Sunday, April 05, 2009

Guest Post: 'What is Constructivism?'

[The following is a guest post by Jeff Sebo -- now blogging at 'The Dear Self'.]

I understand metaethical constructivism* as the family of theories according to which

1. Our moral discourse is cognitive.
2. Some moral statements are true.
3. Moral statements are true or false in virtue of evaluative attitudes.

If this is right, then here's how constructivism relates to the other big "metaethical families":

1. Expressivism. Constructivism and expressivism agree that evaluative attitudes are what ground morality. But unlike expressivism, constructivism says that our moral discourse is cognitive. So whereas expressivism says that our moral discourse directly expresses evaluative attitudes, constructivism says that our moral discourse describes evaluative attitudes.

Relatedly, constructivism differs from expressivism in that, for expressivists, I always use moral concepts to express my evaluative attitudes, whereas for constructivists, I may also use moral concepts to describe others' evaluative attitudes. (More on this below.)

2. Error Theory. Constructivism and error theory agree that our moral discourse is cognitive. But unlike error theory, constructivism says that some moral statements are true.

3. Realism. Finally, constructivism and realism agree that our moral discourse is cognitive, and that some moral statements are true. But they disagree about the nature of moral truth: whereas realism says that moral statements are true or false in virtue of mind-independent moral facts, constructivism says that moral statements are true or false in virtue of mind-dependent moral facts, namely facts about evaluative attitudes.

Technically speaking, then, constructivism is a kind of antirealism. But since constructivism agrees with realism that moral facts are real (albeit mind-dependent), some constructivists prefer to avoid the potentially misleading label 'antirealism'. Korsgaard, for example, calls her view 'procedural realist' instead of 'antirealist', and she calls her opponents' view 'substantive realist' instead of 'realist.'

I should note that this is a very expansive understanding of constructivism; it makes constructivism a much larger family of theories than we might have thought. For example, it means that constructivists can disagree about:

1. Which evaluative attitudes are relevant to morality?

Constructivists have a lot of options here. They can say that the relevant attitudes are moral intuitions, moral beliefs, desires, preferences, aims, plans, choices, and much more. Some pick even more exotic options than these. For example, Rawls picks considered moral judgments. Korsgaard picks reflective endorsement. Street picks takings-to-be-a-reason. Velleman picks judgments-about-what-it-makes-sense-for-one-to-do-in-causal/psychological-terms. And so on.

2. Does morality depend on actual evaluative attitudes or ideal (i.e., fully informed) evaluative attitudes? And if the latter, what does it mean for one to be fully informed?

Most constructivists opt for the latter option here. Rawls says that the moral facts are determined by our considered moral judgments in reflective equilibrium. And in his political theory he idealizes by making people rational and self-interested and ignorant about who they are. Korsgaard says that morality is normative iff we would still see it as justified if we understood it completely. And she argues that humanity has value not because we actually value it, but rather because we would value it if we learned what our present values commit us to. Street argues that we have reason to do something iff it figures in our maximally coherent set of reasons, i.e. the set of reasons that we would take ourselves to have if we made our current judgments about reasons coherent. (Although she has a complicated justification for not calling this an ideal-judgment theory.) And so on.

Generally speaking, the appeal of ideal-judgment constructivism is that it helps us avoid the unintuitive implications of saying that we should do what we actually value doing. (After all, some of us value some pretty crazy things.) The problem is that constructivists have a hard time specifying the ideal conditions without presupposing any values, e.g., the value of knowledge, or consistency, or simplicity, etc. (As many commentators have pointed out, Rawls is especially egregious on this score, since he presupposes reasonableness in his supposedly neutral political theory.)

Usually, constructivists try to get around this problem by arguing that valuing knowledge, or consistency, or simplicity, etc. is constitutive of rational agency. So for example, Street argues that valuing means-end rationality is constitutive of rational agency: if someone says that (a) she takes herself to have reason to eat tacos, (b) she needs to go to Taco Bell in order to eat tacos, but (c) she has no reason whatsoever to go to Taco Bell, then she is failing to be a rational agent at all (as opposed to merely making a mistake about what her reasons are). Kant, Dreier, Railton, Velleman, and many others make arguments along these lines too. The upshot of these arguments is that every rational agent values knowledge, consistency, simplicity, or whatever, and therefore every rational agent values doing what she would value doing if fully informed (in the relevant sense).

3. Whose evaluative attitudes are relevant to morality?

Most self-proclaimed constructivists are internalists, i.e., they say that you should do something iff you value doing it. But a constructivist could also be an externalist, i.e., she could also say that you should do something iff someone else values your doing it. This means that constructivism includes, e.g., divine command theory, which says that you should do something iff God values your doing it, as well as cultural relativism, which says that you should do something iff your culture values your doing it.

This is mainly what I had in mind when I said that this way of understanding constructivism makes it more expansive than we might have thought. I think most philosophers restrict constructivism to what I call internalist constructivism. And I can understand why: what I call externalist constructivism occupies a strange metaethical middle ground that seems not-quite-constructivist and not-quite-realist. On one hand, these theories share the constructivist intuition that value requires a valuer; but on the other hand, they share the realist intuition that value comes from something external to you. Ultimately, this is a terminological issue, and I could see it going either way. But on balance, I think these middle-ground theories have more in common with internalist constructivism than with realism, so I choose to call them a kind of constructivism too.

4. Is there something that you have to value if you value anything at all? And if so, what is it?

This is the question that separates what Street calls 'substantive constructivism' (or 'Kantian constructivism') from what she calls 'formalist constructivism' (or 'Humean constructivism'). Substantivists say that there is something that you have to value if you value anything at all. For example, Korsgaard says that you have to value humanity if you value anything at all. Formalists, in contrast, say that there is nothing that you have to value if you value anything at all. Street is an example of a formalist. (Well, almost: she thinks that you have to value taking the means to your ends if you value anything at all. She also thinks that you have to value constructivism itself, i.e., you have to affirm constructivism as a metaethical theory, insofar as it has normative implications.)

The virtue of substantivism is that it makes morality objective, in the sense that you have certain moral duties independently of what you value in particular (so long as you value something). The problem is that substantivism is hard to establish, especially a substantivism that gives you a robust set of moral duties independently of what you value in particular. And as a result, most (plausible) substantivisms are still quite formalist in spirit. For example, Sartre and Beauvoir are substantivists in the sense that they say that you have to value freedom if you value anything at all. But arguably, this is compatible with your doing anything at all, e.g. lying, killing, stealing, etc.

(To be clear, I don't mean to imply that Sartre and Beauvoir are full-blown constructivists. They don't seem to take a stand on the semantics of moral discourse, for example. But I do think that they come pretty close, and that we should include them in these discussions far more than we do.)

5. Do you have to value anything at all?

Korsgaard says the answer is no; she says that you can fail to value anything at all (even though she personally doubts that you do). Sartre, on the other hand, says that the answer is yes; he says that everyone has to value something.

Of course, how constructivists answer this question will depend on how they answer these other questions. For example, part of the reason why Korsgaard thinks you can fail to value anything at all is that, for her, the relevant attitude is reflective endorsement, which involves stepping back from your desires, asking if you endorse them, and saying yes. Arguably, this is something that you can fail to do. In contrast, part of the reason why Sartre thinks that you have to value something is that, for him, the relevant attitude is choice (if we can call that an attitude). And for Sartre, this is something that no rational agent can escape.

Notice, by the way, how this question interacts with the previous one: If a substantivist says 'yes' to this question, then she can claim an even greater kind of objectivity for morality than before: she can say that every rational agent has certain moral duties, period. Once again, Sartre and Beauvoir are good examples of this: they say that you have to choose freedom if you choose anything at all, and then they say that you have to choose something. Therefore, they might conclude, you have to choose freedom.

6. What kind of normativity are we talking about here?

Stepping back from constructivism about morality: I think that one of the main sources of confusion in discussions about constructivism is that different constructivists are talking about different kinds of normativity. Rawls is a constructivist about morality and political theory. Korsgaard is a constructivist about normativity and morality (insofar as she thinks that morality is normative). Street and Velleman are constructivists about reasons. And so on. (In this sense it might even be misleading to call constructivism a family of metaethical theories, as opposed to, say, a family of metanormative theories, or even something more general still.)

This is important, because one could be a constructivist about one kind of normativity but not about another. For example, one could be a constructivist about reasons but a realist about morality.

One could also be a constructivist about one kind of normativity and then a different kind of constructivist about another. For example, one could think that one's reasons depend on one's own values, one's moral duties depends on one's culture's values, one's legal duties depends on one's fellow citizens' values, and so on. (I actually think that something like this is closest to the truth.)
* I'm not attempting to capture the way that everybody uses the term 'constructivism' here. (I actually think that this would be futile, since everyone uses it differently.) Instead, I'm merely attempting to capture the way that I use it. I think that at least some philosophers also use it this way; but more importantly, I think that using it this way makes constructivism a clear alternative to expressivism, error theory, and realism, and it also explains why people as different as Kant, Rawls, Korsgaard, Velleman, Street, etc. all count as constructivists. (But see the last section for a qualification about what kind of constructivist they are.)

What do you think? Does this capture the way you use 'constructivism,' or does it at least come close to doing that? Does it make sense? If this is right, what else might constructivists disagree about?

-- Jeff Sebo


  1. Nicely written, Jeff. A question:

    How does what you describe distinguish constructivism from a form of realism that, say, defined the good in terms of pleasure or satisfied desires and the bad in terms of pain and frustrated desires? Those count as evaluative attitudes of some sort. The moral facts wouldn't be fully mind-independent, but since this theory would hold that the only things that matter are minds, this wouldn't make it any less realist than psychology is for its theories being mind-dependent.

    (Maybe you wouldn't want to count pleasures and pains as evaluative attitudes, but then just take the point in terms of desire-satisfaction)

  2. This seems to miss out, or at least underplay, the sense in which constructivism involves construction. I'd say that the paradigm constructivist thinks that some moral claims are true, but that they're true because they're the results of a pure constructive procedure ('pure' in the sense that the right result just is what the procedure gives us; there's no independent matter of fact the procedure is intended to discover). Norms are made, not discovered, but are no less real for that. Reflection on what norms apply to us is like playing a game (where the winner is whoever the rules identify as the winner) not like sitting on a jury (where the verdict should track, but may not track, actual guilt or innocence).

  3. Thanks for this, Jeff -- I really should know more about constructivism than I do and this is helpful.

    Like previous commenters, though, I see 1-3 at the top as too weak. They allow subjectivism and relativism to count as forms of constructivism, and I don't think constructivists want to be associated with such riffraff.

  4. Colin,

    Good point. Constructivism would be even more expansive if it included, say, analytic utilitarianism too. Part of me wants to bite the bullet and accept that it does. After all, analytic utilitarianism says that we create value by valuing things (in this case, by taking pleasure in them).

    But part of me also wants to resist making constructivism this expansive. Off the top of my head, I can think of two ways we might restrict it.

    First, what do we mean by pleasure and desire-satisfaction? Are they mere sensations, or do they have representational content? If the former, then maybe we should restrict constructivism to theories that ground morality in evaluative attitudes that have representational content.

    But say that pleasure and/or desire-satisfaction do have representational content. In that case, another option is to make a distinction between two kinds of evaluative attitude: the kind we have "before the fact" and the kind we have "after the fact". Very roughly, pleasure and desire-satisfaction are evaluative attitudes that we have "after the fact": they represent an already-realized state of affairs as to-be-realized. In contrast, moral beliefs, desires, preferences, aims, plans, etc. are evaluative attitudes that we have "before the fact": they represent a possibly-unrealized states of affairs as to-be-realized. And this makes them much more action-guiding.

    I could see restricting constructivism to theories that pick these more action-guiding evaluative attitudes. In which case constructivists should be synthetic utilitarians. They should say something like this:

    1. A state of affairs has value iff I would value it if fully informed.
    2. I would value pleasure if fully informed.
    3. So, pleasure has value.

    Does this make any sense? Can you think of other ways of making a principled distinction?

  5. Sam,

    I intentionally avoided explaining constructivism this way, because I think it invites a lot of confusion. When constructivists say that moral statements are true because they result from a procedure of rational construction, this usually invites the response, "Well what makes that the correct procedure?" And if constructivists want to answer that question without positing any prior and independent value, they have to say: we make it the correct procedure. Either because we value the procedure itself, or because we value other things, like knowledge and consistency, which the procedure, as a matter of fact, helps us to achieve.

    So, yes: any ideal-judgment constructivist will say that moral statements are true because they result from a procedure of construction. But this is because (a) moral statements describe what we value, (b) we value doing what we would value doing if fully informed, and (c) this procedure of construction, as a matter of fact, would make us fully informed. So the procedure has value, but only in this indirect way; and for that reason I think that defining constructivism in terms of the procedure is misleading.

  6. Neil,

    I hear you, but I think we should call these theories constructivist anyway. Every family has an uncle who shows up drunk on Thanksgiving and says racist things to your new girlfriend. Similarly, every metaethical family has members that the other members prefer not to associate with. e.g., we can imagine a realism that says that moral statements are true or false iff a piece of paper floating around somewhere in the universe says so. Few realists would want to associate with such a view, but if anyone defended that view, it would count as a kind of realism. Similarly, few constructivists want to associate with naive subjectivism/relativism, but they still count as kinds of constructivism. (And actually, we can imagine even more embarrassing members of the constructivist family than these, e.g. a constructivism according to which I should do something iff Kirk Cameron values it.)

  7. A couple thoughts:

    I don't think everyone would agree that 'paperism' would count as a form of realism. If someone thought mind-independence was all there was to realism, it might, but that's not that plausible anyway. Sayre-McCord and Railton have both said stuff about how realism involves giving some sort of literal understanding of the claims in question. Hard to make that precise, but it does seem that the only views that should count as realist are ones that give us a picture that matches our commonsense picture well. Paperism doesn't do that, so it doesn't seem realist to me.

    But on the more interesting question of how to characterize constructivism so that clear cases of realism are excluded: for me, at least, the metaphor of 'construction' implies a denial of fundamentality. Obviously, things that people intentionally plan and construct aren't fundamental parts of the universe, but you don't want to limit constructivism to views that see morality as something people actually got together and decided to put together (since, and this seems right, on the way you've described things some views will hold that certain constructions are necessary). So how 'bout this: build it into the third condition that the moral facts are less fundamental than the facts about evaluative attitudes on which they're based. That would rule out brute realism about pleasure being the good, as well as analytic forms of realism.

    The price is that you rely heavily on the notion of fundamentality. But anyone who's not cool with that should talk to Mike Raven anyway.

  8. One more quick thing (then I'll be quiet). This ties the two thoughts above together:

    I think there's some plausibility in saying that the commonsense view of moral facts does take them to be as fundamental as facts about our evaluative attitudes. And, if something like the suggestion I proposed comes out right, this would explain why constructivism doesn't count as realism.

  9. Jeff:

    “So the procedure has value, but only in this indirect way; and for that reason I think that defining constructivism in terms of the procedure is misleading.”

    But if the procedure has value only in the indirect way you describe, these (so-called) constructivists are avoiding a major difficulty with constructivism (as you say, the question 'what makes that the correct constructive procedure?') by... stopping being constructivists! They're no longer taking moral norms to be made, they're taking them to be discovered.

    I'm all in favour of that, but it seems an odd way to define constructivism. Have I misunderstood you in taking that to be your project?

    Or are we perhaps talking about different thinkers? I take Rawls's 'Kantian Constructivism in Moral Theory', for instance, to be a paradigm case of constructivism. How about you?

  10. Colin, fair point about paperism. I still think we should understand each major family (i.e. expressivism, error theory, constructivism, realism) as very inclusive, but I know that a lot of philosophers disagree.

    Sam, I think the exact opposite is the case. If constructivists say that the procedure has value because we value it, then this allows them to be constructivists, because otherwise they would have to say that something (i.e. the procedure) has prior and independent value, and therefore they would be realists.

    As for Rawls: I know he sometimes speaks as though the procedure has prior and independent value. But I think that we should regard him as making a mistake here that contemporary constructivists should correct. (And the mistake is understandable: it takes a long time to cleanse a philosophical theory of its normative assumptions.) So, no, I don't take Rawls as a paradigm case. I rather take him as the dude who got the ball rolling.

    Relatedly, Colin, I know that 'constructivism' implies construction and therefore a denial of fundamentality. But I think that we should resist this too. We should instead read 'constructivism' as implying mere creation, i.e., as implying that we create values by valuing things, rather than discovering them already-existing in the world. (But if your suggestion is the only way to keep analytic utilitarianism out, then I may change my mind.)

  11. Does 'creation' likewise imply a denial of fundamentality?

    I think just going straight for the f-word is better than talking of creation, since I take it that you don't want to exclude theories that don't think there's any act of construction or creation, right?

  12. This is a great discussion.

    I'm very much on board with Colin's insight that even realists might take evaluative attitudes to (superficially) determine the moral facts. So it's a tricky question how to carve this portion of logical space at its joints. I recently took a shot at it in my post on 'Desire-based Objective Value'. Like Colin, I came to the conclusion that we need to look at what considerations the theory treats as fundamental. Constructivist (or broadly 'subjectivist') views treat our subjective perspectives as bedrock. By contrast: full-blown realists (or broadly 'objective' metaethical views), if they appeal to our subjective perspectives at all, do so on the basis of some further, underlying consideration: e.g. that subjectively satisfied people constitute a better world. It's this difference in the fundamental structure of metaethical theories that we want to track. (Or so I propose.)

    I'm also sympathetic to Sam C's intuitive understanding of 'constructivism'. At least, I personally am drawn to the following kind of view: there's just one 'given' (unconstructed) normative fact, namely that which specifies the normatively privileged procedure ('rationality'), from which all else follows (via the procedure of rational construction).

    Now it sounds like you would want to classify me as a realist. That may be reasonable. Even so, there seems a pretty big difference between such 'procedural realism' and a 'substantive realism' which includes all sorts of first-order moral truths in the 'bedrock'. In some ways, the view I've described looks to have more in common with, say, Street than Parfit. (But perhaps these are merely superficial similarities, like those between dolphins and fish. For I grant that there's some sense in which the view I described does "start" with a fundamentally 'realist' commitment, even if it ends up looking more like some of the other metaethical families.)

  13. Perhaps an example would help: Kant is a constructivist if anyone is. Kant thinks that moral norms are those maxims of action that survive a certain reasoning process which is constitutive of practical rationality (and which is expressed by the various formulations of the categorical imperative). That process does not reveal to us what we antecedently value, or what we would value under ideal conditions: it constructs what’s right. It tells us what we should value. I’m not certain whether Jeff’s definition of constructivism lets Kant in or not (that perhaps depends on whether the structure of the will counts as an evaluative attitude). But at minimum, it doesn’t catch what’s central to Kant’s constructivism, what makes him the watershed figure he is, or what makes him a vital influence on (for instance) Rawls or Korsgaard.

    I suppose one way out of this dispute would be to agree that ‘constructivism’ is multivocal. Constructivism(Sam) is the view that a pure procedure can be basic to value; constructivism(Jeff) is a view about which bits of ethics are derived and which basic, and what their relation is.

    (For whatever it’s worth, the Stanford entry on ‘Contemporary Approaches to the Social Contract’ uses ‘constructivism’ in my sense, as the pure proceduralist reading of contract theory: agreement in the specified choice situation is constitutive of justification, not indicative of it. I don’t take this to be an argument-stopper, of course.)

  14. Wow, this is great! A lot to respond to.

    Colin and Richard, I agree with you: a realist can make evaluative attitudes foundational, and in several respects. They can treat them as having final value, as opposed to instrumental value, and they can treat them as having intrinsic value, as opposed to extrinsic or relational value. Analytic utilitarianism does both.

    And this makes me think that I agree with your point about fundamentality too. How about we make a distinction between an evaluative attitude's having value, and an evaluative attitude's object having value? Then we can restrict constructivism to theories that say that evaluative attitudes' objects have final value.

    This would make moral facts less fundamental than evaluative attitudes, in a sense. It would also keep out analytic utilitarianism, because analytic utilitarianism defines pleasure as the good, whereas a constructivist utilitarianism would say that pleasure is good iff the relevant evaluative attitude picks it out.

    And if we add that evaluative attitudes are the ultimate source of value, then we can keep out the kinds of realism that Richard discusses, e.g. a realism that says

    1. x has value iff the mind-independent moral properties supervene on x.
    2. The mind-independent moral properties supervene on x iff we would value x if fully informed.
    3. So, x has value iff we would value x if fully informed.

    Richard, I also agree with you that your procedural realism is, in practice, much closer to Street's constructivism than Parfit's realism. So why do I insist on calling it realism? I guess a few scattered reasons. First, it seems mysterious to me how anything can have mind-independent value in this sense. Even if you insist that only one thing has mind-independent value, all of the typical metaphysical, epistemic, semantic, and normative (e.g. motivational) issues arise for me. And, second, once you allow that one thing can have mind-independent value, why not allow that many things can? Of course you may have good reason to limit yourself to one thing; my point is that the first step is by far the most significant. So I call you a realist for the same reason that I call a monotheist a theist: sure, a polytheist believes in many Gods, but dude, you still believe in God!

    For me this is the whole ballgame. And my sense is that this is the whole ballgame for most constructivists too.

  15. Sam,

    Nice points. I have two things to say in response.

    First: I think that Kant is a constructivist in my sense. And the difference is that he makes his procedure constitutive of rational agency. As I say in my post, a constructivist can specify a procedure that applies to every rational agent if she can show that either (a) this procedure is constitutive of rational agency or (b) every rational agent has to value this procedure -- because she has to value this procedure if she values anything at all, and she has to value something.

    Both of these options specify a procedure that applies to every rational agent, but importantly, neither option requires saying that the procedure has prior and independent value. First, if we say that the procedure is constitutive of rational agency, then this means that we have to follow this procedure in order to be rational agents at all. (And this is a descriptive fact, not a normative fact.) And second, if we say that every rational agent has to value this procedure, then this means that, in fact, this procedure has value for everyone -- but only because everyone (has to) value it.

    Second: I concede that many people, maybe most people, use 'constructivism' in your sense. So why do I insist on this "revisionary" definition? A few reasons. First, as I said, it makes constructivism a clear alternative to expressivism, error theory, and realism. Second, it seems to capture the spirit of constructivists' project, even if constructivists like Rawls and Korsgaard sometimes talk as though the procedure has prior and independent value. Third, and relatedly, it captures where constructivism is heading. Very generally speaking (of course there are exceptions), I think that each constructivist has done a better job than the last of cleansing her theory of prior and independent value. Street, in particular, does an excellent job of this; and I fully expect future constructivists to keep pushing in this direction.

  16. I'm learning a lot from this discussion, Jeff - thanks for putting so much time into it.

    A small point: I don't think you want to base your characterization on an appeal to the objects of evaluative attitudes vs. the attitudes themselves, simply because there's some plausibility in the view that some evaluative attitudes are self-referential (I think Kant somewhere says something suggesting that he saw pain states has having content along the line of "get out of this state!").

  17. Jeff - fair enough. (Deism seems like a nice analogy, actually. "At least I don't believe in an intercessory God!" But yeah, point taken.)

    I'm not so sure that you can accommodate Kantians, though. We can ask: what's the significance of a procedure's being "constitutive of rational agency"? We have a dilemma: if 'rational agency' in Kants mouth, is purely descriptive, then folks might simply not care about being an agent of that sort. (You might try to insist that someone who violates the categorical imperative is not an "agent" at all, but this is mere persuasive definition. Try chiding Hume's knave for failing to be an "agent" in this descriptive sense, and see if he cares.) So, on this horn of the dilemma, you require the Kantian to give up their universalism. Alternatively, they might insist that they are describing a (brutely) normatively privileged psychology -- but that's effectively my position, which you don't consider constructivism at all.

  18. Richard - I'm not sure the dilemma you pose is really a problem for Jeff.

    Your dilemma would be a problem for Kantians insofar as they take themselves to be giving some explanation of normativity that's supposed to be universal. On the dilemma, they either get an explanation or universality, but not both. My guess would be that it's because most Kantians claim to be explaining normativity in terms of something more basic that calling them constructivists make sense, but if you're right that some of them can't coherently do that by their own terms, then there's no reason to worry about how to classify them (from an incoherent position you can derive the distinctive claims of any position whatsoever).

  19. Richard,

    I agree, deism is the perfect analogy! Let's go with that.

    As for Kant: I half-agree with you. The trick is to find a conception of rational agency weak enough to cover the sensible knave, the ideally coherent Caligula, and any other "competent" person, but strong enough to commit every rational agent to a certain procedure of construction. I think that we can do this -- but we might not be able to get everything out of the procedure that Kant wanted.

    Here's what I think we can do. We can find in rational agency a commitment to, e.g., (a) having true beliefs, (b) taking the means to our ends, and (c) doing what makes sense. (This comes from Railton, Street, and Velleman, respectively, among others.) This would already be significant: if we put these commitments together, we get a commitment to a procedure of rational construction that looks very much like reflective equilibrium, i.e., we get a commitment to learning all the facts, and then making our values as simple and coherent as possible in light of those facts.

    We can also get a weak version of the Categorical Imperative, i.e., we can get a principle of universalization similar to the kind found in neo-Kantians like Hare, Mackie, Korsgaard, and Velleman. Will this make the sensible knave, ideally coherent Caligula, etc. wrong to do what they do? Probably not. But it may at least make them pro tanto wrong: as Velleman puts it, his version of the CI, which he gets from our commitment to doing what makes sense, gives every rational agent "rational pressure" to tell the truth, keep promises, etc. This "rational pressure" can be, and often is, outweighed by reasons provided by our contingent aims. But it still guides our practical reasoning to some extent, and its effects are especially apparent over the long run.

    This is not everything that Kant (and Korsgaard) hoped for, and critics are right to call it "formalistic." But at least we can get a procedure of rational construction that makes people like the sensible knave and ideally coherent Caligula rare, as well as a weak Categorical Imperative that places pro tanto checks on the procedure's output.


Visitors: check my comments policy first.
Non-Blogger users: If the comment form isn't working for you, email me your comment and I can post it on your behalf. (If your comment is too long, first try breaking it into two parts.)

Note: only a member of this blog may post a comment.