to the old question asked. But I do not see that it is, or needs to be, meant to. The focus of conceptual engineering is not on providing answers but on asking the right questions.

I believe that the above remarks suffice as a defense of Eklund (2014) against Deutsch. But there is also a third, deeper response to make. While the introduction of new expressions may often just be a matter of ‘syntactic convenience’, to use Deutsch’s phrase, it need not always be so. Sometimes it may be more a matter of conceptual innovation. Introducing a new expression is not the same thing as introducing a new concept. I can introduce a new expression for a concept I already possess. I do not thereby express anything I could not express before; I just do so using a different expression. Introducing a new concept is a matter of coming to be in a position to express something I earlier could not express. In principle, one can agree with Deutsch about cases where it is merely new expressions that are being introduced, but say that in cases where new concepts are introduced, things stand differently. And given the motivation presented in Section 1, the focus is on new concepts.

4. Chalmers on de novo conceptual engineering

I have highlighted the issue of searching for new concepts as opposed to merely new expressions. To what extent have theorists writing about conceptual engineering actually focused on the introduction of new concepts? Not to any extent at all, I will now argue.

A first point is that prominent instances of conceptual engineering that are discussed in the literature do not actually seem to involve the introduction of new concepts. Consider again Haslanger’s proposal that the old concept woman should be replaced. The replacement she proposes is the following concept, also dubbed woman:

S is a woman iff S is systematically subordinated along some dimension (economic, political, legal, social, etc.), and S is “marked” as a target for this treatment by observed or imagined bodily features presumed to be evidence of a female’s biological role in reproduction.¹⁰

But did we not, even before Haslanger entered the scene, have the complex concept of being systematically subordinated economically/politically/legally/socially, and being ‘marked’ as a target for this treatment by observed or imagined bodily features presumed to be evidence of a female’s biological role in reproduction? After all, we had the constituent concepts and grasped their mode of combination. And, one may wonder, did we not then have Haslanger’s proposed concept woman, for isn’t this just that concept? Similar remarks apply to other conceptual engineering proposals.¹¹

In his (forthcoming), David Chalmers distinguishes between what he calls de novo engineering and re-engineering:

De novo engineering is building a new bridge, program, concept, or whatever. Re-engineering is fixing or replacing an old bridge, program, concept, or whatever.¹²

He applies this to conceptual engineering. De novo conceptual engineering is then building a new concept. This may sound like it is conceptual innovation of the exact kind I have advertised interest in. Chalmers reflects:

Many or most of the standard examples in the recent conceptual engineering literature are cases of conceptual re-engineering. Certainly the Carnapian explication literature is very much a literature on re-engineering … The conceptual engineering of belief has largely been re-engineering. There’s also been re-engineering with the concept of truth. More recently there has been a lot of interest in re-engineering social concepts such as the concept of woman and the concept of race.

Many of the examples I gave [of conceptual engineering] … look more like de novo conceptual engineering. Take the cases of epistemic injustice, supervenience, rigid designation, and, indeed, conceptual engineering. These weren’t particularly trying to fix or replace other concepts.¹³

It may sound like the cases of ‘epistemic injustice, supervenience, rigid designation, and, indeed, conceptual engineering’ as presented by Chalmers are instances of conceptual innovation of the kind I am after. But while it is true that those who introduced talk of epistemic injustice, supervenience, etc. did not try to replace anything, the concepts at issue still do not appear to be new concepts. Each can be exhaustively characterized in antecedently available terms. Are they then really results of de novo conceptual engineering?¹⁴

5. Having a concept

My concerns about whether Chalmers’ purported examples of de novo conceptual engineering, and the other examples brought up, strictly speaking involve new concepts naturally invite the question: when, exactly, are we dealing with a new concept?

There are some different distinctions to draw here. I will present them using the helpful taxonomy from Jacob Beck (2017).

First, if I have the concept female fox, do I already have the concept vixen or not? Beck says not but says that even so, the overall expressive power of my conceptual system is not expanded when I acquire vixen, since female fox and vixen have the same ‘contents’. Above I proposed to identify concepts with contents so for me there is no distinction like that. But if one individuates concepts by something other than their contents, Beck’s distinction is crucial. A related issue, not raised in Beck’s discussion, concerns whether it is possible for concepts to be necessarily coextensive but yet have different content. How fine-grained is content? Without taking a stand on this substantive issue, I will say that acquiring a new concept expands expressive power only if the concept is not necessarily equivalent to some concept the thinker earlier had. For me, using this label is just to adopt a terminological convention. Second, Beck distinguishes between a concept being available for employment, and a concept being latent. A concept is available to a thinker, Beck says, ‘just in case she could deploy it in reasoning, thinking, categorizing, remembering, and other cognitive processes without much effort, simply by endogenously shifting her attention’.¹⁵ Latent concepts are concepts that are ‘unavailable yet expressible in terms of concepts that are stored in the thinker’s mind’.¹⁶ Beck distinguishes between different ways in which concepts may be latent. Here is the first way:

One way for a concept to be latent is for it to be unobviously composed from one’s available concepts. Consider the concept burse, which is satisfied by an object just in case it is either green and circular, or blue and enclosed by a prime number of sides, or red and preceded in presentation by a yellow triangle. We can suppose that each of the constituent concepts in this definition—green, circular, or, and, prime, . . . —are available to you. But of course you do not normally go around categorizing things in this peculiar way, and would find it rather difficult to do so.¹⁷

The second way involves having a concept ‘fully formed and stored in the mind’, but ‘isolated from general cognitive processes such as reasoning and categorizing’ It can, for example, be ‘stored inside an innate module in your head’. Using terminology from Rey, Beck says that such concepts are lying in wait.¹⁸

It is the first way for a concept to be latent that is the most relevant for present purposes. I will set aside the issue of concepts lying in wait. Even if there are such concepts, it is unlikely that all potentially useful concepts that are not either available or latent in the first way will be among concepts lying in wait.

Return now to the issue of what it is for a concept to be a new concept for a thinker. One way for a thinker to, in some sense, acquire a new concept is for a concept that previously was latent for her to become available. And this can obviously be extremely useful for the thinker. But in another sense, the thinker does not thereby acquire a new concept. The concept was latent all along.

I am interested in cases of genuine concept acquisition, as opposed to cases where, at most, a latent concept is made available. But I would not want to deny that already conceptual engineering that only has an effect on which concepts we use for what purposes can be important for many purposes, even if it does not involve anything more than making latent concepts available and in that way changes what concepts get employed. Let me mention wo examples, to illustrate. First, regarding making latent concepts available. It is often said that the concept sexual harassment was something we, including the victims of sexual harassment, did not use to have. But arguably it was at least latent (or, if content is fine-grained: acquiring it did not add to expressive power, in the stipulated sense). But already making it readily available for employment can obviously be important. For example, it can help victims of sexual harassment process what they have been subjected to. Second, even an already readily available concept may be used for a new purpose – and that may be innovation in one sense. Here is a possible illustration. Since Gettier (1963) we know that knowledge is not the same as justified true belief. All the same we, at least we philosophers familiar with the relevant epistemological debates, do have the concept justified true belief. And one kind of ‘innovation’ is to decide to use that for the purposes for which we ordinarily use knowledge, however actually those purposes are characterized. Maybe they involve being what we aim for when believing and asserting things, and generally being what we aim for in inquiry. Even if justified true belief was readily available for employment all along, employing it in a different role is a kind of innovation.

But it remains that if all that conceptual innovation ever amounts to is making latent concepts available or affecting which already available concepts we use for what purposes, then conceptual innovation is pretty modest. Anyone inspired by the Nietzschean thought might, and arguably should, feel that it is a lot more modest than what they were looking for. Surely there must be concepts that are not even latent for us, and which are such that acquiring them involves a genuine increase in expressive power.

6. Coming to have new concepts

What I take myself to have established so far is this. One important motivation for conceptual engineering motivates search for new concepts, but actual examples of conceptual engineering at most merely involve making latent concepts available. Already this is an important conclusion. But the conclusion drawn also invites a question: could there be conceptual engineering that involves a search for new concepts? My remarks on this will be tentative. The question is significant, whatever the answer. If there can be conceptual engineering like that, this is new territory to be explored. If there cannot be conceptual engineering like that, this is an important and so far unobserved limitation to the practice of conceptual engineering.

Steffen Koch and Jakob Ohlhorst (manuscript) draw a distinction between what they call moderate and heavy duty conceptual engineering:

Moderate conceptual engineering: A type of conceptual engineering that alters a community’s psychological expressive power by constructing unobviously composed complex concepts out of already available concepts.

Heavy duty conceptual engineering: A type of conceptual engineering that alters a community’s semantic expressive power by introducing collectively foreign concepts.¹⁹

This is obviously closely related to what I am discussing here. Moderate conceptual engineering aims at making merely latent concepts available. Heavy duty conceptual engineering aims at increasing expressive power.

In their discussion of heavy-duty conceptual engineering, Koch and Ohlhorst mention a number of purported examples of genuine conceptual innovation:

Kepler’s vis motrix concept

Mendel’s concept of gene

Maxwell’s concept of electromagnetic field

Einstein’s concept of space–time²⁰

However, they do not elaborate on how the examples are supposed to work. And strikingly, given that they themselves emphasize the distinction between moderate and heavy-duty conceptual engineering, they do not discuss seriously the possibility that in these cases we are only dealing with moderate conceptual engineering: the ‘new’ concepts may be exhaustively characterizable in terms of old ones and thus be merely latent. I am not claiming that Koch and Ohlhorst definitely fail to provide examples of heavy-duty conceptual engineering. Establishing that would require very detailed discussion. The point is just that if someone says that something is an instance of heavy-duty conceptual engineering, the burden is on them to provide a compelling case that it in fact is so. Even if a purported new concept is not explicitly introduced in terms of old ones, it can be necessarily equivalent to a complex concept built up of old ones in such a way that it does not actually increase expressive power.

Koch and Ohlhorst do not just bring up specific examples but also discuss in general terms how one comes to acquire new concepts. They centrally bring up Quinean bootstrapping. The idea comes from Susan Carey.²¹ Carey’s topic is how children acquire new concepts, and it is certainly a reasonable strategy, when considering how we come to have new concepts, to consider the best research on how children manage to do it. There is discussion in the literature of how exactly Quinean bootstrapping is to be understood.²² But here is how Carey herself summarizes it:

(1) relations among symbols are learned directly, in terms of each other; (2) symbols are initially at most only partly interpreted in terms of antecedently available concepts; (3) symbols serve as placeholders; (4) modeling processes – analogy, inductive inferences, thought experiment, limiting case analyses, abduction – are used to provide conceptual underpinnings for the placeholders; (5) these modeling processes combine and integrate separate representations from distinct domain-specific conceptual systems; and (6) these processes create explicit representations of knowledge previously embodied in constraints on the computations defined over symbols in one or more of the systems being integrated.²³

In brief, the idea is that a thinker first has one system of concepts (see 2), and then new, analogous ones are acquired on the basis of these antecedently available ones (see 3, 4 and 5).

Koch and Ohlhorst distinguish between two kinds of cases of concept acquisition: cases where children acquire concepts already known in their community, and cases where entirely new concepts come to be grasped by some thinker. Cases of the latter kind are more significant for present purposes. One example of that kind that Koch and Ohlhorst discuss is that of Kepler’s concept vis motrix, and Kepler is supposed to have acquired this by Quinean bootstrapping. They do not actually pause to argue that this is a case of heavy duty conceptual engineering. And it is not obvious why it would be. Kepler’s supposed conceptual innovation, as described by Koch and Ohlhorst, is this:

Kepler created the foreign concept of a purely physical vis motrix – a predecessor of Newtonian gravitational force – that emanates from the sun and propels the planets. This replaced the placeholder structure of an intelligent anima motrix. He achieved this bootstrap by drawing on the observation that the physical effects of light occur at a distance without any intelligence being required for their explanation while structurally analogous effects occur with the motion of planets.²⁴

This may be a revolutionary new theory, importantly different from the theories of how things work that preceded it. But in what sense is this a new concept? How is this heavy duty conceptual engineering as opposed to merely moderate conceptual engineering? Surely, one may object, for all that is said Kepler’s ‘revolutionary foreign concept’ could simply amount to like anima motrix but purely physical. This may be revolutionary along one significant dimension – maybe earlier thinkers could not conceive of how anything could be like that – but for all that, this is not a new concept. It is built up from elements already available. The point holds even assuming that thinkers preceding Kepler would have thought that for principled reasons, such a concept would be necessarily empty. And the point stands even if for those thinkers, Kepler’s idea involves something like a conceptual error in some sense that somehow goes beyond this. Compare: even if, say, the concept colored idea is incoherent, I can, simply combining concepts I have, form the concept like an idea except that it can be colored (or a property like that of being colored except an idea can have it). One may of course think that a concept like this is necessarily empty. But due to the flexibility of ‘like’ it is hard to see how it can be conceptually ruled out that this concept is empty.

Similarly, allowing for the possibility that Kepler’s concept vis motrix is non-empty maybe required a radical shift in thinking, and so a reconceptualization in one sense, and perhaps a quite significant sense, but for all that the new concept is a construct out of old ones.

There is a nearby general point to make about Carey’s appeal to Quinean bootstrapping. Some critics, such as Georges Rey (2014), have held that a thinker cannot expand the expressive power of her conceptual system by such bootstrapping. One theme is that whatever concepts a thinker can, in some sense, acquire using such bootstrapping are already expressible through the use of a Ramsey sentence. Briefly, in the words of Beck (2017), a Ramsey sentence ‘takes advantage of techniques in quantificational logic to enable multiple new terms to be simultaneously inter-defined with one another and the language’s old terms’.²⁵

Carey herself quotes the following passage from Ned Block (1986) when explaining bootstrapping:

When I took my first physics course, I was confronted with quite a bit of new terminology all at once: ‘energy, momentum, acceleration, mass’ and the like … I never learned any definitions of these new terms in terms I already knew. Rather, what I learned was how to use the new terminology—I learned certain relations among the new terms themselves …, some relations between the new and the old terms, and, most importantly, how to generate the right numbers in answers to questions posed in the new terminology.²⁶

Commenting on this, Rey says that what Block here presents ‘is a description of how one might learn a slew of terms at once by way of, precisely, a ‘Ramsey sentence’’.²⁷ And if this is right, Rey reasons, Quinean bootstrapping only serves to make previously latent concepts available. The idea is that formulating Ramsey sentences requires only resources already possessed by the thinker.

Beck (2017) criticizes Rey on this score. Beck’s main point, which seems to me correct, is that it is only for thinkers possessing the logical concepts needed for formulating Ramsey sentences that definitions by means of such sentences only serve to make latent concepts available, and it is at best far from clear that thinkers generally have the requisite concepts. For example, young children arguably do not. However, Beck’s point is also compatible with, and positively encourages, the view that for thinkers, like sufficiently sophisticated adult thinkers, with the requisite logical concepts, definition by Ramsey sentences do only serve to make latent concepts available. But then for all Beck says, Quinean bootstrapping is not a means for sufficiently sophisticated thinkers to expand the expressive power of their conceptual systems.

7. Conceptual innovation

I do not mean any of my remarks to suggest that conceptual innovation in the non-modest sense discussed is impossible. Children come to acquire what for them are genuinely new concepts. So it undoubtedly happens.

But even so, this may not be thought not sufficient for the aims of conceptual engineering when conceived of as here, as concerned with genuinely new concepts. Conceptual development is not always conceptual engineering. Development can be something that just happens; ‘engineering’ implies a level of control. The mere fact that conceptual development happens does not mean that it is in our control. Of course, ‘conceptual engineering’ is just a label, but there is a substantive point. One might in principle worry: how can there be a reasonable research program in philosophy centered on developing new concepts if such development is not something over which we have control?

How could a philosopher go about ensuring that she comes to have new concepts? One can come to have new concepts completely unintentionally, as when use of a concept undergoes changes over time, not because of conscious decisions but because of changes in how it is in fact used, in such a way that at the end of the process the thinker has a new concept (or the same concept with a different content, if one prefers to speak that way). But that is not an intentional process, so it is not clearly relevant. One can also come to have a new concept through worldly discovery: one might encounter a new phenomenon in the world or come to have a new kind of experience and come to have a concept for that. And one can make sure to be in a position to make such discoveries, by exploring unexplored part of the world, or using new more powerful microscopes, or taking hallucinogenic drugs. But the range of new philosophical concepts one comes to have that way is arguably rather limited.

How much of a problem is this for conceptual engineering, motivated in the Nietzschean way? Not very much, someone might reply. Instead of seeing this as a cause for great concern one might stress that conceptual innovation as a creative process much as other creative processes are. And the mere fact that, for example, coming up with a new theory is a creative process is for good reason not seen as a problem for all those branches of inquiry where theory construction is central. That said, there is a difference between the case of creating concepts and the case of creating theories. There is a way that creating a new theory is easy: you just combine some concepts you have in some new and unfamiliar way. Having the concepts there are, elephants, on and Neptune, I can put them together to come up with the new theory that there are elephants on Neptune. It may not be a good theory but it is a new theory. There is no analogous procedure for cheaply coming up with, in the relevant sense, new concepts, whether ‘good’ ones or not. Putting the concepts elephant, on and Neptune together I can form the concept elephant on Neptune, and that may be new for me, but it is at most a matter of making a latent concept available. One can come up with theories by putting old concepts together in a novel way. Similarly, there is a kind of conceptual invention one can engage in by putting concepts together in a novel way. But this is only moderate conceptual engineering. Conceptual novelty of the kind that I have been discussing requires something more.

As already stressed, Koch and Ohlhorst (manuscript) appeal centrally to Susan Carey’s work on conceptual development and especially her notion of Quinean bootstrapping. And as already discussed, one issue with this concerns the extent to which a thinker actually acquires new concepts using this method: perhaps it is just a matter of latent concepts being made available. Even setting that aside, there is a question of the extent to which we are in control over this bootstrapping process. Carey insists that when this happens in science, the scientists are in control. She says:

As the historical examples discussed in [Carey (2009)] make clear, bootstrapping episodes are often under metaconceptual control; the scientist is consciously engaged in exploring mappings between mathematical structures and physical/ biological/psychological phenomena. But […] metaconceptually explicit hypothesis testing and modelling procedures are not necessary.²⁸

Carey does not elaborate, and it is then left open in what way this is held to be under the scientists’ control. There is a dilemma nearby, related to the issue of Ramsey sentences. If Quinean bootstrapping just amounts to the employment of Ramsey sentences, or in some other way just amounts to making latent concepts available, then it may be under our control – or, more cautiously, it is as much under our control as coming up with new theories is. If, on the other hand, Quinean bootstrapping involves coming up with new concepts as opposed to just making what is latent available, then it is correspondingly more unclear how this process can be supposed to be under our control.

8. Alien logical categories

I have argued that conceptual engineering as actually practiced does not actually involve coming up with new concepts, and I have raised some general concerns about actually being able to do so. Before closing, let me make some constructive remarks about conceptual innovation.

In other work, Eklund (2024), I am looking into the possibility of alien languages, in the sense of languages that are semantically different from familiar languages not merely in that they have words for things we do not have words for, but in that, they have expressions that belong to alien logical categories – not just familiar categories like singular term, predicate, quantifier, etc. Investigating the possibility of such languages and the advantages they might have along various dimensions is an instance of conceptual engineering as I have characterized it here. The explicit focus is on language, but insofar as it is assumed, as is standard, that corresponding to singular terms are concepts with the same logical role (one of referring), corresponding to predicates are concepts with the same logical role (one of predicating), etc., the search for alien languages corresponds to search for systems of alien concepts.²⁹

What might alien languages, and systems of alien concepts, be? This is a difficult question. And while I believe that there are such alien representations, some theorists may reasonably be skeptical. Let me illustrate using an example how some of the discussion might go.

In his (2020), Peter Sullivan says:

[Ramsey] observed that nothing rules out propositions consisting entirely of several expressions of the same type […] He was not suggesting that we could make sense of non-sentences like ‘Socrates Plato’ or ‘mortality senility wisdom’. Any type or category that did self-combine as those familiar ones fail to would be very different from those we employ. It would be employed in thought of a very different logical shape, and altogether alien to us.³⁰

Call a language of the kind Sullivan envisages a flat language. The idea is that in a flat language, all subsentential expressions are of the same type, and sentences are built up from groups of such expressions. These expressions are then neither singular terms nor predicates. For neither singular terms nor predicates work that way.

There are immediate objections to the example and how it is described. In fact, we can easily envisage languages with sentences like ‘Socrates Plato’ and ‘mortal senile wise’. The former can express that Socrates stands in a certain relation to Plato. The second can express that something is mortal, senile and wise. This is of course not how Sullivan intends his example: he thinks of the sentences of what I am calling a flat language as not involving unarticulated elements. But the objection still pounts to a deeper issue. While is easy to concoct languages that on the face of it seem alien, the question is whether what is described is really alien.³¹

I will not further address the objection, or, more generally, the possibility of alien languages, here. All I want to note, in connection with the present theme of conceptual innovation, in this. Suppose that alien languages are possible. Suppose for example that flat languages as envisaged by Sullivan are possible. In describing such a language we are then describing alien concepts: concepts logically different in kind from the ones we actually have. This is going some way toward satisfying the Nietzschean motivation behind conceptual engineering. New concepts are described. If we were to describe the new concepts by providing synonyms using conceptual resources we already have, we would be running the risk of being engaged in merely modest conceptual engineering. So these logically alien concepts can at most be described in general terms, in terms of their general logical or combinatory features. Given only such general descriptions, evaluation of the concepts is hard: how are we to decide whether these concepts are better, in some relevant way, than familiar ones? But while hard, it may, depending on the purpose, not be impossible. One might for example envisage there being possible arguments in metaphysics to the effect that the structure of reality is better represented by a flat language than by a familiar kind of language.

9. Concluding remarks

My main theme here has been this. A central motivation behind conceptual engineering seems to call for a search for new concepts. But when we consider what it is for a concept to be new for us it is unclear to what extent any examples discussed in the relevant literature, or any examples from the philosophical discussion, fit the bill. I say ‘unclear’ and I mean it. As stressed, one might speak in different ways about concepts, and having concepts. But at least one reasonable way of employing concept talk, the examples discussed in the literature do not fit the bill. And more generally, there are principled problems regarding how to go about introducing the new concepts. However, I ended on a constructive note: if the search for alien languages is successful, it provides some examples of new concepts, even if not the kinds of examples that provide exactly what we might have hoped for.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 See Cappelen (2018) and Deutsch 2020). For responses see, e.g. Pinder (2021) and Jorem (2021).

2 Cappelen and Plunkett (2020, 1). I quote the same passage in my (2020). Of course questions may be raised about how exactly Nietzsche himself understood ‘concept’.

3 Cappelen and Plunkett (2020, 1).

4 Cappelen (2018, 62).

5 Deutsch (2020, 3936).

6 Deutsch (2020, 3945).

7 Deutsch (2020, 3946).

8 Deutsch (2020, 3946).

9 See Ichikawa and Steup (2018), section 5.2, and Herzberger (1970), respectively.

10 Haslanger (2000, 39). She goes on to present a more elaborate suggestion but the same remarks apply to that.

11 Kevin Scharp’s (e.g. 2013) prominent proposed replacement of the concept truth is a more complicated case. Scharp introduces his pair of replacement concepts for truth, ascending truth and descending truth, by saying what rules of inference are supposed to govern them. He does then not introduce them by providing synonyms, and there are not necessarily synonyms in the language pre-revision. But Scharp still does give a full characterization using antecedently understood language, and already someone possessing the concepts expressed by expressions of the old language, and who has a general capacity to draw inferences, has the means to perform the inferences constitutive of the new concepts, so there is a clear sense in which Scharp does not introduce genuinely new concepts either.

12 Chalmers (2020).

13 Chalmers (2020).

14 Moreover, even if this is less important for my purposes: Even a self-styled ‘re-engineer’ could in principle seek to re-engineer an already used concept in such a way that it comes to have a content such that no concept we have hitherto used has that content.

15 Beck (2017, 113).

16 Beck (2017, 113).

17 Beck (2017, 113).

18 Beck (2017, 113).

19 Koch and Ohlhorst (manuscript), 6.

20 Koch and Ohlhorst (manuscript, 12).

21 See especially Carey (2009).

22 See the remarks in Beck (2017, 110).

23 Carey (2009, 418).

24 Koch and Ohlhorst (manuscript, 12).

25 Beck (2017, 117).

26 Block (1986, 648). Quoted by Carey (2009, 419), and by Rey (2014, 114).

27 Rey (2014, 114).

28 Carey (2014, 160).

29 For further discussion, see Eklund (2024).

30 Sullivan (2020, 195f).

31 See Eklund (2024), ch. 4, for further discussion of flat languages and the theoretical issues raised.