Hypothetical Imperatives and Normativity
If you want normativity on the cheap...
The notion of (moral or other) norms having “categorical” authority over us—whether we endorse them or not—can seem mysterious. This motivates some people to look to hypothetical imperatives (like, “if you’re thirsty, you should drink a glass of water”) as a model for securing “normativity on the cheap”. The hope is to reduce normativity to the simple and undeniable phenomenon of means-end relations. In this post, I’ll explain why this hope is misguided.
Hypothetical vs Desire-dependent Categorical Imperatives
There are two very different ways that normative advice can depend on desires. To see this, compare:
(1A) If you want to torture children, you should seek psychiatric help.
(1B) If you want to torture children, you should volunteer as a babysitter.
Example 1A, despite its conditional form, invokes categorically authoritative normativity: it specifies what you (really) ought to do if you find yourself in the antecedent circumstances. Note that the implicit ends guiding the advice (protecting children and restoring your sanity) do not depend upon your endorsement. They are—presumably—good ends regardless, and the purpose of the conditional is to indicate how those good ends can be wisely promoted in the specified circumstances.
1B is very different. It is a genuine hypothetical imperative: it expresses a means-end relation by, in effect, presupposing that the goal described in the antecedent has normative legitimacy, and describing what consequent advice would follow from this assumption. But it isn’t really endorsing this advice: it is purely hypothetical. Even if the antecedent condition is actually realized, you can’t “detach” the consequent and turn it into categorical advice. You are always free to reject the normative presuppositions embedded in a hypothetical imperative.1
The Irrigation Model
The key thing to understand about hypothetical imperatives, thus understood, is that they describe relations of normative inheritance. “If you want X, you should do Y,” conveys that given that X is worth pursuing, Y will be too.2 But is X worth pursuing? That crucial question is left unanswered. A view on which there are only hypothetical imperatives3 is thus a form of normative nihilism—no more productive than an irrigation system without any liquid to flow through it.
To get any real normativity out of hypothetical imperatives, you need to add some substantive claims about desirability, or what ends are categorically worth pursuing.
Until then, we’ve just got a huge array of possible normative systems channeled out. We can “hypothetically” predict and compare racist normative advice, anti-racist normative advice, and so on for infinitely many other possible ends. But we need something more to break the symmetry between them and yield actual reasons to do one thing rather than another. To get any concrete advice, we need to fill out just one of the possible channels as the one to follow.
Against Stipulated Answers
At this point, one might be tempted to brush away the problem by simply stipulating an answer. “We use ‘morality’ to talk about distinctively prosocial/co-operative ends,” some suggest. This is empirically false as a universal claim: Nazis and other fanatics commonly ascribe ‘morality’ to acts that align with their anti-social ideologies. Now, I happen to think they’re substantively mistaken, but if you’re just playing word games, anyone is free to stipulate whatever arbitrary meanings they please.
So it’s worth reflecting on what we can say about such fanatics. I take their ‘moral’ talk to aim at capturing what is truly worth doing, but then the acts they call ‘moral’ fail to meet this standard. So on my view, fanatics get things wrong: they are substantively mistaken in believing that their fanatical acts are worth doing.
The stipulator misses all this. The most they can say is that the fanatic’s behavior doesn’t conform to the stipulator’s preferred sense of ‘morality’. But the situation is perfectly symmetric: the Nazi can equally say that the rest of us fail to conform to their preferred sense of ‘morality’. There’s no real disagreement here (the fanatic can perfectly well agree that they aren’t being very co-operative, so we don’t disagree about that), just people with different goals/values, and different stipulated jargon, talking past each other.
Now, if your understanding of normativity doesn’t allow for genuine disagreement between Nazis and altruists, you’ve gone badly wrong. But this disagreement criterion requires commonality of meaning between people with different first-order views about what ought to be done. Robust moral realists accommodate this by taking “what ought to be done” to refer to an independent normative property, the distribution of which people may dispute (without changing topic), just as flat-earthers can dispute the shape of the world without meaning something different by “the world”. Stipulationism—or other reductionist accounts of moral metasemantics—cannot accommodate shared meaning in this way. If you take the meaning of ‘morality’ to be provided by first-order views, different first-order views immediately entail different meanings. So that’s no good.
Quite apart from the semantic issues, the deeper problem in this vicinity is that it makes fundamental normative inquiry impossible. (This is why Parfit thought that his life’s work would be meaningless if metaethical naturalism were true.) G.E. Moore famously noted that even the most fundamental questions in ethics, like ‘Ought we to promote overall well-being?’, are substantive, open questions. You can’t just stipulate an answer. You have to actually engage in critical moral inquiry in order to work out what answer is most justifiable, considering reasons for and against different possible answers.
We should hope to find that our answer has more going for it than the Nazi’s. But that’s not something mere stipulation can achieve. There has to be a real question, with contestable answers.
Subjectivism: will any desire do?
Subjectivists about reasons claim that any actual desire suffices to create categorical reasons. As soon as you adopt an end, pursuing it becomes truly worthwhile, for you.
This raises questions about how to interpret the “for you” clause. In pop-relativist terms, “for you” means “according to your beliefs”. (For example, “true for you” means “you think it is true”.) But to simply note that you believe your ends to be worth pursuing does not yet settle that they are worth pursuing. Beliefs can be mistaken! So the more interesting view in this vicinity is the one on which adopting an end makes it really (categorically) important that you pursue it. On this understanding, the “for you” is just indicating that the categorical normativity is here directing you, rather than anyone else.
The problem with the more interesting view is that it’s implausible. Desires just aren’t that normatively powerful. You can’t turn a terrible idea—like torturing children—into a good one just by desiring it. To think otherwise is to endorse a weird kind of wishful thinking. Maybe desires could act as normative amplifiers or tie-breakers: between all the candidate good options too choose from, desiring one in particular could give you more reason to choose that one. Maybe. But what possible reason could you have for thinking that a mere desire suffices for turning rotten options into worthy ones?
The standard answer is: metaphysical scruples. People worry that categorical reasons are “spooky”, whereas desire-based normativity seems more naturalistically grounded. But I actually think this rests on a conceptual confusion.
Distinguish two kinds of grounding: metaphysical and normative. To say that desires metaphysically ground reasons is to say that reason-facts reduce to facts about desires (and associated instrumental relations). To say that desires normatively ground reasons is to say that what we have reason to do, as a matter of fact, is to fulfill our desires (whatever they may be): desires do the normative work of generating reasons.
The metaphysical reduction puts us back in the vicinity of “stipulated answers”, and struggles to make sense of the normative beliefs of non-subjectivists.4 But also, all the metaphysical benefits here are coming from the reduction of the normative to the natural, not specifically from the subjectivist’s claim that the relevant natural reduction basis is the agent’s desires. Compare reductive hedonistic utilitarianism: the property of maximizing overall happiness is no “spookier”, or less natural, than the property of satisfying the agent’s desires. So why not go with that instead? I don’t see why naturalist scruples should make subjectivism seem especially attractive (compared to other reductionist views).
As for normative grounding, that has nothing to do with your metaphysics. It just comes down to the question of what most plausibly makes an action worth doing. The subjectivist says that desiring something makes it worth doing. Example like desiring to torture children cast severe doubt on this.5 I think a much better answer is that promoting well-being makes something worth doing, though the best answer of all may turn out to be much more pluralistic than this (since, for example, exploring interesting questions is also worth doing, but seems distinct from promoting well-being).
Whatever the right answer turns out to be, I think we’re more likely to find it by doing ethics—reflecting carefully and open-mindedly about what most plausibly matters, and why—than by doing metaethics (thinking about the semantics and metaphysics of normativity)—especially if one’s thoughts here are artificially constrained by “metaphysical scruples” (that is, fear of non-naturalism).
Conclusion
The plausibility of desire-based accounts of normativity may be artificially inflated by the fact that most people have reasonable desires. This creates a striking correlation between desires and reasons. But the correlation remains imperfect, and can be broken entirely by considering more extreme cases. At the end of the day, there’s nothing normatively magical about arbitrary or unjustified desires: you shouldn’t expect to get reasons (for action) out of a state (of desire) that you have no reason to be in.
The purest version of the “hypothetical imperatives” view assigns no special authority to our actual desires. It doesn’t say that we categorically ought to do what will fulfill them.6 Rather, it equally notes all the possible ends, and what would normatively follow from each—were it to be reason-giving. This shows us lots of normative inheritance relations. But as long as it’s all hypothetical, you don’t actually get any normativity out of it. The “beneficence-relative reasons” are just as lacking in normative force as the “Nazi-relative” ones. You could make-believe that there’s normativity running through your preferred channel, and talk as though that were true. But it’s just talk, and the real underlying view here is normative nihilism: the view that there are no normative reasons.
You get what you pay for, alas, and “normativity on the cheap” turns out to be no normativity at all.
One could imagine running a Tortoise-and-Achilles-style dialogue expanding upon this point.
- “I find that I want to torture children, yes, but I don’t endorse it…”
-- “OK, so revise the conditional to begin: ‘If you want to torture children and you endorse it, then…’”
- “But even if I both wanted to torture child and endorsed it, I might not meta-endorse it!” (etc.)
The funny thing is that in this case, unlike in the original dialogue, the Tortoise is quite right to refuse to draw the suggested inference. There’s an unbridgeable gulf between the purely descriptive antecedent and the seemingly normative consequent. To fix this, the antecedent should really be read as implicitly normative: “If you rightly want X, then…”
And this often suffices for everyday purposes: putting aside weird cases involving torture and the like, most people’s actual ends are reasonable ones, worth pursuing. For example, if someone wants to get to the train station on time, you typically have reason to help them attain that goal.
For example, the early Philippa Foot paper, ‘Morality as a system of hypothetical imperatives’.
There are subtle issues here about the relation between epistemic gaps and metaphysical ones, but I think many philosophers are far too quick to assume that there’s little connection between the two. For a corrective, check out Hambly’s excellent Normative Property Dualism Argument, or my own 2-D Argument Against Metaethical Naturalism.
See also Parfit’s objections to subjectivism.
That would be no less ‘spooky’ than any other kind of categorical normativity, as Matt Bedke rightly stresses in his brilliant paper, ‘Might All Normativity Be Queer?’ (See his analogy to ‘subjectively magical’ witches, who can cast spells only on themselves.)
It’s helpful for me when you make clear the views of yours I disagree with!
My strategy is to take the subjectivist view, that any actual desire creates reasons, at least agent-relative ones, and that’s all the reasons that there are.
However, I then note that the boundaries of agents aren’t clear - the desires of a person are in some sense constituted by their (often somewhat conflicting) sub-personal desires; the desires of a group are in some sense constituted by their (often somewhat conflicting) individual desires, etc. I then take morality to be about hypothetical normativity relative to the desires of the biggest group, of which we are all a part.
Everyone’s desires get counted as part of this, even the person who desires to torture children. But the children’s desires count too, and will surely overwhelm those of the would-be torturer.
All argument is hypothetical; if the premises are true, and the axioms of logic are true, and the argument includes no logical errors, then the conclusion is true. So the categorical argument is also hypothetical, it just begins with different assumptions. I guess the difference is, one is allowed to deny the premises of a hypothetical at will, but supposedly the premises of a categorical argument are undeniable, or at least independent of the opinions of agents.
This presupposes that we can identify a perfectly impartial point of view from which to evaluate things, and that we should/must do so. Maybe we can let in a bit of wiggle room, and say we only are obligated to do the best we can to identify this view from nowhere. But why would we want to adopt it? Do we need some meta-argument to make that conclusion? And meta-norms that give the reasons?
Stipulate that such a view from nowhere exists, and we are motivated to find and apply it. How would the choices we face differ from those of persons who disbelieve all that, but are committed to interacting socially and cooperatively for mostly selfish reasons (including within “selfishness” a regard for at least some others)?
Whatever standard gets adopted, someone will violate it. Arguments about what label to apply to that (wrong, bad, mistaken, evil) seem to go only half-way to the bottom line, which is “what happens as a consequence of the violation?”