January 01, 2011

Solutions to the Is-Ought Problem

There are two solutions to the is-ought problem: Define sentences that use ought normatively in terms of merely factual propositions, or deny that ought is meaningful at all. I'm going to present one way to do the former, and in doing so, undo the philosophical study of ethics by reducing it into studies for other fields.

I know that sounds a bit big-headed at first. We've had the unresolved is-ought problem for as long as we've had Hume's philosophy, whose most relevant passages that establish the is-ought problem can be found in A Treatise of Human Nature:
"Those who affirm that virtue is nothing but a conformity to reason; that there are eternal fitnesses and unfitnesses of things, which are the same to every rational being that considers them; that the immutable measures of right and wrong impose an obligation, not only on human creatures, but also on the Deity himself: All these systems concur in the opinion, that morality, like truth, is discerned merely by ideas, and by their juxta-position and comparison. In order, therefore, to judge of these systems, we need only consider, whether it be possible, from reason alone, to distinguish betwixt moral good and evil, or whether there must concur some other principles to enable us to make that distinction.
"If morality had naturally no influence on human passions and actions, it were in vain to take such pains to inculcate it; and nothing would be more fruitless than that multitude of rules and precepts, with which all moralists abound. Philosophy is commonly divided into speculative and practical; and as morality is always comprehended under the latter division, it is supposed to influence our passions and actions, and to go beyond the calm and indolent judgments of the understanding. And this is confirmed by common experience, which informs us, that men are often governed by their duties, and are detered from some actions by the opinion of injustice, and impelled to others by that of obligation.
"Since morals, therefore, have an influence on the actions and affections, it follows, that they cannot be derived from reason; and that because reason alone, as we have already proved, can never have any such influence. Morals excite passions, and produce or prevent actions. Reason of itself is utterly impotent in this particular. The rules of morality therefore, are not conclusions of our reason.
"No one, I believe, will deny the justness of this inference; nor is there any other means of evading it, than by denying that principle, on which it is founded. As long as it is allowed, that reason has no influence on our passions and action, it is in vain to pretend, that morality is discovered only by a deduction of reason. An active principle can never be founded on an inactive; and if reason be inactive in itself, it must remain so in all its shapes and appearances, whether it exerts itself in natural or moral subjects, whether it considers the powers of external bodies, or the actions of rational beings.
"It would be tedious to repeat all the arguments, by which I have proved [Book II. Part III. Sect 3.], that reason is perfectly inert, and can never either prevent or produce any action or affection, it will be easy to recollect what has been said upon that subject. I shall only recall on this occasion one of these arguments, which I shall endeavour to render still more conclusive, and more applicable to the present subject.
"Reason is the discovery of truth or falshood. Truth or falshood consists in an agreement or disagreement either to the real relations of ideas, or to real existence and matters of fact. Whatever, therefore, is not susceptible of this agreement or disagreement, is incapable of being true or false, and can never be an object of our reason. Now it is evident our passions, volitions, and actions, are not susceptible of any such agreement or disagreement; being original facts and realities, compleat in themselves, and implying no reference to other passions, volitions, and actions. It is impossible, therefore, they can be pronounced either true or false, and be either contrary or conformable to reason."
-- David Hume, (A Treatise of Human Nature, Book III, Part 1)
I'm going to eliminate the issue formally by actually showing how normative claims are made and how we can undo the error that "a conformity to reason" does not yield claims whose illocutions are different from a speaker's assertion of facts.
How do we even argue ethics in non-academic environments? Well, it's complicated, but I think that we could simplify it with a sample dialog. Suppose that someone (A) says the following to another person (B):
A: You really ought to quit smoking.
B likes to smoke. He has no intention of quitting, and he doesn't really appreciate A's prescriptive tone, so we can reasonably say that a conversation of this sort might continue with this:
B: Why should I?
How does A (or anyone who asserts what another person ought to do) respond? Well, he probably does so most easily by saying this:
A: You ought to quit smoking because Φn.
Here, Φn is just some other sentence or sentences (which we can translate into logical forms and then define into a single proposition), all of which argue [Φn ⇒ *Ought*(b,Φm)] ∧ [*Ought*(b,Φm) ≝ "B ought to quit smoking." (or an equivalent translation of that sentence)]. This is a problem for the is-ought problem because we are always arguing for prescriptions on the basis of assertions of fact. This means that we're always going about ethics the wrong way (that is, arguing on facts for a proposal of what we ought to do) or it means that normative or prescriptive claims translate into some sort of truth-apt assertions to make the discourse coherent.

The issues that are at the center of all ethics are just these questions: What is this Φn, exactly, and what axioms prove Φn? All of the schools of contemporary normative and metaethics answer these questions in different ways, but they never deny that the model under which the problem arises doesn't exist. No one argues that arguments for and against normative claims and prescriptions don't exist, that people don't actually argue this way. Ethicists' burdens are to interpret what the arguments assert and then to argue why they are correct or incoherent.

It's pretty straightforward from here. We have no reason to deny that Φn exists. In fact, logic tells us that there are trivially many reasons that we could pose for *Ought*(b,Φm) via the weakening rule ([P ⇒ [Q ⇒ P]]). Our burden, however, is not the providence of just any assertions that we could vacuously write as sufficient conditions for *Ought*(b,Φm) because *Ought*(b,Φm) is what someone in B's situation is denying (B asserts ¬*Ought*(b,Φm)). Φn must convince B of *Ought*(b,Φm).

Here are just a few examples of what Φn could be:
Φn1: A: Smoking causes lung cancer.
Φn2: A: I'll kill you [B] right now if you don't quit smoking.
Φn3: A: You're smoking at a gas station.
Φn4: A: I am speaking on behalf of God.
These examples of false and true assertions could convince B. But we should always remember that B may very well have retorts (Φo) that frustrate A's effort to convince him of Φm. For some instances:
Φo1: B: Good! I hope I die of lung cancer!
Φo2: B: You couldn't kill a fly, pussy!
Φo3: B: This is a smokeless cigarette.
Φo4: B: God doesn't exist.
However, all of these assume some coherent means of blocking an argument which we could express as follows: That I ought do something means that all of the propositions which identify what it means to say that I ought do something are true. When we do this, however, we also outline all of the merely factual propositions (Φq) that comprise the prescriptive proposition *Ought*(b,Φm).

Below I have produced a model that has some decent efficacy in outlining all of the constituents of a prescriptive or normative statement, which I will write from A's perspective:
  1. You, B (b), want Φp. [Want*(b,Φp)]
  2. You want that you don't want Φp, but that you get it. [*Want*(b,¬*Want*(b,Φp)) ∧ *Want*(b,*Get*(b,Φp))]
  3. If you get Φp, then you won't want Φp. [*Get*(b,Φp) ⇒ ¬*Want*(b,Φp))]
  4. You are more likely get Φp given Φm, but are less likely to get Φp given not-Φm. [P(*Get*(b,Φp)|Φm) > P(*Get*(b,Φp)|¬Φm)]
  5. Your doing something is Φm. [Φm = *Do*(...b...)].
  6. That you ought Φm is the previous assertions (Φq) 1 through 5. [*Ought*(b,Φm) = [Φq1 ∧ [Φq2 ∧ [Φq3 ∧ [Φq4 ∧ [Φq5]]]]]]
  7. The facts of Φn imply (4).n[P(*Get*(b,Φp)|Φm) > P(*Get*(b,Φp)|¬Φm)]]
English has made good progress in conflating two different senses of the word ought, which has made this reduction much easier to achieve than it would be in a different language that differentiated its predictive sense (e.g. "He ought to be here by now,") over its more obligatory sense (e.g. "People ought to floss their teeth regularly."). This ambiguity is actually a good clarification, as ethical statements, then, reduce into wants and expectations of others' wants, and those things can be empirically investigated matters. The prescriptive language is some simple abbreviation for the more complex argumentation, but that argumentation is not, itself, moral, just a common tool of speech that we use to influence social behavior (which is what language primarily does for us).

The problem with our regular talks on these subjects is that we (when we resemble people A) assume all of the assertions that we don't make in a discourse, all of which we (when we resemble people B) can quite easily deny. We're not challenging what "ought be" on the grounds of some other prescription (because we could always regress by demanding more justification for that prescription) or another "value on High". We just submit and challenge them on the grounds of the assertions of our wants and incentives.

We can look at Φn again, but this time add the remaining constituents to make them more closely match the members of Φq. We can see that by explicitly stating Φq1 and Φq4 and some immediately accepted consequences from the assertions that were made above parenthetically, that A has indeed presumed these facts when he was arguing with B.
Φn1: A: Smoking causes lung cancer.
(You don't want to die of lung cancer, but you more likely will if you continue smoking, yet you more likely won't if you stop smoking.)

Φn2: A: I'll kill you [B] right now if you don't quit smoking.
(You want not for me to kill you, and I more likely will if you do continue smoking, and I more likely won't if you stop smoking.)

Φn3: A: You're smoking at a gas station.
(If you smoke at a gas station, you more probably will explode. You want not to explode, and you more probably won't explode if you quit smoking.)

Φn4: A: I am speaking on behalf of God.
(God is vengeful if ye obeyeth not his Word. He shall not smite thee if ye cease thy smoking; but his wrath will be upon you with such continued acts of sin. To the more, you wish not be smitten by him.)
If I have not yet convinced you that you are talking about facts over wants specifically when you are making prescriptions, I'll leave comments open to objections. (This skeleton should present enough to state objections to the details or to request them if you want them.)

I have mentioned already that this tactic of reinterpretation completely removes ethics as a self-standing philosophical study. Any issue over what all people ought to do will have to comprehend the universal wants of humanity or the universal facts that determine all people's individual wants. No philosopher will be able to give an adequate account of such things, since individual objections regarding personal wants cannot be refuted by mere logical analysis. Ethicists and other philosophers have to take many of their assertions at face-value. For instance, a philosopher often cannot reasonably charge that one's claims to have peculiar or unfeasible wants are self-deceptive. His investigative tools are just inadequate to raise that kind of objection. The tools for the sincere investigation of human nature must be empirical, not philosophical. Without this direct interference against philosophical musing on ethics, all claims over what is "right" versus "wrong" remain, as some disgruntled commentators have rightly seen it, veiled emotional battles for singular Utopian ideals.


  1. What about cases where what is asserted is that one ought to want x?

    1. Appeals of that sort are not a major problem with the theory. There are many cases where the possesion of a specific desire actually hinders a more pressing and inviolable want (like avoidance of lung cancer).

  2. Where x is not free, define (for an agent B) *Want*(b,x) as Φm.

    *Want*(b,x) ≝ Φm.

    Do you suspect a problem with this formulation?

  3. My worry is no doubt ill-formed, but here it is.

    I take it that defenders of the is/ought distinction are happy to have some empirical considerations involved in ought judgments. But they are going to balk at full reduction of ought to is. So, take one of your examples:

    You ought to quit smoking.

    Why? Well, there is an empirical component -- smoking causes lung cancer -- and also a normative component -- you ought to avoid lung cancer.

    If one replies, "Yes, but I want lung cancer." Then barring some very weird motives, I think the right reply is, "Oh, that's crazy! You *ought* to want to avoid lung cancer."

    Now, you might try to back up a level and say that the right reply is to look at whether this person wants to want to avoid lung cancer. But I think that we don't really reduce norms to wants in this way. Even if this benighted person is self-consistent in wanting to want lung cancer, he is still *wrong* to want lung cancer. He ought not have such a desire.

    1. You could argue that the desire for lung cancer stems from faulty affective forecasting — a misunderstanding of lung cancer that prevents him from taking into account the aspects of the disease that he does not want. So to argue that he ought not want lung cancer, you again appeal to relevant wants that you presume he has (ex. to die a horribly painful death, worry his family, waste his life, etc.).

  4. There are two criticisms that I would make about the idea of a self-standing "normative component." The first I put at the end of my blog: "Without...direct interference against philosophical musing on ethics, all claims over what is "right" versus [what is] "wrong" remain, as some disgruntled commentators have rightly seen it, veiled emotional battles for singular Utopian ideals."

    Consider B as someone who would challenge anyone's authority to make any normative claims against his behavior. "You don't have any more special access to any absolute normative scale than anyone else," would be B's Φo, and that would counter A well because very few are going to convincingly argue that they have special, absolute access to the truth of what people ought to do.

    Put it another way. A can call B's motive "weird" or "crazy" and can poison the well, but A doesn't really argue any fact outside of what A merely asserts someone ought to do when he makes those accusations. Until A defends his belief with a fact about B's actual wants, not just about what A insists that B's wants be, B has no clear reason to consider A's prescription.

    In order to argue why B is wrong to want lung cancer, or to want to want lung cancer, A has to argue a point in fact about some "good" in avoiding lung cancer. B's problem isn't a problem of consistency no matter how he answers in his semantic ascents, since there's no logical contradiction occurring on his end (e.g. "I don't want to sleep right now, but I want to want to sleep right now," is still logically consistent). B's problem, if there is one at all, is that he does not see some fact that would make him quit smoking or make him seek longevity over self-destruction.

    Here's a second criticism, which I find many people accept at face value when I tell them:

    'Φn ⇒ *Ought*(b,Φm)' can be well-formed for HOL. Fine, but does it actually state a fact? If it does, then *Ought*(b,Φm) has a truth-function, and it can't be circularly defined (that is, it can't appeal to "being good," because "being good" is just "being as we ought to do something.") If it doesn't, then either you have to provide an entirely different semantics for both *Ought*(b,Φm) and Φn because you won't have any means of not arbitrarily switching between truth-functional and non-truth-functional values in a semantic model and because it will suffer from Hume's objections that lead to the is-ought divide.

    "Deontic logics" do junk in the latter area, but in my estimation, those logics are doomed because there's always some inadequate relation between truth-functions and "permissibility" functions, which makes logical operators get goofy interpretations under the mixed semantic model.

    If ethics has access to any facts at all, those facts have truth-functions.