The phenomenon objection to conceptual engineering

Abstract

Conceptual engineering is sometimes presented as an alternative to conceptual analysis. But one important objection to conceptual analysis threatens to carry across: that philosophy investigates phenomena—knowledge, truth, freedom, etc.—rather than concepts of those phenomena. This poses a prima facie problem insofar as conceptual engineering targets concepts or terms rather than phenomena. Call it the ‘phenomenon objection’. I begin by examining recent discussions of the phenomenon objection by Cappelen and Scharp, rejecting their responses. I then clarify and strengthen the objection, discussing the challenge that the ‘strong phenomenon objection’ poses to conceptual engineers. Finally, I develop a new response to the strong phenomenon objection. In doing so, I motivate, articulate and defend an account of philosophy and its problems on which conceptual engineering is the appropriate method for tackling many philosophical problems—as well as, perhaps surprisingly, some problems in the sciences.

1 Introduction

People sometimes seek to change and improve their linguistic or conceptual repertoires for some given purpose. One may seek to revise, replace or eliminate a term or concept, or else to introduce a brand new one. This is called conceptual engineering.

Conceptual engineering is sometimes presented as an alternative philosophical method to conceptual analysis:

[…] while philosophers often have been concerned with our actual concepts or the properties or relations they stand for, philosophers should also be asking themselves whether these really are the best tools for understanding the relevant aspects of reality, and in many cases consider what preferable replacements might be. Philosophers should be engaged in conceptual engineering. (Eklund, 2014: 293)

A growing number of philosophers have begun to suggest that we depart from the Standard Model and its dogged pursuit of conceptual analysis, towards a more revisionary venture—conceptual engineering. Conceptual engineers aim to improve or to replace rather than to analyse; to create rather than to discover. (Nado 2021b: S1509)

However, at least one objection to conceptual analysis threatens to apply analogously to conceptual engineering. The objection is that conceptual analysis misconstrues the nature of philosophy: philosophy is the study of philosophical phenomena such as knowledge, freedom, the right and the good, etc., but conceptual analysis instead focuses on the concepts of those phenomena.¹ In recent work, Cappelen and Scharp raise this objection in connection with conceptual engineering.² Just like conceptual analysis, conceptual engineering is seemingly focused on concepts (or terms) rather than the philosophical phenomena they represent. Call that the phenomenon objection.

Examining the phenomenon objection is important for at least two reasons. Firstly, the phenomenon objection is more serious than has been acknowledged. Neither Cappelen nor Scharp discusses the objection in depth and, as I argue below, neither provides a compelling response. Moreover, with the objection suitably strengthened, there is no easy line of response. If I am right, the phenomenon objection is an open problem for conceptual engineers, deserving further attention.

Secondly, the phenomenon objection draws attention to important metaphilosophical questions. The objection purports that, given the alleged nature of philosophy as the study of philosophical phenomena, conceptual engineering is ill-suited to tackling philosophical problems. A key thought underlying the objection is this: that how we understand the nature of philosophy directly affects the viability of understanding conceptual engineering as a philosophical method. This raises important metaphilosophical questions: Is there a way of understanding the nature of philosophy such that conceptual engineering is a coherent method for tackling its problems? If so, what is it?. Engaging with the phenomenon objection forces the conceptual engineer to tackle such questions head-on. And this is important: compelling, positive answers to such questions are essential to any plausible view of conceptual engineering as a philosophical method.

Accordingly, herein, I use the discussion of the phenomenon objection as a springboard to develop an improved understanding of how and why conceptual engineering can be used to solve philosophical problems. In responding to the phenomenon objection, I motivate, articulate and defend an account of philosophy and its problems on which conceptual engineering should have the decisive role in solving many traditional problems in philosophy—as well as, perhaps surprisingly, some problems in the sciences.

I begin, in § 2, with a brief introduction to conceptual engineering in philosophy. In §§ 3–4, I critically examine Cappelen’s and Scharp’s discussions of the phenomenon objection. In § 5, I develop what I call the strong phenomenon objection, drawing out some of the challenges it poses to conceptual engineers. In §§ 6–9, I develop a response to the strong phenomenon objection, defending the view that conceptual engineering can play the decisive role in tackling many traditional philosophical problems (and some scientific ones). I close in § 10.

2 Conceptual engineering in philosophy

Conceptual engineers seek to improve linguistic and/or conceptual repertoires. For the most part they seek to improve upon existing concepts or terms, and they do this by improving upon contents.^{3, 4} In the case of concepts, this might be a representational, expressive and/or cognitive content; in the case of terms, this might be a semantic meaning or speaker meaning. So construed, there are various ways to engage in conceptual engineering: introducing a new vehicle with a new content, allowing us to think or say new things; eliminating a vehicle whose content was in some way problematic; or retaining a vehicle whilst revising and improving its content in some way. There are no specific constraints on the dimensions of content evaluation. We might seek semantic improvements (e.g. contents that are more precise), political improvements (e.g. contents that facilitate enhanced social justice), theoretical improvements (e.g. contents that carve nature at its joints), and/or something else.

Several philosophers have urged a central methodological role for conceptual engineering in philosophy. Let me give three examples.

Olsson (2015) develops a methodological approach to epistemology he calls explicative epistemology. The approach is to define key philosophical terms using Carnap’s (1950) method of explication. Simplifying, the method involves assessing proposed definitions against four criteria:⁵

• Similarity: the definition should apply in most cases in which the term is ordinarily applied.

• Exactness: the definition should be given in terms of a system of scientific concepts.

• Fruitfulness: the definition should be useful for the formulation of universal statements.

• Simplicity: the definition should be as simple as the above criteria permit.

To illustrate, Olsson investigates whether one can have knowledge in a Gettier scenario⁶—such as, for example, a commuter who forms the justified, true belief that it is 8:47am by accurately reading the station clock, which by chance happened to stop exactly twelve hours earlier. Adopting the explicative epistemological methodology, Olsson sketches an argument that, counterintuitively, one can have knowledge in such a scenario. In particular, he argues that a reliabilist definition of ‘knowledge’ does well against the four criteria overall despite having counterintuitive implications in some Gettier scenarios.⁷ According to Olsson, this illustrates how conceptual engineering can be used to tackle epistemological problems.

Sally Haslanger’s ameliorative project involves using conceptual engineering to tackle certain social philosophical questions such as ‘What is gender?’, ‘What is race?’ and ‘What is marriage?’. According to Haslanger 2000/2012, the ameliorator starts ‘by considering what we want the concept[s] in question for’ (p. 226), and develops revisionary definitions accordingly.⁸ Seeking ‘effective tools in the fight against injustice’ (ibid.), Haslanger develops what she takes to be revisionary definitions for concepts such as man, woman, and racialized accordingly. Later, in Haslanger 2006/2012, she offers a different gloss on the project, appealing to a form of metasemantic externalism on which the meanings of social kind terms are fixed by the nature of relevant paradigms and arguing that an ameliorative analysis ‘better reveals what we mean’ by the term in question (2006/2012: 398). But, more recently, Haslanger has reemphasised the role that conceptual engineering can play in ameliorative projects:

to ask, what is marriage, really? is to ask what forms of domestic partnerships (if any) promote a well-functioning and just society. When activists have claimed that same sex couples can be married […], it wasn’t based on what we have meant all along, but on what we should have meant. (2020: 257)

So, at various times, Haslanger has planned to use conceptual engineering to tackle activism-directed social philosophical questions of the form ‘What is X?’.

Finally, on Kevin Scharp’s (2020) radical therapeutic program, conceptual engineering has a central role to play, alongside conceptual analysis, when tackling nearly all philosophical problems. He claims, first, that ‘philosophy is for the most part the study of what have turned out to be inconsistent concepts’ (p. 397). On his view, concepts are constituted by principles that guide our deployment of those concepts, and a concept is inconsistent if and only if its constitutive principles jointly imply a falsehood. Given this, inconsistent concepts are problematic as, in some situations, they guide us towards accepting falsehoods. Scharp recommends that, in such cases, philosophers seek to replace the inconsistent concepts with consistent counterparts. Thus, according to Scharp, there are two aspects to tackling philosophical problems: the philosopher should analyse inconsistent concepts, drawing out and evaluating their constitutive principles; and, building on these analyses, the philosopher should engineer consistent replacements.

3 Cappelen on the phenomenon objection

Cappelen (2018) formulates the phenomenon objection thus:

Philosophy, on [Williamson’s (2007)] view, isn’t about analysing concepts […]. It is, instead, about various features of the world: knowledge, freedom, morality, perception, language, etc. […] Someone attracted to this picture of philosophy would and should be prejudiced against the very idea of conceptual engineering. On the face of it—in the very name!—it seems to be advocating concept-fiddling, which is just what should be avoided. (Cappelen, 2018: 146)

His response draws on three features of his Austerity Framework.

The first feature: conceptual engineers aim to change the semantic meaning of terms. Thus, for example, we might imagine Olsson’s explicative epistemologist seeking to change the intension of ‘knowledge’ from justified true belief to reliably-formed true belief.⁹ For concreteness, let us suppose that the explicative epistemologist succeeds, with the change of intension taking place at t ₁.

The second: terms are associated with topics such that, when a term’s semantic meaning changes within the limits of revision, the associated topic survives but undergoes a qualitative change.¹⁰ So, assuming for the sake of argument that the semantic change to ‘knowledge’ at t ₁ is within the limits of revision, utterances of (1) and (2) at t ₂ > t ₁ are true.

• (1) The topic associated with ‘knowledge’ has not numerically changed.

• (2) The topic associated with ‘knowledge’ has qualitatively changed.

As I understand Cappelen, he intends topics to be amalgams of their associated term’s intensions across time, so that the topic associated with ‘knowledge’ consists of justified true belief before t ₁, and reliably-formed true belief after t ₁.¹¹

The third: one can use a term to say something about its associated topic. So, one can use

• (3) Knowledge has qualitatively changed.

to express that the topic associated with ‘knowledge’ has qualitatively changed. If one does this at t ₂, one’s utterance of (3) is true—even though, at t ₂, the proposition semantically expressed by (3), that reliably-formed true belief has qualitatively changed, may be false. According to Cappelen, as sentences such as (3) can be truly uttered post-revision, conceptual engineering is ‘about the world’ (2018: 137).

Cappelen’s response to the phenomenon objection is this:

What I have described is a process that operates directly on extensions and intensions (i.e. things in the world). In that sense, the process I describe as conceptual engineering is about knowledge, freedom, what is right, women, marriage, and salad. (pp. 146–147)

Here, the extension of ‘knowledge’ is knowledge, the extension of ‘freedom’ is freedom, and so on. So Cappelen’s response to the phenomenon objection is that conceptual engineering is ‘about’ knowledge, freedom, etc., in the sense that it operates directly on those phenomena.

However, the response is not compelling. The problem is that, given Cappelen’s response, conceptual engineering operates on phenomena that are uninteresting from the relevant theoretical perspective. In the above example, the explicative engineer supposedly operates on knowledge at t ₁. So let us suppose that, at some unknown time, the explicative engineer (tenselessly) utters:

• (4) Knowledge qualitatively changes at t ₁.

For Cappelen, there are three interpretations of ‘knowledge’ available: its semantic meaning before t ₁; its semantic meaning after t ₁; and its associated topic. For Cappelen’s response to be compelling, one of these needs to be such that, with (4) so interpreted, there is an epistemologically interesting sense in which (4) is true.

The first two interpretations of ‘knowledge’ are its semantic meaning before t ₁, justified true belief, and after t ₁, reliably-formed true belief. However, with ‘knowledge’ interpreted either way, the utterance of (4) is false.¹² Merely changing the meaning of ‘knowledge’ from justified true belief to reliably-formed true belief does not qualitatively change justified true belief or reliably-formed true belief. Such a semantic change does not, for example, change the nature of justification, nor does it change which belief-forming methods are reliable. Neither justified true belief nor reliably-formed true belief qualitatively changes at t ₁. On neither interpretation has the explicative engineer operated on knowledge (so interpreted) at all.

On the third plausible interpretation, ‘knowledge’ denotes the topic associated with ‘knowledge’. Now, according to the Austerity Framework, this is something that qualitatively changes at t ₁: topics are things that conceptual engineering operates on. The problem is that Cappelen’s topics are gerrymandered and grue-like, uninteresting from the relevant theoretical perspective. In the present example, the topic associated with ‘knowledge’ consists of justified true belief before t ₁ and reliably-formed true belief after t ₁. But this topic is a mere theoretical construct: it is an amalgam, stipulated to consist of disjoint parts, designed simply so that Cappelen may endorse (1) and (2). From a purely epistemological perspective, the stipulation is ad hoc: it has no role to play in epistemological theorising. For example, there is no reason for epistemologists to investigate the amalgam as such—they need not account for ‘what happened to justification at t ₁’, nor to explain ‘why reliable belief formation was not part of the amalgam until t ₁’, and so on. Once justified true belief and reliably-formed true belief have been accounted for, the relevant theoretical work has been done. Cappelen’s topic, the amalgam consisting of justified true belief before t ₁ and reliably-formed true belief after t ₁, is not an additional target of epistemological theorising. And accordingly it is of no particular epistemological interest that conceptual engineering can operate on such a phenomenon.

On the available interpretations of (4), then, (4) is either false or epistemologically uninteresting. Either way, Cappelen has not identified an interesting sense in which, from the relevant theoretical perspectives, conceptual engineering operates on (or is otherwise ‘about’) knowledge, freedom, and so on.

For completeness, I note that there is another phenomenon that, according to the Austerity Framework, conceptual engineering operates on: semantic meanings. The explicative epistemologist, in our example, successfully numerically changes the intension of ‘knowledge’. But this observation is not helpful in the present context: changing the intension of ‘knowledge’ does not, as such, constitute a change to any epistemological phenomena.¹³ It is a matter of mere semantics. The explicative epistemologist has not changed the nature of justified true belief or reliably-formed true belief, for example, but has merely changed which of those we call ‘knowledge’. This is not an epistemologically interesting sense in which conceptual engineering has operated on (or is otherwise ‘about’) knowledge.

Let me summarise. Suppose that an epistemologist objects to conceptual engineering on the grounds that it is mere concept-fiddling. Cappelen’s response is that, to the contrary, conceptual engineering also operates on knowledge. From the epistemologist’s perspective, however, this response is not compelling. Given the Austerity Framework, the epistemologist can seek to use conceptual engineering to operate on two things in the vicinity. Firstly, she can seek to qualitatively change an epistemologically-uninteresting amalgam of historical semantic meanings of ‘knowledge’; secondly, she can seek to numerically change the semantic meaning of ‘knowledge’. Neither of these options, however, makes conceptual engineering attractive to the epistemologist seeking to investigate knowledge.

4 Scharp on the phenomenon objection

Kevin Scharp presents the phenomenon objection thus:

Here is another objection […]. Philosophy isn’t the study of the concept of truth or the concept of knowledge or any of the other concepts. Instead, philosophy is the study of certain phenomena, like truth, knowledge, freedom, justice and the rest. (Scharp 2020: 413)

Here, in outline, is his response:

If our philosophical concepts are as defective as I have suggested, then there is no reason to expect there to be a property of truth or a property of knowledge or any of the rest. At least, not if one thinks of the property of truth as anything like what our concept of truth leads us to think it would be like […]. If the [constitutive] principles for these concepts are inconsistent, then no property can satisfy them. If they are seriously inconsistent, then no property can even come close to satisfying them. […] Hence, it makes most sense to think of philosophy as the study of certain concepts—there isn’t much else for it to be about. […] The very idea that there is something like truth or knowledge or freedom or justice or virtue for us to investigate at all is inconsistent. Of course, we have the concept of truth and the concept of knowledge and all the rest, and philosophy is primarily the study of these concepts. (pp. 413–414)

Thus, Scharp responds to the phenomenon objection by characterising philosophy, contrary to the objection, as the study of concepts.

There is, however, an interpretive difficulty here.¹⁴ Scharp argues that there are no properties of truth, knowledge, etc., that are anything like we are led to believe by our concepts of truth, knowledge, etc. But Scharp appears to conclude that there are no properties of truth, knowledge, etc., simpliciter. The latter is particularly salient in Scharp’s final comments on the objection:

So if there is no such thing as truth or knowledge or freedom or virtue, then what is there? We don’t know. And we won’t know until we have done far more conceptual engineering. (p. 414)

Here, as I read Scharp, he takes himself to have argued that there are no properties of truth, knowledge, etc., at all. So interpreted, Scharp’s view is this: the concepts truth, knowledge, etc., are inconsistent; so there are no phenomena (truth, knowledge, etc.) or properties (of truth, of knowledge, etc.) to serve as the targets of philosophical investigation; so philosophy is best understood as an investigation into the inconsistent concepts.

In contrast, on the alternative interpretation, Scharp accepts that there may be properties of truth, knowledge, etc., but denies that such properties are anything like our concepts lead us to think. Moreover, due to the inconsistency of the associated concepts, such phenomena (or properties) would be unavailable for philosophers to investigate.

[T]here is no property of being true, not if that property is anything like what the concept of truth leads us to think it would be like. Philosophy cannot be about the property of being true because there is no property of being true for philosophers to investigate. (pp. 413–414, my emphasis)

So interpreted, Scharp is arguing that philosophers should not investigate truth because the property (if it exists) does not satisfy the constitutive principles for truth. There are two ways Scharp might defend this point. Firstly, he might argue that, if truth does not satisfy the constitutive principles for truth, then philosophers simply have no idea what kind of thing truth is, which renders truth epistemically inaccessible to them. Secondly, he might argue that, if truth does not satisfy the constitutive principles for truth, the faulty concept would derail any investigation into truth, rendering the investigation worthless. Either way, Scharp concludes that philosophers should instead study truth itself. Mutatis mutandis for knowledge, freedom, etc.

On both interpretations, we can see Scharp’s response as consisting of two parts. He first argues from the inconsistency of truth, knowledge, etc., to the claim that philosophers aren’t in the position (for one reason or another) to investigate truth, knowledge, etc. He then concludes that philosophers are, or should be, investigating the inconsistent concepts instead. For present purposes, I grant the first part of Scharp’s response. However, even if Scharp is right that philosophers are not in the position to investigate truth, knowledge, etc., it does not follow that philosophers are, or should be, investigating the inconsistent concepts instead.

Consider first a case from the sciences. Scharp has elsewhere argued (2013b: 433–435) that the Newtonian concept mass is inconsistent. Let us suppose that Newtonian physicists were therefore not in the position to investigate mass. Even given this supposition, though, we would not say that Newtonian physicists were (or should have been) investigating the concept mass instead. It is better to conclude that they were (permissibly) investigating various phenomena in the vicinity, such as inertial mass and rest mass, albeit from a confused vantage point that prevented them from properly distinguishing the phenomena. Here’s another example. For Scharp, it is tempting to conclude that astronomical discoveries in the early 21st century showed planet to be inconsistent—an inconsistency that astronomers corrected when they redefined ‘planet’ in 2006. Let us suppose that planetary scientists prior to 2006 were therefore not in the position to study the property of planethood. Even given this supposition, though, we would not conclude that those planetary scientists were (or should have been) investigating the concept planet instead. It is better to conclude that they were (permissibly) investigating various phenomena in the vicinity, Mercury through Pluto, albeit from a confused vantage point that prevented them from distinguishing astronomically salient phenomena.

Now, in such cases, it is appropriate for the researchers in question to spend some time investigating their concepts. Once the inconsistencies of mass and planet were discovered, the respective scientists were right to investigate those inconsistencies. But it is a mistake to claim that the scientists were (or should have been) doing this instead of investigating phenomena. The investigation into the inconsistent concepts was subsidiary, important only insofar as it supported the primary investigation into physically or astronomically interesting phenomena in the vicinity.

The point carries across to philosophy. Firstly, according to Scharp, truth is inconsistent and, therefore, philosophers are not in the position to investigate truth. Nonetheless, on the present line of thought, philosophers have been (permissibly) investigating phenomena in the vicinity. Perhaps philosophers have been investigating what Scharp (2013a,b) calls ascending truth and descending truth, albeit from a confused vantage point. But there are other phenomena besides. For example, the inconsistency of truth notwithstanding, philosophers are in the position to notice that, whenever p&q, it is also p; and it is appropriate for philosophers to investigate why that is the case, what other similar relationships hold, and so on. Secondly, according to Scharp, knowledge is inconsistent and philosophers are not in the position to investigate knowledge. Nonetheless, again, philosophers have been (permissibly) investigating phenomena in the vicinity. Perhaps philosophers have been investigating justified true belief, reliably-formed true belief, contextualist knowledge, and so on. And there are other phenomena besides. For example, the inconsistency of knowledge notwithstanding, philosophers are in the position to notice that agents often judge the world to be a certain way, and to notice that they act on that basis; and it is appropriate for philosophers to investigate and theorise about the sources of such judgements, to identify and explain patterns in the relationship between such sources and the likelihood of a subsequent action’s success, and so on.

So, pace Scharp, there is no simple inference from the claim that philosophers are not in the position to investigate truth, knowledge, etc., to the conclusion that philosophers are, or should be, investigating the inconsistent concepts truth, knowledge, etc., instead. Moreover, if we take the analogy to the sciences seriously, we will be sceptical of the conclusion. Even if truth, knowledge, etc., are inconsistent, then investigations into those inconsistencies are subsidiary to the primary investigations into alethic, epistemological or other interesting phenomena in the vicinity.

5 The strong phenomenon objection

It is helpful, at this point, to look more closely at the phenomenon objection. Neither Cappelen nor Scharp spell the objection out in detail, making it difficult to pin down. In this section, I clarify and strengthen the objection, before discussing the challenges conceptual engineers face in responding to (what I will call) the strong phenomenon objection.

Consider Cappelen’s formulation: someone who views philosophy as ‘about’ phenomena ‘should be prejudiced against the very idea of conceptual engineering’, because conceptual engineering advocates ‘concept-fiddling’ (2018: 146). So construed, the phenomenon objection might be reconstructed thus:

• (P1) Philosophy is about phenomena—knowledge, freedom, the right and the good, etc.—rather than concepts of those phenomena.

• (P2) Conceptual engineering advocates concept-fiddling (or term-fiddling).

• (C) Philosophers should not engage in conceptual engineering.

Cappelen responds by denying (P2): he claims that conceptual engineering does not advocate concept-fiddling, but instead operates on phenomena.

Now, as it stands, (C) is far too strong. To see this, consider an analogy. The sciences investigate phenomena—geologists study rocks, mathematicians study numbers, etc.—rather than concepts of those phenomena. But conceptual engineering plays an important, ancillary role in the sciences. Geologists have developed technical definitions of ‘rock’, as have mathematicians of ‘number’. Such technical definitions help investigators to state problems and ask questions, enabling and facilitating subsequent investigations. Given the general success of the sciences, the default view is thus that conceptual engineering is an appropriate, ancillary tool for investigating phenomena in the sciences. But this plausibly implies that scientists should sometimes engage in conceptual engineering—or, at the very least, that it’s not the case that they should not engage in it.

The point carries across to philosophy. Even if philosophy is an investigation into phenomena, there is an ancillary role for conceptual engineering to play. Technical definitions of ‘possible world’ facilitate investigations into modality, technical definitions of ‘rigid designator’ facilitate investigations into semantics, technical definitions of ‘autonomy’ facilitate investigations into morality, and so on. It should not be controversial that philosophers should sometimes engage in conceptual engineering. As it stands, even given the premises, (C) is not particularly plausible.

But, importantly, many supporters of conceptual engineering do not merely argue that philosophers should be engaged in conceptual engineering; they argue that philosophers can and should use conceptual engineering to play the pivotal role in solving philosophical problems. The three views I introduced above—Olsson’s explicative epistemology, Haslanger’s ameliorative project, and Scharp’s radical therapeutic program—are all examples of such a view.¹⁵ Olsson plans to tackle and solve epistemological problems by redefining key epistemological terms; Haslanger plans to answer social philosophical questions of the form ‘What is X?’ by ameliorating the concept expressed by ‘X’; and Scharp plans to tackle and solve nearly all philosophical problems by replacing inconsistent concepts with newly engineered, consistent alternatives. On such views, as I understand them, conceptual engineering is to be given (what I call) the decisive role in solving at least some philosophical problems.

Now, at face value, this proposed role for conceptual engineering in philosophy outstrips the role it plays in the sciences. Geologists tend not to afford conceptual engineering the decisive role when tackling geological problems, as such problems would not generally be solved by redefining ‘rock’ or engineering rock. And mathematicians tend not to afford conceptual engineering the decisive role when tackling number theoretical problems, as such problems would not generally be solved by redefining ‘number’ or engineering number. After all, the process of conceptual engineering is no substitute for empirical or logical investigation. Conceptual engineering cannot evidence causes, discover structures, observe correlations, explain new phenomena, or prove theorems. At face value, then, the role that conceptual engineering plays in the sciences is merely ancillary, restricted to improving technical terminology: it is not given the decisive role in solving scientific problems.¹⁶ While conceptual engineering may help investigators to state problems and ask questions, enabling and facilitating subsequent investigations, at face value it does not thereby solve the problems being investigated.

This puts significant pressure on the supporter of conceptual engineering. If philosophy is an investigation into phenomena, and if the sciences—the archetypal successful investigation into phenomena—afford conceptual engineering a merely ancillary role, then ceteris paribus we should expect conceptual engineering to play a merely ancillary role in philosophy.

These reflections suggest that, as formulated above, the appropriate target of the phenomenon objection has been misidentified. The appropriate target is not the view that philosophers should engage in conceptual engineering. It is rather the view that philosophers should afford conceptual engineering the decisive role when tackling philosophical problems. I therefore recommend reformulating the conclusion thus:

• (Cʹ) When tackling philosophical problems, conceptual engineering should be afforded at most an ancillary role.

This is my preferred formulation of the conclusion.

We cannot simply replace (C) with (Cʹ), however, as (P1) and (Cʹ) are formulated in different terms: (P1) is formulated in terms of what philosophy is ‘about’, whereas (Cʹ) is formulated in terms of how to tackle philosophical problems. We could try to bridge the gap by adding an additional premise of the form: ‘if philosophy is about phenomena, then philosophical problems are such-and-such’. But, for the sake of simplicity, I prefer to reformulate (P1):

• (P1ʹ) Philosophical problems principally concern philosophical phenomena—knowledge, freedom, the right and the good, etc.

So formulated, the first premise describes the nature of philosophical problems rather than philosophy itself. Nonetheless, (P1ʹ) is underpinned by the view that philosophy is an investigation into philosophical phenomena.

Turn now to (P2). The claim that conceptual engineering advocates concept-fiddling leaves open the possibility that conceptual engineering advocates many things, only one of which is concept-fiddling. But the phenomenon objection only gets off the ground when we assume—plausibly—that concept-fiddling is all that conceptual engineering advocates: that conceptual engineering as such operates on concepts (or terms) and does nothing else. Any other effects—enhanced social justice, deepened understanding of phenomena, qualitative changes to philosophical phenomena—are indirect or downstream causal consequences. This suggests the following reformulation:

• (P2ʹ) Conceptual engineering as such operates on concepts (or terms) and does nothing else.

Finally, notice that the inference from (P1ʹ) and (P2ʹ) to (Cʹ) is not valid. To make the inference valid, we need an additional premise to the following effect: methods like conceptual engineering should play at most an ancillary role when tackling problems like philosophical problems. Inserting such a premise, and relabelling for ease of subsequent discussion, we obtain the following:

• (P1) Philosophical problems principally concern philosophical phenomena—knowledge, freedom, the right and the good, etc.

• (P2) Conceptual engineering as such operates on concepts (or terms) and does nothing else.

• (P3) When tackling problems that principally concern phenomena, a process that as such operates on concepts (or terms) and does nothing else should be afforded at most an ancillary role.

• (C) When tackling philosophical problems, conceptual engineering should be afforded at most an ancillary role.

And, from (C), we can plausibly infer the following:

• (Corollary) Philosophers should not afford conceptual engineering the decisive role when tackling philosophical problems.

I call that the strong phenomenon objection.¹⁷

If successful, the strong phenomenon objection poses a problem for conceptual engineers. The view that conceptual engineering should have a merely ancillary role when tackling philosophical problems would be a significant concession, substantially weakening its purported promise. This is true even for philosophers who defend ameliorative projects, projects of conceptual engineering that have a significant political component. Ameliorators (in the Haslangerian tradition) explicitly see their projects as tackling philosophical problems.¹⁸ For example, in Haslanger 2000/2012, she states that her.

priority in this inquiry is not to capture what we do mean, but how we might usefully revise what we mean for certain theoretical and political purposes. (p. 224, my emphasis)

The aim of ameliorative projects is not to ignore the philosophical question and to pursue social justice instead. The aim is rather to recognise that tackling philosophical problems about social kinds can be part of the social justice project. Haslanger does not seek to ignore the philosophical question ‘What is it to be a woman?’, but rather seeks to answer the question using a theoretical framework that (i) puts social hierarchies of privilege and subordination at centre stage, and (ii) can serve a political function in the pursuit of enhanced social justice. Ameliorators, then, are not mere activists. Like other recent conceptual engineers, they aim to solve philosophical problems by engineering relevant concepts or terms. The strong phenomenon objection, if successful, directly undermines the role afforded to conceptual engineering by its supporters.

Moreover, there is no easy response to the strong phenomenon objection. To draw this out, I consider each premise in turn.

(P1). The first premise is underpinned by a substantive view of philosophy, namely that philosophy is an investigation into philosophical phenomena. This is a well-known view within the analytic tradition—see e.g. Kornblith 2002, Williamson, 2007, Deutsch, 2021–but is by no means universal. We have already seen that, for Cappelen, philosophy is concerned (at least in part) with operating on phenomena; we have also seen that, for Scharp, philosophy is principally the study of concepts. There is also ongoing work in experimental philosophy that is naturally construed as investigating our (folk) concepts of knowledge, freedom, the right and the good, etc.; there are many substantive defences of broadly ‘conceptual analytic’ methods in the recent literature; and, more generally, there are plenty of ways to conceive of philosophical problems other than as problems arising during the investigation of philosophical phenomena.¹⁹ So there is clear precedent for denying (P1).²⁰

But, for the supporter of conceptual engineering, this is not enough. While (P1) is underpinned by a substantive view of philosophy, it also encodes a substantive understanding of philosophical problems. So, if the supporter of conceptual engineering responds to the strong phenomenon objection by rejecting (P1), she needs to offer an alternative understanding of philosophical problems—one on which those problems can be solved by conceptual engineering.²¹ Standard ways of understanding philosophical problems, however, tend not to be suitable. For example, when philosophical problems are seen through the lens of conceptual analysis or experimental philosophy, conceptual engineering (unsurprisingly) appears inappropriate for tackling them. At face value, a more promising line is to understand philosophical problems as problems about which concepts we should have. But, while suggestive, it remains unclear (to me, at least) why this would be a plausible interpretation of many philosophical questions. A question such as ‘Is free will compatible with determinism?’ is easy to understand as concerning free will and determinism, or the concepts thereof, or even folk intuitions on the matter—but I find it difficult to interpret the question as concerning which concepts we should have. Of course, such a view might be made to work. The point here is simply that, for the conceptual engineer who wants to respond to the strong phenomenon objection by denying (P1), work remains to be done.

(P2). The second premise encodes a standard conception of conceptual engineering as a process for improving our conceptual or linguistic repertoires—where such improvements are made by operating on (contents of) concepts and terms. Like (P1), there is precedent for denying (P2): Nado characterises conceptual engineering as operating on categorisation procedures, and Nimtz characterises it as operating on social norms.²² However, similarly to the above, such denials of (P2) do not constitute successful responses to the strong phenomenon objection. The objection does not turn on the thought that conceptual engineering operates on concepts or terms; it turns on the thought that conceptual engineering operates. Whereas philosophy is (given (P1)) an investigative endeavour, conceptual engineering is a practical one: the former helps us to understand things, the latter to change things. This is why, at face value, conceptual engineering is inappropriate for solving philosophical problems. It isn’t important whether the objects of change are concepts, terms, categorisation procedures or social norms. Denying (P2), at least by denying that conceptual engineering operates on concepts or terms, is not obviously a promising strategy for responding to the strong phenomenon objection.

(P3). From the present perspective, the third premise looks highly plausible. It captures the intuitive thought that underlies the whole objection—namely, that we do not solve problems concerning phenomena by amending our terms or concepts. The premise derives its plausibility, in large part, from the sciences. Scientific problems are archetypal examples of ‘problems principally concerning phenomena’ and, at face value, scientific problems are not solved merely by amending our terms and concepts. This suggests two strategies for denying (P3): arguing that philosophy is exceptional insofar as some of its problems can be solved merely by amending terms and concepts; or else arguing that some scientific problems can be solved merely by amending terms and concepts after all, and that some philosophical problems are like those scientific problems in that respect.

Firstly, then, one could deny (P3) by endorsing a view on which philosophy is exceptional in some way. The challenge on this line of response is to say why philosophy is exceptional in such a way that makes sense of the proposed role for conceptual engineering. For one example, it would not be enough to claim that philosophy is exceptional because it investigates normative phenomena; one would also need to say why problems about normative phenomena in particular are suited to being solved merely by amending terms and concepts. Perhaps such an explanation can be provided; but I, at least, cannot see why the distinction between normative and non-normative phenomena would mark the boundary for the appropriate use of conceptual engineering for investigating phenomena. The challenge for the supporter of conceptual engineering adopting this strategy, then, is to give a plausible account of what is distinctive about philosophy that bears out the particular suitability of conceptual engineering as a philosophical method.

Secondly, one could deny (P3) by arguing that some scientific problems can be solved merely by amending terms and concepts, and that some philosophical problems are like those scientific problems in that respect. This view may, at first sight, appear to be a non-starter. After all, as noted above, conceptual engineering cannot evidence causes, discover structures, observe correlations, explain new phenomena, or prove theorems—in short, it cannot do the things we expect a scientific problem-solving method to do. Nonetheless, it is the strategy that I adopt in what follows. I argue first (§ 6) that, contrary to intuition, conceptual engineering is what plays the decisive role in tackling (and possibly solving) at least one scientific problem. I then argue (§ 7) that at least one philosophical problem is like that scientific problem in that respect. I then extend the case (§ 8) to many traditional philosophical problems. I conclude that, contrary to the strong phenomenon objection, philosophers should afford conceptual engineering a decisive role when tackling many traditional philosophical problems.

In this section, I have developed the strong phenomenon objection and explored the challenge it poses to the conceptual engineer. I have suggested that, whichever strategy one adopts, work needs to be done. In the remainder of the paper, I develop my preferred line of response.

6 The Pluto problem

At the turn of the century, ‘planet’ was effectively defined by nine instances, Mercury through Pluto, with an implicit generalisation to other solar systems. However, astronomical discoveries had highlighted the need for terminological change. It had become clear that, in terms of important astronomical properties (orbit, constitution, mass, etc.), Pluto was quite different to the other eight instances, and very similar to thousands of trans-Neptunian objects, especially those in the Kuiper belt. However, nothing in contemporary astronomy called for any particular terminological refinement. The issue first arose to prominence when the Hayden Planetarium in New York eliminated ‘planet’ as a standalone technical term from their displays, opting instead for ‘terrestrial planet’, ‘Jovian planet’ and ‘Kuiper belt object’.²³ Later, different astronomers offered various definitions of ‘planet’ in terms of important celestial properties—such as (a) orbiting the sun, (b) being in hydrostatic equilibrium, and (c) having cleared one’s orbit of debris—with different definitions carving up astronomical phenomena in different ways. Two definitions rose to prominence, one [(a)+(b)] counting Pluto and thousands of trans-Neptunian objects as planets, the other [(a)+(b)+(c)] not. Given the need to co-ordinate terminology, and in lieu of a clear theoretical or empirical reason to settle on either definition, the IAU resolved the matter by vote. In this way, the problem was solved. In the astronomically interesting sense, Pluto was deemed not to be a planet. Had the vote gone another way, the solution would have been different.

It is worth pausing to consider what problem the astronomers were tackling. At first blush, we might identify the problem with the question ‘Is Pluto a planet?’. But that cannot be right. That question, posed prior to 2006, had an easy answer: ‘yes’.²⁴ More generally, it is a mistake to think of investigations into phenomena as principally aiming to answer specific questions that are set out in advance. It is better to think of the aim of investigations into phenomena as being to understand, explain or theorise the relevant phenomena. From this perspective, problems that arise during such investigations are ultimately challenges to understanding, explaining, or theorising those phenomena. The astronomers’ problem, then, is better construed as a complex set of challenges concerning:

• (i) what the structure of the solar system is;

• (ii) what within that structure the phenomena of interest to astronomy are;

• (iii) whether any of those phenomena align roughly with the prior taxon consisting of Mercury through Pluto; and

• (iv) which of those phenomena Pluto is an instance of.

Call this complex set of challenges the Pluto problem.

Of course, we might still gloss the Pluto problem as ‘Is Pluto a planet?’. Such a gloss is simple and convenient. The present point is simply that we should not identify the Pluto problem with its gloss. Solving the Pluto problem did not require astronomers to explicitly answer the question ‘Is Pluto a planet?’, but rather to advance their understanding of astronomical phenomena in relevant respects. And, by engaging in conceptual engineering, the astronomers:

• (i) imposed a theoretically informed structure on the solar system;

• (ii) picked out what within that structure the phenomena of interest to astronomy were; and.

• (iii) highlighted a phenomenon that aligned roughly with the prior taxon consisting of Mercury through Pluto—a phenomenon that, as it happens,

• (iv) Pluto is not an instance of.

Which is to say, conceptual engineering played the decisive role when astronomers solved the Pluto problem.

There are two features of this case to draw out. The first feature is this. The astronomers were at a stage in their inquiry during which they faced a difficult choice about how to define some apparent phenomenon of interest, which we might denote ‘planethood’.²⁵ Two factors made this choice difficult. First, there were two theoretically informed characterisations—descriptions composed of object-level terms drawn from currently accepted theories—of ‘planethood’ available. Second, further investigation into ‘planethood’ would not have told decisively in favour of either characterisation: they faced a question not about the nature of (say) Pluto, but about how to delineate various phenomena. Call such a stage of inquiry a choice point.²⁶

To move beyond a choice point, investigators must engage in conceptual engineering: they must decide how to define the relevant technical term. They have multiple theoretically informed characterisations that can serve as definitions for the term, but lack a decisive reason to favour one characterisation over the others. Thus, in making the decision, theorists must make an informed judgement about which characterisation will serve as the best definition for the term, and subsequently treat that characterisation as a technical definition of the term.

There is no algorithm for making such decisions, and there may be no complete list of the criteria that theorists can use. But the kinds of criteria that might typically play a role are familiar: how well the characterisations cohere with the broader theory; how likely the characterisations are to lead to future developments and discoveries; how simple the characterisations are; and so on.²⁷ These conceptual engineering decisions, then, are not arbitrary; resolving a choice point is an important step in ongoing theory development.

The second feature of the case is this. The astronomers were tackling a scientific problem—the Pluto problem—whose solution depended on how the choice point was resolved. Defining ‘planet’ in one way yielded one solution (glossed as ‘Pluto is not a planet’), and defining it another way would have yielded another solution (glossed as ‘Pluto is a planet’). Call such a problem—one whose solution depends on how a choice point is resolved—a puzzle. Puzzles, then, are solved by conceptual engineering. They are solved by deciding how to define the technical term intended to denote the apparent phenomenon of interest.

On the present line of thought, then, at least one scientific problem is a puzzle, and conceptual engineering is what solves puzzles. So scientists rightly afforded conceptual engineering the decisive role in solving at least one scientific problem.

7 The problem of free will and determinism

Let me now turn to a philosophical problem, specifically the problem we might gloss as ‘Is free will compatible with determinism?’.

If we are thinking of philosophy as investigating phenomena—as I am—then, as above, we should not identify the underlying problem with its gloss. The underlying problem is a challenge to understand, explain, or theorise various metaphysical phenomena—especially those associated with agential self-direction and deterministic spatiotemporal relations—and it does not turn essentially on (say) the semantics and metasemantics of any particular question. The underlying problem is better construed, for example, as a complex set of challenges concerning: the different senses in which an agent might have self-direction; which (if any) of those constitute metaphysically interesting properties (perhaps in the specific sense of underpinning moral responsibility); and which (if any) of those are compatible with a deterministic universe. For concreteness, call that the problem of free will and determinism.

There are at least two theoretically informed characterisations of the apparent phenomenon denoted by ‘free will’, roughly: (a) the power to do otherwise; and (b) being such that, had one chosen to do otherwise, one would have done otherwise. Moreover, given the considerable investigation into free will that has already taken place, we might reasonably expect further investigation into associated phenomena not to tell decisively in favour of either characterisation. If we find that thought plausible—more on which below—we will conclude that we are at a choice point with respect to ‘free will’. Moreover, depending on which characterisation we choose to serve as a definition for ‘free will’, we get different solutions to the original problem—a different judgement about which sense of agential self-direction is metaphysically most interesting, and a different conclusion about its compatibility with determinism. We might gloss those solutions as ‘incompatibilism’ and ‘compatibilism’, respectively.²⁸ From this perspective, the philosophical problem of free will and determinism is a puzzle; it is solved by conceptual engineering.

Is it plausible that, as I suggested above, further investigation will not tell decisively in favour of either characterisation, (a) or (b)? I believe so. I doubt there is a unique ‘intuitive conception’ of free will, and I doubt there is a unique property of metaphysical interest in the vicinity. Moreover, we cannot get an independent, epistemic grip on free will—like we might with (say) concrete cognitive processes underpinning action—to investigate its make-up. Just like in the case of ‘planethood’, there is an essential element of choice in how we characterise the apparent phenomena.

One might wonder whether metasemantic externalism speaks against this line of thought: perhaps the reference of ‘free will’ is fixed by externalist mechanisms that guarantee that free will is the metaphysically salient property in the vicinity, so that there is a fact of the matter whether free will is (a), (b) or neither. However, firstly, metasemantic externalism does not guarantee that there is a metaphysically salient property in the vicinity. At best it would establish that if there is such a property, then ‘free will’ refers to it. And, secondly, even so, this would not explain how further investigation could decide between (a) and (b): metasemantic externalism does not imply that we have epistemic access to the nature of the external properties that fix the reference of our words.

So how, on the present view, is the problem of free will and determinism to be solved in practice? Consider the individual philosopher. First, she engages with the relevant literature and does preliminary research to obtain an informed theoretical understanding of the choice. Second, drawing on that understanding, she decides which of (a) and (b) is more metaphysically interesting in the relevant sense. Third, she offers that characterisation as a technical definition of ‘free will’. Finally, with metaphysical phenomena so carved, she glosses the solution she has obtained to the original problem either with ‘free will is not compatible with determinism’ or with ‘free will is compatible with determinism’.

Consider now a community of philosophers. There are various ways in which they might proceed. Following the astronomers’ precedent, they might take a vote on whether (a) or (b) is more theoretically interesting or practically useful for the metaphysical investigation at hand. Or, more likely, the community will naturally split into two factions, each endorsing a different characterisation, developing both theoretical avenues for future philosophers to choose between.

These comments accord surprisingly well with intuition. If a philosopher believes that one definition of ‘free will’ is better than the other, she may well believe that there is a clear solution to the problem of free will and determinism. For example, if she believes that ‘free will’ should be defined as the power to do otherwise, she will likely accept that there is a clear solution to the problem (‘incompatibilism’); a philosopher convinced that ‘free will’ should be defined in terms of counterfactual choices will agree that there is a clear solution, but disagree on what that solution is. In contrast to such philosophers, those who feel ill-placed to make an informed judgement about how to characterise the phenomenon would be unsatisfied with merely making an arbitrary choice. And, if a future generation of philosophers all endorse the same definition, then they will likely agree that the problem of free will and determinism has been solved.

8 Other philosophical problems

I have so far argued that at least one traditional philosophical problem is a puzzle, and is thus to be tackled (and perhaps solved) by conceptual engineering. I now sketch a case that the conclusion generalises—that many traditional philosophical problems should be tackled in this way.

To begin, note the following: many traditional philosophical problems are glossed as questions of the form ‘What is X?’. Such problems are typically of long-term philosophical interest just when there are conflicting theoretically-informed characterisations of X available, with little prospect of further investigation into the relevant phenomenon telling decisively in favour of one particular characterisation. For example, if there were no conflicting theoretically-informed characterisation of (say) ‘knowledge’, ‘truth’, ‘the right and the good’ or ‘gender’ available, then the corresponding ‘What is X?’ questions would not draw much interest from philosophers. And, in each case, the longevity of the debate is explained in part by noting that further investigation is unlikely to settle the issue: firstly, I doubt there are unique ‘intuitive conceptions’ of knowledge, truth, the right and the good or gender; secondly, I doubt there are unique properties of metaphysical interest in the vicinities thereof; and, thirdly, we cannot get an independent, epistemic grip on knowledge, truth, the right and the good, or gender—like we might with (say) cognitive processes underpinning memory, mathematical properties of logical systems, cultural processes affecting folk moral judgements, or biological sex—to investigate their make-ups. So the corresponding ‘What is X?’-questions indicate choice points and are glosses of puzzles.

But there is nothing special in this respect about knowledge, truth, the right and the good, or gender. I could have sketched the above argument (mutatis mutandis) for any traditional philosophical ‘What is X?’-question, of which there are many. If that is right, then many traditional philosophical problems—those glossed by ‘What is X?’-questions—are puzzles and, as such, conceptual engineering should be afforded the decisive role in tackling them.

As before, a philosophical problem glossed by ‘What is X?’ is not to be identified with its gloss. In each case, the underlying problem is better construed as a complex set of challenges to understand, explain or theorise relevant phenomena. For example, ‘What is knowledge?’ might gloss a complex set of challenges pertaining to: the variety of mental states an individual might be in; whether any of those states have epistemologically interesting properties (such as being factive); and whether any of those states aligns roughly with our use or understanding of ‘knowledge’. To be explicit, there is no algorithm to articulate those challenges, and there may be some disagreement about the details. Regardless, one solves a philosophical problem glossed by ‘What is X?’ by advancing one’s understanding of the relevant phenomena with respect to an appropriate collection of underlying challenges.

More speculatively, I suggest that we might obtain solutions to many other traditional philosophical problems by solving ‘What is X?’-style problems. For example:

• Fixing a solution to the problem glossed by ‘What is free will?’ might yield solutions to the problem of free will and determinism, the problem of whether we (humans) have free will, and so on.

• Fixing a solution to the problem glossed by ‘What is knowledge?’ might yield a solution to the Gettier problem, the problem of radical scepticism, and so on.

• Fixing a solution to the problem glossed by ‘What is truth?’ might yield a solution to the liar and other semantic paradoxes.

• Fixing a solution to the problem glossed by ‘What are the right and the good?’ might yield solutions to a variety of first-order moral dilemmas such as the trolley problem, the drowning child, and so on.

• Fixing a solution to the problems glossed by ‘What is gender?’ and ‘What is sex?’ might yield a solution to problems surrounding the relation between sex and gender, problems about the reality of gender, and so on.

Whether or not this is right depends on the details of each case, and there will be complexities. For example, while formal accounts of truth (e.g. Tarski 1983) tend to yield solutions to the liar paradox, informal accounts (such as the correspondence and coherence theories) tend not to. So, in general, we cannot determine in advance whether we would obtain a solution to a given problem involving ‘X’ by solving the problem glossed by ‘What is X?’. Nonetheless, I take it to be plausible that, once we solve a ‘What is X?’-style problem—of which there are many in philosophy—we will obtain the solutions to a range of additional philosophical problems in the vicinity. If this speculation is right—and nothing herein turns on it being so—then conceptual engineering might turn out to have the decisive role in tackling a wide range of traditional philosophical problems.

9 The picture

I have presented philosophy and the sciences alike as investigations into phenomena. On such a view, I have suggested, we should not identify scientific and philosophical problems with particular questions. Instead, we should think of those problems as potentially complex sets of challenges to understand, explain, or theorise relevant phenomena. For some of these problems, such as the problem of determining Pluto’s constitution, empirical methods will play the decisive role in reaching a solution. In other cases, such as the problem of proving Fermat’s Last Theorem, mathematical methods play the decisive role. And in yet other cases—puzzles such as the Pluto problem and the problem of free will and determinism—conceptual engineering will play the decisive role.

To say that a class of methods plays the decisive role is not to say that other methods are irrelevant. For example, given the obvious practical limitations, mathematical methods were essential in determining the constitution of Pluto. And, while conceptual engineering plays the decisive role in solving puzzles, other methods are likewise relevant. When attempting to resolve a choice point, one should be theoretically informed, and complementary methods can help us to better understand the relative theoretical interest of competing characterisations. For example, when tackling the problem of free will and determinism, it would be appropriate to use a variety of strategies: we might assess our intuitions in various actual and counterfactual cases; we might consider psychological research into how decisions are made; we might examine physical deterministic systems and their alternatives; we might consider the ethical implications of denying free will; we might gather intuitive samples and look for commonalities (Kornblith, 2002); we might deploy our ordinary counterfactual reasoning faculties (Williamson, 2007); and so on. Such strategies are merely ancillary and would not solve the problem per se, but they would put us in a better position to solve the problem using conceptual engineering.

It may be tempting at this juncture to mark a possible difference between philosophy and the sciences. Pretheoretically, we would not expect many traditional scientific problems to be puzzles; we would expect most scientific problems to call for logical or empirical methods, leaving the Pluto problem as an outlier. In contrast, I suggested in § 8 that many philosophical problems may turn out to be puzzles. If both points are borne out, then, on the present picture, philosophy is different from the sciences. But such a difference, if it obtains, is not obviously interesting from a philosophical perspective: it is a difference in prevalence, rather than a difference in kind. This having been said, it remains an open question on the present picture just how prevalent puzzles are in the sciences. For example, consider Mark Wilson’s (2006) view of the sciences. On Wilson’s view, applying extant theory to a new phenomenon (a new ‘patch’) typically involves slight and often unnoticed adjustments to the theoretical concepts involved—his examples include force, rigid body, hardness, colour, and many others—resulting in the mere façade of a univocal theory. On such a view, conceptual engineering processes are deeply interwoven with scientific practices and, when faced with a new problem arising from available data, it may be that conceptual engineering processes are called upon to find a solution to the problem. If such a view is on the right lines then, contrary to expectation, many traditional scientific problems may also turn out to be puzzles, suggesting a deep continuity between philosophy and the sciences. So, on the present view, it remains an open question whether (and to what extent) philosophy is different from the sciences.

I close by explicitly addressing the strong phenomenon objection. I have presented philosophical problems as principally concerning philosophical phenomena [consistent with (P1)], and I have understood conceptual engineering as such to operate on terms [consistent with (P2)]. However, against this background, I have given an account of philosophical (and scientific) problems on which conceptual engineering should play the decisive role in tackling a subset of philosophical (and scientific) problems [contrary to (P3)]. On the view developed herein, then, contrary to the strong phenomenon objection, philosophers should afford conceptual engineering a decisive role when tackling (some) philosophical problems.

10 Closing remark

The key insight that underpins the view developed herein is twofold: that the questions that represent philosophical and scientific problems, such as ‘Is Pluto a planet?’ and ‘Is free will compatible with determinism?’, should be seen as mere glosses of underlying problems; and that the underlying problems are sets of challenges in understanding, explaining or theorising relevant phenomena. From this perspective, there is no antecedent reason to expect our particular concepts and terms to be essential to the problems we are tackling. And it is, on reflection, unsurprising that, when one reaches a choice point in one’s inquiry, conceptual engineering should play the decisive role in tackling whatever problem is under investigation.

Conceptual engineers who find this view unattractive, but who want to afford conceptual engineering a decisive role in tackling philosophical problems, will need to offer an alternative response to the phenomenon objection. As part of this, they will need to offer their own account of philosophy and its problems; and they will need to explain why, with philosophy so understood, conceptual engineering should be afforded a decisive role in tackling those problems.²⁹