Downplaying the change of subject objection to conceptual engineering

ABSTRACT

Conceptual engineering projects have been criticized for creating discontinuities of subject-matter and, as a result, discontinuities in inquiries: call this the Change of Subject objection. In this paper, I explore a way of dealing with the objection that clarifies its scope and eventually downplays it. First, two strategies aimed at saving subject-continuity are examined and found wanting: Herman Cappelen’s appeal to topics, and the account in terms of concept function. Second, the idea is introduced that one can begin an object-level inquiry either with a ‘semantically conservative’ approach, whereby semantic change is not permitted, or with a ‘semantically progressive’ approach, whereby semantic change is permitted. This distinction helps one significantly downplay the Change of Subject objection.

KEYWORDS:

• Conceptual engineering
• concepts
• change of subject
• topic
• function
• inquiry

1. The change of subject objection to conceptual engineering

Conceptual engineering is characterized by Cappelen (2018, 3), and Cappelen and Plunkett (2020, 7), as the process of ‘assessing and improving our representational devices’. In the current debate, paradigmatic projects falling under this rubric involve: (a) an inquirer, or a community of inquirers, focussing on the meaning of a certain term (usually regarded as proxy for the concept the term expresses), (b) arguing for such meaning’s defectiveness along a certain dimension, and (c) proposing a change that they deem conducive to some form of representational progress. In the history of analytic philosophy, authors who explicitly go through these steps are Tarski (1935), Carnap (1950) and Quine (1960). In more recent times, we find this methodology explicitly adopted by Haslanger (2000), who aims at offering an ameliorative revision of the concepts of race and gender; and Scharp (2013), who urges a replacement of the truth concept in order to overcome the liar paradox. However, the number of theorists who can be interpreted as pursuing a conceptual engineering project – although they do not explicitly frame their inquiry in this way – is potentially much higher (see Brun 2016). Cappelen (2018) includes (among others) Clark and Chalmers on ‘belief’, and Railton on moral ‘good’ and ‘ought’. One could add, to name but a few authors, Hempel on ‘confirmation’, Craig on ‘knowledge’, Parfit on ‘personal identity’ and Sider on ‘existence’.

The changes operated by conceptual engineers are standardly described in terms of revision, or in terms of replacement. Revision typically occurs when a limited number of relevant features of one and the same concept is changed. Replacement is the introduction of a numerically new concept intended to supplant the initial one. Depending on how one accounts for the semantics of concepts, revision and replacement can, in turn, be thought of as targeting different aspects: for example, a concept’s intension, its extension, or the rules or principles that govern its use.

While different descriptions may suit different paradigmatic projects of conceptual engineering, it is usually agreed that the extent of the conceptual revision may, especially to unsympathetic eyes, become problematic. For, if the changes effected in order to improve a concept were to exceed a certain limit – which might vary from case to case – the conceptual engineer may face an objection, which I will call the Change of Subject Objection.¹ In this paper, I will focus on a formulation of this objection that emphasizes how a change of subject produces a discontinuity in an inquiry (this formulation is to be found especially in Cappelen 2018, 101–102; Cappelen and Plunkett 2020, 12–13). A disclaimer: in what follows, I will for ease of exposition describe conceptual engineering as affecting the intension and extension of a term. While this way of framing conceptual engineering may not do justice to all the projects currently gathered under this terminological umbrella, it will allow me to more fruitfully engage with currently influential views, such as Herman Cappelen’s, which deal specifically with the change of subject objection.

Here is the formulation of the objection I will consider. Note that it is a rather broad formulation, targeted at inquirers whose investigation leads to semantic changes. This breadth of scope is, to my mind, a plus, for there may well be a general problem of subject-matter continuity for any inquirer who faces semantic changes, whether such changes result from one’s will to ‘engineer’ a concept or not. Suppose one’s inquiry starts with some initial first-order question, for instance, ‘What are the Fs?’, ‘Are the Fs such-and-so?’ Suppose these questions are formulated in a language in which the term ‘F’ has a certain intension and extension, which we could imagine to be sufficiently determined for the purposes of the investigation. Suppose now that conceptual change is brought about by the inquirer, in the form of a revision of the intension and/or extension of the term ‘F’ (whether such change results from an ‘engineering plan’, or not, is not relevant at this stage). If the change exceeds certain limits, it seems that, whatever answer one gives at the end of the inquiry, it will not be an answer to the initial question, because one of the terms it contains, ‘F’, has undergone excessive semantic change. Consider the following example inspired by Cappelen and Plunkett (2020).

Philosopher Erika embarks in an investigation aimed at answering the question ‘Is free will compatible with determinism?’ In the language she is employing, ‘free will’ expresses the concept Free will, whose intension could be spelled out as ‘the capacity to do otherwise’. During her investigation, Erika finds that the term (and related concept) better describes a certain number of cases if such intension is changed into ‘being the sole source of one’s actions’. In her eyes, Erika has ameliorated the concept because she has increased the fit between the conditions of its application and the data provided by untutored intuitions. Suppose she now answers the initial question saying ‘Yes, free will is compatible with determinism’. A critic could accuse her of having changed the subject, for the post-engineering term ‘free will’ does not have the same intension as the pre-engineering ‘free will’. The answer that free will is compatible with determinism does not concern the property of being able to do otherwise, it concerns another property. Therefore, it is not strictly speaking an answer to the initial question. The semantic change has produced a discontinuity in the initial inquiry.

Inquiry discontinuity may have effects that threaten the very feasibility or implementation of conceptual engineering projects. For, if an inquiry gets disrupted by subject-change, then the disruptor might cause communication breakdowns with other speakers (more on this in section 6). This might be a hurdle to the smooth completion of a conceptual engineering project. To see this, suppose Erika now wishes for the revised concept of free will to enter the conceptual repertoire of her linguistic community. To do so, she starts asserting ‘Free will is compatible with determinism’, and other germane sentences, using ‘free will’ according to her own (supposedly non-standard) definition. This might, on some occasions, create a ‘positive’ communicative disruption that eventually leads her fellow conversationalists to revise their concept of free will (see the notions of ‘linguistic intervention’ and ‘conceptual activism’ discussed by Sterken 2020; Cantalamessa 2021). It is safe to say, however, that on many other occasions this kind of use would create misunderstandings, and confusion as to what is at stake in the very question under examination. This kind of confusion would plausibly pose a significant obstacle to the achievement of Erika’s goal – namely, the introduction of the revised free will concept within a wider linguistic community. If inquiry disruption has these implications, then the task of addressing the inquiry-discontinuity version of the subject change objection will at some point connect with the task of addressing problems regarding the feasibility and implementation of conceptual engineering projects.

The inquirer targeted by the change of subject objection seems to have two reply options: (i) either argue that (provided certain conditions) no subject discontinuity has been produced; or (ii) argue that the subject discontinuity is not problematic. In Sections 2 and 3, I will examine two responses of the first kind, which invoke the notions of ‘topic’ and ‘concept function’ respectively. After having found these strategies wanting, I will move on to exploring (in Sections 4–6) an answer of the second kind. I will argue that certain inquiries can be approached by factoring in subject-change from the very beginning. This helps one significantly downplay the subject-change objection.

2. Cappelen’s topics

In response to the objection just presented, Cappelen (2018) invokes topics. He argues that, since topics are more coarse-grained than intension and extension, we can have continuity of topic even if the intension or extension of some key concepts has undergone change. In turn, one may say that continuity in topic ensures continuity in inquiry.

Cappelen starts by presenting the notion of samesaying. For many expressions, we can issue disquotational speech reports in terms of samesaying. Suppose that A says ‘Serena is strong’ in context C₁ and that B says ‘Serena is strong’ in context C₂. Suppose A and B use ‘strong’ relative to different comparison classes (say, tennis players and celebrities). As a result, the occurrence of ‘strong’ in C₁ has a different intension and extension than the occurrence of ‘strong’ in C₂. From our speech-reporting context C₃, we can issue a collective, cross-contextual report like the following: ‘A and B both said that Serena is strong’.

Cappelen builds on this result in order to argue for the existence of topics (2018, 113). We may reconstruct his argument as follows. First, as already seen, we can correctly report A and B as both having said that x is F even though their uses of ‘F’ had different intensions and/or extensions (in their respective contexts). If this is the case, then A and B said the same. If A and B said the same, then their topic is the same. If this is the case, then one can have sameness of topic despite differences in intension and/or extension.

In his examples on samesaying, Cappelen seems to focus on speech-reporting contexts (like C₃) where A and B said the same for (hypothetical) current conversational purposes (within C₃). However, purposes can shift within one and the same speech-reporting context. This implies that, with little reflection, it is possible to also correctly describe A and B as not saying the same. The following example illustrates the idea.

Suppose a journalist is collecting evidence regarding whether a certain prisoner was tortured in a detention facility. She examines reports issued by the United Nations, which define torture as ‘any act inflicting severe suffering, physical or mental, in order to obtain information or to punish’, as well as reports by the U.S. department of justice, which defines torture as ‘any act inflicting pain rising to the level of death, organ failure, or the permanent impairment of a significant body function’.² Suppose the prisoner did experience organ failure as a consequence of how they were treated; so, the U.N. documents and the U.S. Justice Department documents both state: ‘The prisoner was tortured’. There is a clear sense in which these reports say the same – that the prisoner was tortured. Yet, the journalist might wish to highlight that strictly speaking, the U.N. documents and the U.S. documents do not say the same. The U.N. documents say that the prisoner was the victim of acts inflicting severe suffering, physical or mental, in order to obtain information or to punish; the U.S. documents say that the prisoner was the victim of acts inflicting pain rising to the level of death, organ failure, or the permanent impairment of a significant body function. Although both statements happen to be true of the prisoner in question, fully spelling them out might help one to stress that, if things had gone otherwise and the prisoner had not suffered organ failure, the documents would have made contradictory claims. This might be considered deeply problematic.

In light of these observations, one may ask: are the U.N. and the U.S. remarks about the same topic? To the extent that they say the same, it seems like they are, for they are about whether the prisoner was tortured, or about torture. To the extent that they do not say the same, it seems like they are not, for they are about whether the prisoner underwent different types of treatment (with certain specific consequences). Which considerations settle the question? In this particular case, there seems to be no determinate answer.

One could reply that, as long as it is possible to report the two occurrences as saying the same in a broad sense of ‘saying the same’, then their topic is the same. This may be compatible with them not saying the same in some other, narrower sense of this term. Yet, the author of the reply could go on, the broad sense of samesaying should have pride of place within our linguistic practices, because it prevents judging that topic changes too much and too often and, therefore, ensures stability in the way conversations are accounted for.

According to this reply, the broad sense of samesaying trades precision for stability as far as topic is concerned. Is that a generally desirable trade-off, though? Many conversational contexts, including contexts where (e.g.) the violation of human rights is at issue, and philosophical contexts, demand higher precision than normal. In these higher-precision contexts, our judgements may speak against samesaying reports in exactly this broader sense. Plus, in the same contexts, one could always grant samesaying in the broader sense, but regard it as too loose to give informative data as to sameness of topic given the standards of precision at hand. Appeals to the broad sense of samesaying can therefore be easily undercut.

The foregoing discussion was focussed on just one example. Yet, I believe the phenomenon I am pointing at can involve any term, and importantly also terms that are of interest to conceptual engineers. I leave it as an exercise to the reader to devise similar examples with terms like ‘marriage’, ‘gene’, ‘knowledge’, and so on.

The upshot of this is that samesaying judgements are unstable.³ In each particular instance of conceptual engineering, it is possible to counter samesaying reports with reports to the effect that the speakers did not say the same. So, in each particular instance of conceptual engineering, it will be difficult to offer convincing arguments about samesaying. Since samesaying determines (or is a diagnostics for) sameness of topic, it is going to be equally difficult, in each particular instance of conceptual engineering, to offer convincing arguments as to sameness of topic.

In all fairness, Cappelen does allow that whether two concepts are about the same topic can be contested (2018, 116, 119), and seems to regard this instability of verdicts as a feature, not a bug. Yet, if my argument is right, it will basically always be possible to contest sameness of topic. If this is so, then it seems like the topic-strategy just fares poorly when it comes to applying it to particular cases. This may raise the question: why commit to a strategy with such systematic application issues? That such application issues obtain looks like a good reason for discarding the strategy, not for adopting it.

It could also be pointed out that all these remarks do is show that the notion of samesaying has not been adequately operationalized. While the notion of samesaying could certainly use some regimentation, I am not very optimistic that such regimentation could be accomplished. First, the notion of samesaying is a term of art; surely, it does rely on judgements about whether speakers ‘said the same’, which Cappelen deems pre-theoretical (cf. Cappelen 2018, 108). Yet, either such judgements are completely untutored, and hence predictably unstable too; or they are in turn based on the notion of ‘what is said’ as a term of art. The problem is whether ‘what is said’ has been adequately operationalized too. A quick review of the current literature shows that there is no agreement as to what ‘what is said’ is.⁴ This is a further source of instability. Second, even if all these problems were solved and ‘samesaying’ were operationalized, it would still harbour a good amount of context-sensitivity. That is, samesaying may remain relative to a number of parameters whose value would have to be fixed contextually, thus potentially generating divergent samesaying verdicts. This by itself would seem sufficient to derive the instability of the notion of topic, and hence to question the effectiveness of the strategy proposed by Cappelen. Therefore, even if my arguments only showed a lack of operationalization, it seems reasonable to doubt that operationalizing ‘samesaying’ will be conducive to a better answer to the change of subject objection.

3. Appeal to functions

Another possible strategy to deal with the change of subject objection is to argue that changes in intension and/or extension need not alter the function of a concept. And as long as the concept’s function is preserved, continuity of inquiry is arguably safe.

This line of argument is compatible with a number of recent approaches. According to Prinzing (2017), a concept’s function is that restricted range of uses corresponding to what the concept was originally ‘designed for’ (intentionally or not). Simion and Kelp (2020) characterize a concept’s function as its etiological function, i.e. the function that it came to acquire as a result of its evolution in a system of practices. Haslanger (2020) and Thomasson (2020) also emphasize the role of function as a guide for conceptual engineering and as a resource to appeal to when it comes to the change of subject objection.

Haslanger, for instance, reconstructs the semantic changes undergone by the concept Family in the last decades. She argues that, although such concept now carves up the logical space differently, by admitting more human groups as families, the concept has maintained its function: providing the means for ‘processing information about the coordination of domestic life, e.g. intimacy, sex, raising of children, economic partnership, intergenerational transfers of traditions and property’ (2020, 250). As a result, we can say that one and the same concept, Family, evolved, and not that one concept, say Family₁, were replaced by a numerically different concept, Family₂. Functional continuity may in turn be recruited in order to argue for inquiry continuity.

The approaches just surveyed all seem to accept that, first, a concept acquires its function as a result of how linguistic and other relevant human practices develop; and, second, that no speaker in particular is responsible for, or in control of, the function that a concept acquires throughout its history. This implies that philosophers who assign a function to a certain concept might get things wrong in a number of ways.

One of the ways in which philosophers might err is in how finely they individuate the concept’s function. Consider the concept Family. It would seem natural to attribute to it the function sketched by Haslanger, namely, ‘tracking the coordination of domestic life’ which includes ‘intimacy, sex, raising of children, economic partnership, intergenerational transfers of traditions and property’, etc. However, one may observe that, prior to (roughly) the late twentieth century, the coordination of such practices in the Western world was based on hetero-normative rules and embedded in a marriage-centred legal system. Post late twentieth century, the coordination of the same practices (in the Western world) became governed by other rules, which were non-hetero-normative and not embedded in a marriage-centred legal system. This may lead one to conclude that the concept Family has two different functions: pre late twentieth century, it has the function of tracking the coordination of domestic life in a hetero-normative, marriage-centred system; post late twentieth century, it has the function of tracking the coordination of domestic life in a non-hetero-normative, non-marriage-centred system.

Which way of individuating the concept’s function is the right one? I wish to argue that attempts to answer this question run into a significant amount of epistemic indeterminacy. To start with, let us suppose that concept functions exist mind-independently.⁵ The evidence at our disposal does not reveal whether the ‘real’ function of a concept is correctly individuated in the coarse-grained or in the fine-grained way. If we look, for example, at how the concept Family is used pre and post late twentieth century, it seems to have the same function of ‘tracking the coordination of domestic life’. Yet, if we look at the rules and legal systems that inform domestic life pre and post late twentieth century, we might be inclined to deny that the concept has the same function during the former and the latter period.

This affects judgements as to continuity of inquiry. If the concept maintained its function, then it seems that inquiries about family pre and post late twentieth century are in continuity: a putative question about family posed in (say) 1952 could be felicitously answered by using the concept Family in 2019, and vice-versa. However, if the concept did not maintain its function, then there would seem to be a discontinuity between inquiries about family pre and post late twentieth century. It would then seem infelicitous (at least in some cases) to use Family with the function it had in 1952 in order to answer a question about family posed in 2019, and vice-versa.

In sum, our evidence – about language use, our laws, and our institutions – does not allow us to clearly identify the concept’s function in one way or another. If this is the case, then establishing sameness of function will become tricky. Since sameness of function would determine continuity of inquiry, establishing continuity of inquiry would become tricky as well.⁶

The function theorist may reply that, with my previous argument, I merely showed that mind-independent concept function is epistemically hard to come by. So what? We can still say that a concept has the function that it is most reasonable to assign to it after informed reflection about the aims and purposes that it serves for us. This option is close in spirit to accounts offered by Brigandt (2010) and Nado (2019), who both emphasize the relativity of a concept (and hence of a concept’s function) to aims and purposes. However, epistemic indeterminacy arises here as well. This is because informed reflection seems incapable to adjudicate whether Family has the same function in 1952 and 2019, for reasonable considerations support both the function-identity verdict and its opposite. The former verdict is supported by considerations regarding use, while the latter is supported by considerations regarding the cultural and legal environment that grounds such use. If this is so, then the same problem affects the individuation of ‘attributed’ function just as it affects the individuation of ‘real’ function.

The function theorist might still reply that, as long as sameness of function can be achieved relative to at least some aims and purposes, continuity of inquiry may be ensured. Yet, first of all, an explanation would be owed as to why certain aims and purposes matter more than others. Furthermore, even if such explanation were successful, if my argument is on the right track, it will still be basically always possible to contest function attributions, because it seems that a concept’s function can basically always be rendered both in a coarse-grained and in a fine-grained way. This means that the functionalist strategy can be potentially undermined in each of its particular applications. Such a systematic application problem may ultimately constitute grounds for discarding the strategy.

4. A different strategy: semantically conservative and progressive inquiries

We have reviewed two strategies to deal with the change of subject objection: appealing to topics and appealing to concept function. Both strategies are centred on some key element that is supposed to ensure continuity of subject and of inquiry, but both have troubles offering criteria of identification for it. Topic is unstable to the extent that samesaying verdicts can be contested; and function is unstable to the extent that judgments about its granularity can be contested too. Are there better ways of dealing with the objection?

The topic and function strategies are examples of what, in the first section, I identified as ‘reply option number (i)’ – namely, defending subject continuity. The topic strategy pursues this reply by directly offering a theory of subject that can ensure such continuity, while the function strategy strives to achieve subject-continuity via continuity of representational devices.⁷ In what follows, I will pursue what I identified as ‘reply option number (ii)’ – namely, arguing that subject discontinuity is not problematic. According to my proposal, one can admit that change of subject occurred in certain key cases; yet, one can also maintain that this need not affect inquiry continuity. How can such view be motivated?

In what follows, I suggest that one can distinguish between two ways of approaching an inquiry centred on some initial, first-order question, Q. First, I will draw attention on a ‘semantically conservative approach’, according to which semantic change is not allowed during one’s inquiry centred on Q. Second, I will draw attention on a ‘semantically progressive approach’, according to which semantic change is allowed. If the initial first-order question Q is approached in a semantically progressive way, one may say that it becomes embedded in a broader type of inquiry, which we could call ‘Semantically Progressive Inquiry’ (SPI). A SPI includes both first-order and meta-linguistic stages. By contrast, if Q is approached in a semantically conservative way, it is embedded in a type of inquiry that does not allow certain meta-linguistic issues to arise. We could call it ‘Semantically Conservative Inquiry’ (SCI).

In the remainder of this section, I will provide some examples of SCI and SPI. My aim is to convince the reader that these two categories are not mere stipulations, but rather describe something that already exists – although it is not yet labelled in terms of ‘SCI’ or ‘SPI’. Having argued that SCIs and SPIs do exist, I will contend that explicitly engaging in a SPI would benefit the inquirer targeted by the change of subject objection.

4.1. Semantically conservative inquiry.

To better illuminate SCI, imagine a child who, addressing an adult, asks a question with the goal of being educated about something: for instance, ‘Do fishes lay eggs?’ The adult’s response, in order to be appropriate, ought to provide the child with insight into the current, shared definition of what a fish is, as well as into the intension and extension of the word ‘fish’ as they are currently accepted in the ordinary language. No tinkering with the definition and the semantics of ‘fish’ is expected or allowed.⁸

Similarly, a scientist may ask the plain, first-order question: ‘How many exemplars of fish are there in the Lake Tohopekaliga?’ If all she aims for is gathering data about the ichthyic population of the lake according to the meaning of ‘fish’ currently accepted by the scientific community, and no further reflection as to what a fish is, or what ‘fish’ means, is expected or allowed, then her research is an example of SCI.

We thus have perfectly mundane cases of inquiries – centred on some first-order question, Q – which it is arguably appropriate to describe in terms of SCI, because they can consistently be reconstructed as ‘leaving no room for’ (not expecting, or not allowing) changes in the semantics of some key expression either contained in Q, or that is relevant to Q’s answer (e.g. ‘fish’).

4.2. Semantically progressive inquiry

To illustrate SPI, consider the following case, also reported in detail by Peter Ludlow (2014) and Mark Pinder (2020b). Up until the 1990s, the word ‘planet’ as used by astronomers was associated with a certain (more or less precise) definition. According to this definition, Pluto counted as a planet. Yet, in 1992, astronomers began to find out that Pluto was part of an asteroid belt, dubbed ‘Kuiper Belt’, which contained numerous objects with the same chemical composition as Pluto, some of which even larger than it (such as Eris, whose discovery was announced in 2005 (cf. Brown, Trujillo, and Rabinowitz 2005)). This would have meant calling these objects ‘planets’ too. Yet, such result was deemed undesirable, as Steven Soter explains:

If Pluto is included as a planet, we have no physical basis for excluding Eris, dozens of other large spherical [Kuiper Belt Objects], and Ceres. The term ‘planet’ would then lose any taxonomic utility. (Soter 2006, 2513)

This prompted the astronomers’ community to rethink the definition of ‘planet’. After a special Planet Definition Committee was appointed, the International Astronomical Union gathered in Prague in 2006, in order to vote a new and more precise definition of ‘planet’.

The experts’ decision⁹ arguably affected, through the division of semantic labour, the very meaning of the term ‘planet’ as scientists use it. In this scenario, we can imagine that the question ‘What is a planet?’ posed by the astronomers during their 2006 meeting was the start of a SPI. Why so? Because the astronomers did not wish to answer that question on the basis of the current definition (and term’s semantics) at the time. Rather, they wanted to (i) work out a new definition, which would in turn (ii) change the word’s meaning, which would in turn (iii) allow to formulate a better answer to the initial question, in light of the new discoveries.

Based on this example, we may characterize a SPI as an inquiry in which the subject (whether individual or collective) who asks the initial first-order question, Q: (a) assumes that some expression occurring in Q, or relevant for Q’s answer, has a suitably determinate meaning,¹⁰ (b) despite this, the same subject is open to revise such meaning, in order to achieve a certain purpose. Such purpose may relate to the agent’s pursuit of epistemic values (truth, accuracy, explanatory power), as well as non-epistemic values (practical convenience, applicability, but importantly also values related to moral, social and political progress, like socio-economic equality, democratization, emancipation, empowerment, sustainability). Whether the envisaged semantic change is to have an epistemic or non-epistemic import, the ensuing inquiry will likely consist of both first-order and meta-linguistic stages.

If the reconstruction of the planet case is correct, there exists at least one example of SPI that pre-dates my conjuring of the category ‘SPI’; the category is therefore not a mere stipulation. Still, some more details about this type of inquiry should be filled in.

First, what motivates an inquirer (or community thereof) to start a SPI? The inquirer’s (or inquirers’) dissatisfaction with a certain category or terminology, and their being open to definitional and semantic revisions to achieve certain epistemic or non-epistemic goals, are arguably central factors. In the case of Pluto, the category (and term) ‘planet’ came to appear unsatisfactory, and the astronomers’ community became expressly open to introducing definitional (and semantic) changes to better achieve certain epistemic and non-epistemic goals – such as representational accuracy or taxonomic expediency. Also questions like ‘What is a fish?’, ‘What is a gene?’, ‘What is a family?’ were perhaps, at some point in history, asked with a semantically progressive attitude, in light of the need to revise the attendant categories and terminologies in order to achieve certain epistemic or non-epistemic benefits.

I do not wish to suggest that these projects were ever overtly framed in terms of SPIs, or that someone engaged in a SPI need be immediately aware that they are. In many cases, scientists may not even think of themselves as ‘revising a term’s intension and extension’. Still, if one could reasonably reconstruct their inquiry as stemming from dissatisfaction with a certain definition, category or terminology, and as making certain semantic revisions permissible, one could legitimately portray them as pursuing a SPI. If this were so, any first-order inquirer (or community thereof) – be they natural scientists, social scientists, philosophers, etc. – may turn out as a semantic revisionist. All it takes is an adequate reconstruction of their inquiry in terms of SPI. Such reconstruction may be adequate if, for example, it met with sufficient consensus inside the very community conducting the first-order inquiry, or inside a community that purports to give a historical account of that inquiry. In turn, the consensus may be reached because the reconstruction itself exhibits some a priori virtues such as being charitable; or, because it is confirmed by more than one method of investigation; or, it finds support from more than one authoritative source.

4.3. The teleological unity of semantically progressive inquiry

Having engaged in a SPI, the inquirer, or the community of inquirers, may reasonably expect, at the end of their investigation, an answer to the initial question Q in which the semantics of a certain key expression e is different from the semantics of e at the time of Q’s initial occurrence. Moreover, since the inquirer, or the community thereof, was open to semantic revision from the very beginning, they may, or even should,¹¹ be ready to change the subject to some degree – especially in light of the fact that semantic revision is the main culprit for subject-change. How can they deal with inquiry discontinuity then? As I wish to propose, they can argue that, even if subject-change happens, the continuity of inquiry need not be broken.

The inquirer(s) can contend that the unity of inquiry afforded by a SPI is, at bottom, a teleological one. Suppose Erika’s inquiry starts on June 7th, 1994 with the question ‘Is free will compatible with determinism?’. If this question is part of a SPI, then semantic or subject changes will not in appropriate circumstances jeopardize the continuity of inquiry, because what keeps the various stages of Erika’s inquiry together is the common goal of answering the question posed at the beginning of her inquiry, on June 7th, 1994. In the course of a SPI, such goal is presumably pursued by both (i) engaging in object-level investigation and, (ii) when needed, by ascending at the metalinguistic level in order to operate semantic changes (although, as already noted, this stage may figure only in an inquiry’s reconstruction, not necessarily in the inquirer(s)’ self-conception). If all goes well, changes in intension and extension will not compromise such teleological unity.

Of course, this is not tantamount to saying that all changes in intension and extension will be allowed for the sake of teleological inquiry continuity. If this were to happen, the continuity of a SPI would become utterly trivial. It is entirely compatible with the present approach that some changes of intension or extension may go too far, and therefore cause a teleological discontinuity in a SPI, for instance in causing the inquirer to drop the initial question, Q, and to focus on a different question, R.

Yet, what makes a question the same or a different one? This will likely have to be established on a case-by-case basis. The following examples may, nevertheless, begin to shed light on what one’s guiding principles would be. First, in the case of Erika’s revision of the free will concept, one might argue that the initial question, Q (‘Is free will compatible with determinism?’) still stands, because the notions of compatibility and determinism have not been touched. Q still concerns whether ‘free will’ – whatever content this word will acquire – is compatible with determinism.¹² In other cases, however, Q might undergo changes that ultimately make it reasonable to discard it and replace it. Consider the question ‘Are whales fish?’. Revising the concept Whale implies revising the concept Fish. If such revision were effected while investigating the question ‘Are whales fish?’, it might cause one to lose sight of the very question at issue, for it may be unclear whether ‘fish’ is to retain its initial semantics or not. This unclarity might be a good reason for dropping Q and for starting a new inquiry based on a different question, R, phrased in more precise terms (for example: ‘Do whales breathe through gills?’). In yet other cases, it might be reasonable to drop Q entirely, without replacing it. Imagine that one’s initial question, Q, is: ‘Is phlogiston toxic?’, and imagine that, as a result of one’s inquiry, ‘phlogiston’ is ruled as extensionless. It seems like Q cannot be felicitously asked anymore and has to be dropped.

These are, admittedly, just examples. Indeed, one may feel that the identification conditions for questions are still too indeterminate. In light of this indeterminacy, the critic may point out that the present proposal fails to provide clear criteria for deciding whether a teleologically unified SPI is the same or not, in a very similar fashion to its competitors: the topic strategy, too, fails to provide criteria for inquiry-sameness by failing to provide criteria of topic-sameness, and mutatis mutandis for the function strategy. So why should one be drawn to the proposal sketched so far?

One significant advantage is that the notion of SPI, which includes a metalinguistic and a teleological aspect, provides an additional level of analysis for inquiries that allows one to capture the complexity and instability of judgments concerning sameness and difference of subject. This level of analysis is not accessible to the versions of the topic and function strategy presented in this paper and currently defended in the literature.

To illustrate the advantage, let us return to Erika’s inquiry on free will. The topic and function strategy would plausibly try to argue that the subject of Erika’s inquiry did not change because the inquiry’s topic, or the concept’s function, did not change. Yet, both topic-sameness and function-sameness can be easily questioned, as I have argued previously. What if there were an additional level of analysis where these sameness judgments could be accommodated, so as to make it possible to hold that something (let’s say, subject-matter) has changed, but also, something else has remained the same? The idea of a SPI can do this job: it allows one to state that, even though Erika’s subject matter changed, her ‘semantically progressive inquiry’ did not lose its continuity.

This additional level of analysis is not accessible to the topic and function strategies (as presented in Sections 2 and 3), which only focus on one key element (topic or function), and therefore lack the right degree of flexibility to accommodate cases in which sameness-judgments and difference-judgments arise within the same conversational context. Indeed, if these strategies subscribed to the idea of SPI, and of the metalinguistic and teleological unity of inquiry, they would plausibly have access to this advantage. To my mind, then, this makes the SPI-based strategy overall more attractive than its rivals: it offers tools that even its rivals can benefit from.

The conclusion of this section is that one can set up one’s inquiry as a SPI, in such a way that it is permissible to operate semantic revisions on relevant key terms. Such revisions may lead one to changing the subject, but still they won’t (in favourable circumstances) break the continuity of inquiry, because what crucially holds a SPI together is the goal of answering the initial question, and not semantic or subject continuity.

5. Change of subject and semantically conservative vs. progressive inquiries

The change of subject objection is problematic within a SCI, but need not be so within a SPI. I will devote this section to comparing the impact of the subject change objection on a SCI and SPI respectively, and explain how this can benefit conceptual engineering projects.

In the course of a SCI, if the answer to Q is not in line with the presupposed semantics of the key terms initially occurring in Q (be they known or not by the questioner), it is in some sense infelicitous. The infelicity may be explained as a lack of relevance: an answer in which the semantics of a certain key term differs from the semantics of the same term as it occurred in the question is strictly speaking irrelevant to the question itself. It fails to answer the question that was originally posed. It may answer another question that could have been posed, but in fact was not. This is perhaps the sense in which the conceptual engineer fears ‘providing answers to questions that weren’t being asked’ (Haslanger 2000, 34). The irrelevance is made problematic by the fact that a SCI makes no provisions regarding semantic changes along the way. Note that irrelevance is not a problem per se, but only in relation to the expectations built into the structure of an inquiry that is semantically conservative.

In a SPI, however, there need be no expectation that the semantics of the key terms will stay the same. In addition, there may, or even should¹³ be a certain degree of openness to change of subject. If such a change happens as the result of a revision of the intension and/or extension of some key term occurring in the question, or that is relevant for answering the question, this need not be a cause for concern, because the expectations of the inquiry as it unfolds under a semantically progressive approach would allow such change. Of course, as already mentioned, not any change will be tolerated for the sake of teleological inquiry continuity. The extent of the permitted change might be set by the inquirer at the beginning of the investigation, or even as the inquiry itself unfolds, and is heavily dependent on contextual factors. Still, SPI allows a degree of flexibility with respect to semantic and subject changes that is altogether absent in a SCI.

As a consequence, if one’s inquiry is structured as a SPI, the change of subject objection loses its bite, for it no longer points at an undesirable outcome in light of the initial expectations of the inquiry. It points at an outcome that is, or at any rate should reasonably be (under appropriate conditions and to appropriate degrees) allowed, precisely in light of the initial expectations of that kind of inquiry.

Given that projects in conceptual engineering aim at assessing and, if needed, improving semantic aspects of our concepts, there are good reasons to believe that, when a conceptual engineer starts by asking what seems to be an object-level question (e.g. ‘What is a woman?’, ‘Is free will compatible with determinism?’), they intend to pursue a SPI, or at least they should do so, as this would seem a reasonable way to frame and carry out their investigation. If this were so, conceptual engineers need not worry too much about the change of subject objection; for the alleged problem flagged by the objector would simply be (in the appropriate circumstances) a possible, expected outcome that is part of the otherwise teleologically unified enterprise.

The previous considerations lead to a response that is partly a concession and partly a downplaying of the change of subject objection. It is a concession, because it agrees that the objection points at something problematic for certain kinds of inquiries, namely SCIs. It is a downplaying, because it shows that the change of subject phenomenon is (in appropriate circumstances) factored into, and hence expected as a possible outcome of, other kinds of inquiries, namely SPIs.

6. How to sidestep the change of subject objection

The same considerations also suggest the following advice for inquirers who wish to deal with the change of subject objection. They should become sensitive to whether their inquiry is carried out as a SCI or a SPI and, if helpful, they should explicitly frame it as a SPI. This would help them (at least partially) ‘immunize’ their projects from the change of subject objection.

One could make it explicit that one’s inquiry is being carried out with a semantically progressive attitude by formulating one’s questions and answers accordingly. Suppose one were investigating the question ‘Is marriage necessarily between a man and a woman?’ The inquirer sensitive to the SCI vs. SPI difference should pause and ask: ‘Is it admissible for the purposes of this inquiry to change any semantic aspect of the word ‘marriage’ – thereby exposing this investigation to the possibility of changing the subject?’ If yes, any such changes may be tracked at different stages of the inquiry. First, they may be tracked in the way of formulating the initial question, ‘factoring in’ the possibility of such modifications. This would mean asking a more complex question like: ‘Are marriages (where the semantics of ‘marriage’ is open to revision) necessarily such-and-such?’ Second, they may be tracked in the subsequent answer, which may be articulated along the following lines: ‘Yes/no, marriages (modulo such-and-such semantic changes in the term ‘marriage’) are/are not necessarily such-and-such’.

The formulations just recommended also help one address the following worry, which may be called the question-answer mismatch worry.¹⁴ Suppose the initial question, Q, is individuated narrowly, that is, according to the semantic content its occurrence has at the beginning of inquiry. Suppose the answer, A, is also individuated narrowly, i.e. via the content it has acquired at the end of the inquiry, which included semantic revision. In this scenario, how can A really answer Q? Conversely, suppose Q is narrowly identified, but A is not narrowly identified, that is, the interpretation of A at the end of inquiry is left open. Why was Q narrowly identified in the first place? This worry can be allayed by introducing a metalinguistic element in both answer and question. Suppose I ask: ‘Are marriages (where the semantics of ‘marriage’ is open to revision) necessarily such-and-such?’. This question can be relevantly answered by uttering: ‘Yes/no, marriages (modulo such-and-such semantic changes in the term ‘marriage’) are/are not necessarily such-and-such’. If both question and answer refer to a metalinguistic stage in the inquiry, their connection can be more clearly represented, and the question-answer mismatch problem is less likely to arise.

The critical reader may at this point observe that the previous recommendations seem to suggest that framing one’s investigation as a SPI is the easiest thing to do: all it takes is deliberately signal that one is open to semantic and subject change. However, this deliberate choice may be contested. If the objector argued compellingly that the investigator’s inquiry should be carried out as a SCI, they could legitimately level the subject-change charge.

In response, it may be tempting to argue that very few first-order inquiries would be properly carried out as SCIs and that, in general, it is best to situate one’s first-order inquiry within a SPI, as it is simply more rational to be ready to question the quality of one’s concepts than to accept them uncritically (this thought is reflected in Cappelen’s (2018, 5–6) idea of ‘representational skepticism’ as opposed to ‘representational complacency’).

Regardless of such wide-ranging (and a bit over-intellectualistic) considerations about rational inquiry, though, one may simply point out that, at least in the cases that matter for conceptual engineering, inquirers usually have good reasons to suspect that certain key concepts should be revised. For instance, the grounds Haslanger has to suspect that the concepts of race and gender are covertly discriminatory – and hence call for revision – seem to be supported by history, sociology and psychology. The grounds Scharp has to regard the concept of truth as inconsistent – and hence as calling for replacement – find adequate substantiation in the liar paradox. Similar considerations seem to apply to other projects that involve the concepts of knowledge, belief, moral goodness, etc. The burden of proof would be on the objector to argue that nothing speaks in favour of revising (or at least trying to revise) such concepts.

The objector could also note that the present account allows communication breakdowns to obtain after an inquiry has led to conceptual revision. Yet, these breakdowns would not matter to the present view, as long as the teleological unity of inquiry could be maintained. That seems like blatantly allowing subject change!

While it is true that misunderstandings and merely verbal disputes could sometimes happen post-concept revision, the present proposal also has the resources to show that semantic changes can be seamlessly incorporated into a teleologically unified SPI. To achieve this, the actors involved in the SPI should become more sophisticated in asking and answering questions. This means, for example, asking questions that explicitly signal the possibility of meaning change, like: ‘What are Fs (where the semantics of “F” is open to revision)?’; or answering questions by explicitly mentioning any semantic changes effected, like: ‘Fs (modulo such-and-such change in the semantics of “F”) are Gs’. These more complex questions and answers make it apparent that one’s inquiry has both object-level stages and meta-level stages, which can be seen as continuous in virtue of their aiming towards the same goal – namely, answering the initial question. Thus, a more refined sensitivity towards communication within an inquiry can reconcile unity and semantic change.

Note that these would be good measures to take even if there were no such thing as conceptual engineering. Languages evolve, and terms change meaning across contexts and times, whether anybody deliberately brings about these changes or not. So, paying attention to potential semantic shifts, and injecting some ‘metalinguistic awareness’ in our speech practices, are good policies no matter what.

A further objection comes from the externalist camp: ‘If externalism is true, there can be no such thing as a semantically progressive inquiry’. For, if they are fixed externalistically, neither the intension nor the extension of a concept can be changed by a semantically progressive inquirer. A way around this problem consists in arguing that, if externalism is true, practitioners of SPI (and a fortiori, conceptual engineers) need only change or improve those aspects of a concept that can actually be controlled (cf. Riggs 2019).¹⁵ For instance, even if externalism is true, one may engage in the project of manipulating speaker-meaning, as opposed to semantic meaning, which is supposedly independent of the beliefs and intentions of language users (Pinder 2019). Alternatively, one might wish to intervene in the very external environment fixing the relevant semantic facts, as suggested by Simion and Kelp (2020). If this is so, then semantically progressive inquiries can be safely carried out; one just needs to accept that they will change whatever aspects of conceptual representations can be under the inquirer’s control, whether or not externalism is true.

7. Conclusion

I have argued that theorists targeted by the change of subject objection may downplay such objection, by framing their inquiries as Semantically Progressive Inquiries (SPIs). In a SPI, the terms – and relatedly, the concepts – relevant to the investigation are allowed to undergo semantic change. This may, for all the inquirer knows, lead to a change of subject. In a SPI, though, subject-change need not imply inquiry discontinuity, since inquiry continuity can be thought of as teleological and metalinguistic, i.e. as tied to the overarching goal of answering one’s initial question, Q, by going through both first-order and meta-level steps (note that such continuity may result from even just a reconstruction of the inquiry). In this picture, even if a change of subject obtains, SPI need not be disrupted. As a result, the change of subject objection loses its bite. Finally, potential communication breakdowns can be prevented by increasing the inquirers’ metalinguistic awareness.

Acknowledgments

Versions of this paper were presented at the Czech Academy of Sciences, at the University of Zurich, and at the University of Vienna. I am grateful to the philosophers sitting in those audiences for their helpful and constructive comments. Thanks to Manuel Gustavo Isaac and Steffen Koch for their feedback during the revision process of this article.

Disclosure statement

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

Notes

1 Peter Strawson’s (1963) objection to Carnap’s method of explication (Carnap 1950) is usually associated to it, but in this paper I wish to set Strawson’s objection aside. For a careful study of Strawson’s arguments, see Pinder (2020a).

2 See United Nations (1984, 85) and U.S. Department of Justice (2002, 340A), as referenced by Plunkett and Sundell (2013, 19).

3 For a similar point, see also Schroeter and Schroeter (2020).

4 The debate on what is said is divided between minimalist views (Cappelen and Lepore 2005; Borg 2004) and pragmatic-enrichment views (Bach 1994; Recanati 2001, 2004; Carston 2002; Sperber and Wilson 1986). According to minimalists, what is said is exclusively determined by semantic factors (which may include the assignment of reference to indexicals and demonstratives); according to the pragmatic enrichment theorists, what is said is supplemented with contextual elements whose retrieval is not semantically triggered, but depends on pragmatic processes. For a recent alternative, see Schoubye and Stokke (2016).

5 Note that I am not canvassing, or assuming anything about, the ontology of concepts. I am describing the assumptions of potential, other authors.

6 This granularity problem is presumably one of the reasons why function theorists tend to use qualifications like ‘proper function’ (Thomasson 2020, 444), ‘central function’ (Haslanger 2000, 35), ‘essential feature’ of a concept’s function (Prinzing 2017, 8). Do these qualifications escape the granularity problem? There are reasons for thinking they do not. In the family case, one may ask: are hetero-normativity or the role of marriage essential, or proper, or central to the function of the concept Family? Our current practices do not allow to give a determinate answer to this question. Indeed, whether these aspects are essential, proper, or central seems to be exactly what fuels many heated controversies between subjects with different religious and ideological preferences in the contemporary public sphere.

7 There could be other strategies that pursue these broadly construed approaches. One may complain that I am not ruling out such alternative strategies. While this is true, it would, however, be up to the supporters of the broadly construed approaches to present these putative alternative strategies and show their viability. As to my proposal, since it will ultimately rest on the idea of teleological continuity, I take it to depart from both the subject-continuity and the representational-device continuity approaches.

8 Neither the child nor the adult need be aware of any of the above. However, there is a reconstruction of the child’s and adult’s exchange whereby it would be rational for them to assume that ‘fish’ has a certain definition as well as a determinate intension and extension, and that any answer to the initial question ought to give the asker access to them.

9 The decision is stated in resolution B5 of the International Astronomical Union (https://www.iau.org/static/resolutions/Resolution_GA26-5-6.pdf).

10 The inquirer could have a different assumption as well: that the key term they are ready to revise does not have a determinate semantics. This is stressed by Haslanger, who writes: ‘[O]rdinary concepts are notoriously vague’; as a consequence, ‘there is room to stretch, shrink, or refigure what exactly we are talking about’ (Haslanger 2000, 34). This is echoed by Dutilh Novaes: ‘we users of ordinary languages often do not know exactly what it is that we are talking about when using everyday life vocabulary. At the very least, there is considerable leeway and flexibility in usage’ (2020, 1030). (Recall, also, that a famous example of explication by Carnap (1950, 11–15) starts from the quite imprecise ‘cold’ and ‘warm’, and shows the process of their explication in terms of the more precise, quantitative concept ‘temperature’). If the inquirer assumes the indeterminacy of the concept under scrutiny, their response against the change of subject objection may also be that, given the vagueness of many of our concepts, there seems to be no subject to change. Although my proposal allows this move, it provides a response to the change of subject objection that can also build on the assumption that there is a determinate semantics and hence there is a ‘subject’, and that it is OK to change it provided our inquiry has the right structure. I consider this a plus that the indeterminacy-based response lacks.

11 It would be tactically unwise to be open to semantic change, but not to subject change, for this would certainly make one’s position vulnerable to the subject-change objection.

12 Thanks to Naomi Osorio-Kupferblum for helpful discussion on these issues.

13 Of course, the conceptual engineer is free to effect semantic changes while not being open to subject change to any degree. As noted in fn. 11, however, this may be tactically unwise.

14 I owe this objection to an anonymous referee.

15 Alternatively, one may insist that it is possible to change our concepts, as a community and in the long run, even if externalism is true (see Koch 2021).