Words Don't Matter

Ethics, schmethics, and semantics: Against philosophical language policing

Some philosophers seem strangely fixated on ordinary language. Either that, or they assume that English words perfectly correspond to natural joints in reality.¹ They might, for example, think it a substantive metaphysical question whether tacos qualify as a kind of sandwich (as though our word ‘sandwich’ marked a fundamental distinction in reality—a real property that tacos might either possess or fail to possess, independently of our linguistic dispositions). Worse, some take normative ethics to be illuminated by close examination of the semantics of the English word ‘ought’. This strikes me as deeply wrongheaded.

Just imagine a linguist trying to intervene in a heated debated between utilitarians and Kantians: “Guys, stop! You don’t need to argue any more: I just did a thorough empirical survey and it turns out that the English term ‘wrong’ picks out both pushing the guy off the trolley footbridge and telling the truth to the murderer at the door. Common sense ethics wins!”

Obviously this is daft. We’re not really interested in the English word ‘wrong’. Imagine having a strange condition which causes you to confidently confuse different words, such that you’ve mistakenly muddled up the English meanings of ‘cow’ and ‘wrong’ all this time without realizing it. This wouldn’t mean that your normative questions are “really” questions about cows; all it means is that you will struggle to properly express yourself to other English speakers. We use words for communication, but at the end of the day, it’s our thoughts that matter, philosophically, not our words. And neither other language-speakers nor the linguists that study them have the slightest authority over the contents of your thoughts.²

Empirical data about people’s linguistic dispositions has no bearing on normative questions, or indeed anything else of broader philosophical interest (besides whatever interest you might have in how language itself works). If it turned out that I was a brain-in-a-vat and there was no English language—beyond the private language that I speak myself and hallucinate on the part of others—I cannot see how that would (or should) change any of my (fundamental) philosophical views in the slightest.

This could be you! Would it change the fundamental philosophical facts?

So words don’t matter. If needed, we can always just introduce new words—start speaking philosophical English, or “schmenglish”, if ordinary English turns out to be ill-suited for expressing our philosophical thoughts and ideas.

I’m belaboring this point because so many other philosophers seem to treat semantic interventions as a respectable philosophical methodology, and I don’t understand why. In my philosophical utopia, semantic interventions would simply prompt the immediate reply, “Words don’t matter!”,³ and save everyone a lot of time, with absolutely no loss (as far as I can tell) of substantive philosophical insight.

Alas, our actual world is no utopia, so philosophers sometimes take seriously semantic interventions that illicitly narrow our philosophical thinking. Examples follow.

Predicative and Attributive Goodness

Consequentialism is standardly formulated as claiming that we should promote the good. Some try to block the very expression of this view by claiming, with Geach and Thomson, that the word ‘good’ only has an “attributive”, role-relative sense (where, like “a good knife” or “a good thief”, a “good X” is good as an X), and not a “predicative” sense that allows separating “good outcome” into the two parts being an outcome and being just plain good.

I’ve always been baffled by this move. Even if we grant the bizarrely restrictive linguistic claim, does anyone seriously believe that there are no consequentialist thoughts to be expressed in moral philosophy? Obviously, we can just use different words if we have to. For example, instead of talking about “good outcomes”, we can talk about ranking outcomes in order of their preferability (or their impartial preferability, or perhaps their pre-moral preferability—prior to taking into account any special deontic reasons we might have to disprefer outcomes in which we violate our moral duties).

Words don’t matter. You can’t rule out a normative view via language policing. So appealing to the thesis that ‘good’ isn’t predicative as an attempt to preclude the very expression of the consequentialist idea is just silly. Nobody should take this sort of move seriously for a moment. If you want to argue that there are no reasons for preferring some outcomes over others, feel free, but you need to argue for that view on substantive philosophical grounds.

Satisficing and Absolute Gradable Adjectives

I learned at a recent conference that philosophers have recently argued against satisficing accounts of rationality on the semantic grounds that ‘rational’ functions as an absolute gradable adjective, such that anything short of perfect rationality qualifies by definition as ‘irrational’. (Here’s a recent example in a leading journal.)

I thought the conference presenter offered a compelling case for disputing that restrictive linguistic claim, before briefly mentioning what I think is really the more important point, namely, that we could always choose to use words differently in any case. We cannot simply assume that the English word ‘rational’ expresses a philosophically interesting concept. Instead, I think we should think substantively about rationality, determine whether we think rational criticism aptly applies to everything that is less than perfectly rational, and then adopt terminology that allows us to most easily express the philosophically important truths in this area.

Again, I’m just not interested in anyone’s linguistic dispositions when I’m grappling with a philosophical problem. So I disagree with Siscoe, for example, when he writes:

Why think that a theory of practical rationality needs to answer to our ordinary uses of ‘rational’? How we apply the concept of rationality is one constraint on philosophical theorizing. If a theory can explain the ways that we speak about rationality, then this provides prima facie evidence in favour of that account. Granted, prima facie evidence is always defeasible: error theories can be given to explain why natural language systematically leads us astray. It is this precise issue, though, that makes our inquiry relevant. Determining which of maximizing and satisficing more readily explains the way that we speak, and which is saddled with providing an error theory, is thus of philosophical import. If it’s true that Sorensen’s views on rationality can make better sense of the semantics of ‘rational’, then acknowledging this fact can help to make progress on the strongest account of practical rationality.

I don’t believe that “the way that we speak” has even prima facie bearing on normative questions. It rather strikes me as strictly irrelevant to the questions of what we should care about, what we ought to do, and when we’re apt to be criticized. The correct answers to these normative questions simply do not depend on anyone’s linguistic dispositions. (Again, I could not express the questions, or my answers to them, without language. But that’s very different from claiming that which philosophical proposition is true depends on contingencies of the English language. And others’ dispositions to assent or dissent to various premises will naturally influence the dialectic, by affecting how relevant or persuasive others find various arguments. But that will naturally emerge in the give-and-take of argumentation; we don’t need—or want—surveys or linguistic data to pre-empt this process of philosophical dialectic.)

You might take a more moderate view based on viewing peer judgments as a kind of higher-order evidence. Maybe you (unlike me) give some weight to how diverse others apply a normative term as potentially overriding your own normative judgments. But this isn’t properly philosophical evidence, IMO. Compare: crude opinion surveys would not qualify as helpful interventions in a first-order philosophical debate. (“Surveys suggest that commonsense morality is correct!” is not a paper we’d expect to see in a top philosophy journal, even if you think that others’ philosophical opinions constitute prima facie evidence.) And linguistic data is on a par with opinion surveys. The first-order debate can best proceed without such distractions.

Extending ‘Ought’ to Non-Maximal Options

Some miners are stuck in a mineshaft, but you don’t know which one of two it is. You can block either shaft to protect it from the rising water level, at the cost of flooding the other. If you block the correct one, you’ll save everyone. If you block the wrong one, all will drown. If you do neither, a few will die but most will eventually be saved.

Given these stipulations, we can all agree that you objectively ought to block the correct mineshaft, and objectively ought not to block the wrong one. But consider the more generic question, “Should I block a mineshaft?” Rationally—given your evidence—you should block neither, of course. But objectively ought you to block a mineshaft?

I think this is a non-question. We can disambiguate two more specific questions one could be asking, and beyond that I don’t see any further substantive issue to dispute. One could ask:

(1) Would doing as I objectively ought to do entail blocking a mineshaft? (Answer: Yes, blocking the correct mineshaft obviously entails blocking a mineshaft.)

(2) Would it be objectively advisable for me to block a mineshaft—given how I would actually end up blocking a mineshaft, would an omniscient advisor want me to do that? (Answer: it depends which mineshaft you would actually block. If you would actually block the wrong shaft, then the answer is clearly ‘No’, you objectively ought not to do what you would actually do were you to block a mineshaft.)

These are two clear questions with very clear answers. When philosophers debate how the objective ‘ought’ applies to non-maximal options, they are (it seems to me) engaging in the terminological debate of which of these two candidate meanings we should give to the ambiguous question. I don’t think it matters much, as long as we’re clear about what we mean; but if I had to pick, I’d opt for the second as seeming more normatively relevant and action-guiding.

In an old blog post, I lamented that an otherwise-interesting paper seemed to rule out the advisability interpretation of ‘ought’ by linguistic fiat, in claiming that ‘ought’ is upward-monotonic in the following sense: “if you ought to do a certain act X, and X-ing entails Y-ing, then you ought to do Y.” (Obviously the advisability of X-ing does not entail the advisability of the more generic act Y, if the agent would actually Y in a disastrous way instead.)

In an interesting exchange in the comments, Jack (one of the paper’s co-authors, and an old friend) clarified—very reasonably!—that he didn’t mean their linguistic arguments to rule out any normative views:

Upon reading this paper, one could be convinced that ethicists should not be concerned with what people (objectively) ought to do. Indeed, one could be convinced that ethicists instead should be concerned with what people a-ought to do.

So that’s all well and good. Except, I think it’s pretty clear that philosophical terms aim at philosophical relevance and perspicuity. If the concept of objective advisability (what Jack calls ‘a-ought’) is in fact more relevant to ethics than the upward-monotonic English ‘ought’, then ‘ought’ in the mouth of a moral philosopher expresses a-ought, not English-ought. Why would we care about English words at all, when they fail to track what philosophically matters?

So I say: Let ‘ought’ default to expressing whatever is the most philosophically interesting and respectable concept in the vicinity (unconstrained by shallow linguistic concerns), and add artificial markers (‘E-ought’) when needed to differentiate non-philosophical usage as tracked by English-language linguists and semanticists. We can then reframe Jack’s upshot as claiming that in our concern for what people ought to do, ethicists should not be concerned with what people E-ought to do. But we already knew that, based on the more general (and a priori) dictum that words don’t matter.

Conclusion

This risks coming off as an anti-semantics manifesto, so I should stress that I’m really only targeting semantic overreach. There may be very interesting questions in philosophical semantics (I take no view on that). But they don’t settle the other questions of philosophy. To suggest otherwise constitutes an objectionable form of philosophical language-policing. If my philosophical idiolect diverges from ordinary English, so much the worse for ordinary English!⁴

Words don’t matter; ideas do. Philosophical questions and concepts should be assessed for their intrinsic interest, not for how or whether they correspond to natural language terms. We should generally want to expand our expressive powers, so as to be able to consider a wide range of candidate views. Semantic explorations may be helpful, i.e. drawing attention to otherwise neglected possibilities or new candidate views to consider.⁵ But attempting to rule out a rival position as inexpressible strikes me as the absolute cheapest of cheap shots, not to be tolerated as a serious philosophical methodology.

Once all the candidate views are on the table—and all our thoughts rendered expressible as best we can manage it—we can turn to the critical project of substantively evaluating the rival views on their merits. I can’t think of a single case where I’ve found semantic “evidence” concerning natural language to be helpful or relevant to this critical stage of philosophical evaluation. Can you? (And would you radically revise your broader philosophical beliefs upon learning that you’re a brain in a vat and there’s really no such thing as the English language?)

1. Traditional debates over the analysis of knowledge, for example, often left unexplained why this was assumed to be a concept of fundamental philosophical interest, over and above those of truth and rational credences. (If anyone knows the answer—or even just has a justified true belief on the matter—I’d love to hear it. I’m guessing it’s linked to taking knowledge to be the correct “norm of assertion”, or something along those lines. Which could well be a good basis; I just wish it had been made more explicit when I was first introduced to the topic.)

2. Except insofar as you choose to cede such authority via metalinguistic deference: I might think about “whatever it is that others in my linguistic community generally mean by ‘arthritis’,” for example; but nothing stops me from instead simply thinking in my personal idiolect, if that’s what I prefer.

3. Or perhaps a sarcastic, “Oh no, I’ve been refuted by the language police!” But that’s maybe a bit too harsh.

4. Assuming my concepts better illuminate the underlying philosophical substance, that is. I certainly don’t mean that they’re special just because they’re mine. If you think I’m getting at uninteresting philosophical questions, you should of course ask different ones instead! But I do want to insist that what matters is the intrinsic interest of the ideas, not whether the concepts involved correspond to natural language terms.

5. Examples may be various forms of contextualism, assessor-relativism, etc.