« Efficient Economist's Pledge | Main | Rarity Anomalies Remain »

April 17, 2009


The Holocaust should never be used as an example of anything if you can possibly help it, cf Godwin's Law. And I don't feel bound by this example because unlike Robin I don't feel bound to measure utility for another person based on what they think they want.

I haven't found an example where I can't either challenge the assumption that the unpalatable thing maximises utility (like this story) or accept the "unpalatable" conclusion (eg torture vs dust specks).

I realize we all have the "Uncle Tom's Cabin" model of slavery that we use to drive our visceral hatred of slavery, but do you think people would actually be in favor of the "Gone With the Wind" model? On the one hand I agree that utilitarian stories are important to supporting our inclinations, but what accounts for the difference in how we think of government by another individual (slavery) and official government. It seems like other philosophical principal is necessary.

Citizens of classical Greece would have found it easier to identify with a story in which slavery passed a utilitarian test than one where it didn't.

Sumner wrote "it ought to be possible to find some real world example, in some country, at some point in history, of an actual public policy that clearly would pass the utilitarian principle, and yet which shocks our moral sensibilities. And I’m not sure that it is possible."

Yes it is for a society in a Malthusian trap that lets all deformed babies die. And if this isn't enough imagine that the society has enemies that wish to torture, rape and kill all of the members of the society.

Paul: Maybe this isn't the place for this discussion, but I'll try to provide an example which might give one pause. In general, I think Overcoming Bias suffers from too many ill-informed discussions in moral philosophy and that we would all be better off reading an introductory textbook on the subject, but I can't resist. I hope it's not too crude.

Suppose there is a very beautiful woman who is not strongly psychologically effected by sleeping with other people (an unusual characteristic to be sure, but likely possessed by a small minority of women). She doesn't particularly want to do so with most men, but neither would it cause her much distress. Is this woman morally obligated to go around sleeping with unpopular/unattractive men? It seems very likely that total utility would be greatly increased. (Consider, for instance, what the men would be willing to pay).

I read the topic post several times and was amazed by how difficult it was to fit to a coherent framework for evaluation. I then visited the referenced blog posts and got the necessary context—once again, "rationalists" laboring under the transparent assumption of an objective basis for subjective ranking.

Objective measures of expressed agreement—yes.
Objective measures of subjective utility—not even wrong.

To start, it seems that most of those involved in such debates would do well to read this.

I assume we understand that maximizing the sum of utilities is an approximation to what we want. Certainly, given two positions of equal or near equal sum of utilities we would like a position where there is more "equality". Should the objective function then not look something like

maximize sum_i (u_i) - (positive constant)* (sum_i (u_i - u)^2)
u = 1/n * sum_i (u_i)
(+ other constraints)

This kind of formulation is pretty standard wherein you want to maximize mean but are simultaneously interested in minimizing variance. I assume our other constraints would be values like liberty, which would likely put pressure on the u_i's to be disparate since our abilities are different and given a relatively liberal world, free markets would likely ensure that people land up with very different utilities.

This is a fascinating idea, but I'm sure we could imagine any number of stories where actors fail to achieve some utilitarian outcome but still receive moral praise. Usually these stories have a component of "sacrificing one's self for some [non-utilitarian] virtue or cause." While one could perhaps argue that the promotion of the promoted values might be good instrumental virtues to promote utilitarian outcomes, I don't think it can be said that that's why they're being praised. But perhaps the operative principle with these stories of heroism is that we like to see other people make sacrifices for our causes and are willing to praise their doing so.

Part of the problem with these debates is that "pass the utilitarian principle" and "shock our moral sensibilities" are treated as constants across all parties when in fact they aren't. Utilitarianism allows a lot of wiggle room by tweaking values and weightings, and different people have different moral sensibilities.

Jess Riedel uses the example of a promiscuous woman who runs around maximizing utility while offending our moral sensibilities. There are two ways to address this dissonance. 1) Construct a utility function that weighs down the benefits of promiscuity while increasing its costs, say by introducing third-parties adversely affected by the behavior. 2) Allow that promiscuity, given the conditions Jess specifies, is an inoffensive thing.

The real question has to do with promiscuity--or whatever topic a story addresses--not with what moral theory you use to frame the analysis.

Ben: the problem is nothing to do with promiscuity per se; the framing of the problem makes it pretty explicit that if she were enjoying these encounters there would be no moral problem for consequentialists, and I think most readers here would also be happier if she were appropriately paid for the encounter.

I think this is a special case of the more general observation that the rule of equal consideration is impossible to live by; after giving away enough sympathy sex, she'll reach the stage where the slight extra ennui caused by one more such encounter outweighs the happiness it might bring him, even if it would seem the other way to an external observer.


Objective measures of subjective utility—not even wrong.

That is incorrect. Robin (and the article he mentions) is discussing utilitarianism: the view that the right act is the one that maximizes happiness (note that it says this in the first quoted sentence). Happiness is not the same as the 'utility' that economists talk about. These are two different concepts that are both sometimes given the name 'utility' and this poor choice of naming leads to a lot of confusion. For what it is worth, the ethics usage is much older than that in economics.

I suppose this argument really is a very effective counter to the hypothetical scenarios where an efficient outcome would be morally shocking. It seems so obvious once stated...reading Bryan Caplan's posts on the subject, the subject seemed more complex.

Still, for practical purposes, a lot of the hypothetical scenarios that people bring up would be best solved for efficiency by allowing personal liberty. If you imagine a world where for many people, the slavery transaction is efficient (maximizes joint utility), then the best way to reach the efficient amount of slavery is to allow legally enforceable contracts. It seems to be the cheapest way to reveal the relevant preferences. But then, that's the entire justification for free markets, so I guess that goes without saying.

...Hm, there's really not much insight there.

@Toby: Did you read the simple page that I linked to in my comment. Does it not address exactly these points?

"how do we know that slavery was so awful? Because we have been exposed to accounts of slavery in the arts which vividly showed how the suffering of slaves was immeasurably greater that the frivolous pleasures of the slave-owner."

More plausibly: Because we have been exposed to accounts of slavery in the arts which vividly showed the suffering of slaves, period. The comparative weight of the slave-owners' pleasures seems neither here nor there, emotionally speaking. We would sympathize with the portrayed slaves no less if we saw the slave-owners gleefully gloating about how incredibly much they enjoy oppressing people.

N.B. You can't equate "non-utilitarian grounds" with "abstract philosophical principle". It's uncontroversial that many moral intuitions are driven by sympathy and other emotions. The question (at least according to one methodology) is what abstract principle best systematizes our responses. Utilitarianism is one proposed 'abstract philosophical principle'. There are others, that also take human suffering into consideration but in different ways (e.g. giving priority to the worst off, or discounting sadistic/vicious "pleasures", etc.).

So the argument here seems like a non-starter.

So the argument against utilitarianism is that a utilitarian would support slavery if it did more good than harm?

Such story arguments against utilitarianism (or any moral position, really) are usually guilty of equivocation. "Slavery" is a word that includes its horribleness in our understanding of it (at least today).

Consider that if a utilitarian were to consider unpaid forced labor ethical, relative to the "typical" situation, either the worker would have to be less oppressed, or the benefit greater. But such situations exist: for example, volunteer work is free unpaid labor, because the worker is willing; the military draft is dangerous forced labor, but the benefit is perceived as much higher to society.

A slavery that was defensible on utilitarian grounds would probably not be recognized as slavery at all.

A coherent account of moral decision-making presumes nothing about the subjective states of others, which by definition cannot be known. Any moral agent always only expresses its own nature within its environment. All other metaethical models suffer from the well known internal contradictions of consequentialist and the infinite regress of deontological ethics.

To the extent that actions are assessed as promoting the agent's values, such actions are considered (subjectively) "good."

To the extent that actions are assessed as promoting the agent's values over increasing scope of consequences, such actions are considered (subjectively) "better".

To the extent that actions are assessed as promoting an increasing context of values (of agency over time, or agency over a group) over increasing scope of consequences, such actions are considered (increasingly objectively) increasingly "moral" or "right."

These inequalities suggest an "arrow of morality" by which any agency can always point in the direction of perceived increasing rightness.

In general, actions are assessed as increasingly moral to the extent they are seen as promoting an increasing context of increasingly coherent, fine-grained, hierarchical, present but evolving values by means of methods seen as increasingly effective, in principle, over increasing scope of consequences. Lather, rinse, repeat with agency (its nature, therefore its values) evolving with respect to "what works."

This coherently and extensibly explains the basis of our evolved moral instincts, and our culturally reinforcing moral codes, and suggests a framework for increasingly intentional development in the direction of increasingly effective promotion of an increasing context of values naturally seen (necessarily from inside the system) as increasingly right.

Slavery is not inherently immoral. Indeed its very real benefits can easily be demonstrated on a consequentialist basis within a particular context. But, with increasing context its inefficiencies—as it fails to exploit synergies inherent in interactions between near equal agents—become increasingly apparent, thus increasingly "wrong."

Obscure? Probably so. I apologize for lacking the bandwidth to elaborate. I post this only as a possible seed of thought that might take root in a fertile mind.

> The Holocaust should never be used as an example of anything if you can possibly help it, cf Godwin's Law.

Please kill this horrible, horrible meme. Genocide is far too common an occurance to be neglected as a possible circumstance within which for moral principles to operate. Principles which cannot so operate are badly-formulated principles.

Bringing up the holocaust to attack utilitarianism is a disingenuous argument, as it implies that utilitarianism is somehow more sympathetic to the idea than other moral philosophies.

In fact, it is virtually impossible to make a plausible case for a holocaust using utilitarian principles. And I would say it is much harder than under the most common moral codes (religious ones.) Islamic radicals, for instance, frequently call for a holocaust by citing their religious scriptures.

Jeff Riedel, I would praise the woman strongly if she so acted, but she is not obligated to do anything. In a saner society she would be offered large amounts of money to exercise her comparative advantage.


Did you read the simple page that I linked to in my comment. Does it not address exactly these points?

Ah, I had a quick look at it and only saw the information on economists' utility. Now I see that it has a quick gloss on the original conception of utility as happiness at the top. It thus talks about these points, but it does not sufficiently address them. If it really were meaningless to talk about whether one person was happier than another, then your points would be correct. However, it is in fact possible to make such statements (eg I am happier now than McCain was at the time he lost the election). No-one really doubts that this is true, as I am quite happy now and he was no-doubt rather crushed at the time. Economists point out that we don't have a fool-proof method of accurately performing interpersonal comparisons in all cases, but this is a long way from it being meaningless in all cases. It is quite conceivable that there is a fact of the matter as to whether I am currently happier than McCain, even if there is no easy way to find out. Utilitarians (and many other philosophers) believe this.

Suppose there is a very beautiful woman who is not strongly psychologically effected by sleeping with other people (an unusual characteristic to be sure, but likely possessed by a small minority of women). She doesn't particularly want to do so with most men, but neither would it cause her much distress. Is this woman morally obligated to go around sleeping with unpopular/unattractive men? It seems very likely that total utility would be greatly increased. (Consider, for instance, what the men would be willing to pay).

What Eliezer said. To elaborate, "obligations" aren't morally primitive or simple, and don't proceed in any simple way from utility maximization; they are the sorts of things "individual rights" are taken as in Interpersonal Morality – concepts used in interpersonal judgment, that are said to exist in a non-socially-relative sense when their use would have good consequences. There being an "obligation to X" means something like "it should be regarded (by an individual or society) that you should do X, and if you do not do X you should be called a bad person, shamed, not cooperated with, or other appropriate social punishments." It seems very likely to me that for a person or society to hold anyone obligated to have sex would have seriously bad consequences, so I hold that no such obligation exists.

Robin - could you clarify why you think this is "a powerful argument"? As noted in my above comments, it simply isn't true that "our initial visceral reaction against slavery" implies an implicit commitment to utilitarianism over any other moral theory according to which suffering is bad. So it looks to me like an extremely poor argument -- am I missing something?

Richard, responding to your request, I added to the post.

"The argument, I think, is more that we overgeneralize from the stories where we first picked up our morals. For example, we first hear stories where slave owners gain less than slaves lose, and then come to see all slavery as bad."

But this just assumes that in first hearing such stories, what we are responding to is specifically the fact that slaves lose more than owners gain. Why think that? It seems overwhelmingly more plausible that, right from the start, what we are responding to is simply the suffering we see portrayed.

Nick, Eliezer: As I understand it, utilitarianism is a moral theory which makes normative claims about the actions of individuals, e.g. agents should perform whichever actions maximize utility. So, utilitarianism says either "the woman ought to act in this way" or "the woman ought not act in this way". This is what I mean by the woman being "morally obligated".

This is a completely separate from how other agents ought to act with respect to the woman (a different normative claim) or how society should structure its laws/customs with respect to these actions to best bring about certain states of affairs (an empirical claim).

In other words, Nick, I think I disagree with your statement (as I understand it) that the term "utilitarianism" is used to refer to a collection of judgments about how society should be structure to maximize utility. If so, then utilitarianism would merely be an empirical science and we would still be left with the moral question "ought we implement utilitarianism?"

Eliezer, would you say the woman is a Bad Person for not so acting? Would you say she is a worse person than someone who, for instance, breaks into a poor graduate student's office and steals his laptop?

Daniel Franke: Genocide is a fine example to use, but use a different example of genocide. Unless you're drawing on specific unique features of Nazi Germany in your example, reaching for the Holocaust as your example of evil every time strikes me as a kind of intellectual laziness.

Richard, if the efficiency violation in the first influential stories where we learn to see behaviors as immoral were an incidental feature, we should see many such stories where efficiency was not violated. The hypothesis that the efficiency violation was crucial to embracing the learned moral predicts fewer such exceptions.

Robin - if there are truly as few stories where efficiency is not violated as you suggest, then it is not merely the first stories in which efficiency is violated, but in all likelihood the first, the middle, and the last. On what basis do you claim that it is the first stories which are critical?

Furthermore, if there are truly so few stories were efficiency is not violated, then might this not reflect an underlying reality that efficiency tends to be violated? In that case, then causality becomes a question: maybe the moral conclusion is not caused by exposure to such stories, but rather, both the moral conclusion and those stories are both caused by the fact about efficiency.

Furthermore, is efficiency really violated in the stories, or is the reader's reaction one in which the harm to the victim outweighs the benefit to the victimizer? This may be a result of a pre-existing moral commitment on the part of the reader, which is thus elicited from the reader by the story rather than impressed into the reader by the story.

In short, the hypothesis seems highly speculative and the slim evidence offered in support seems consistent with many other hypotheses which seem to me on their face to be more credible.

Finally, I would like to repeat Richard's objection to "non-utilitarian grounds, i.e. based on some abstract philosophical principle." This is a false dichotomy.

Jess is right. It's hard to get the three classic modes of deontic logic to apply in any form of utilitarianism (these modes are permissibility, impermissibility, and obligation). On its face, utilitarianism seems to allow only two modes, depending on whether an action increases or decreases utility, and those two modes are impermissibility and obligation. So, Jess's woman is in fact obliged to sleep with as many unnattractive men as possible, unless there is some use of her time that more effectively maximizes utility -- murdering the congenitally unhappy, perhaps? -- in which case sleeping with as many men as possible becomes morally impermissible.
This unfortunate consequence of utilitarianism was recognized even by Mill, whose proposed solution was formally similar to Nick Tarleton's, but you'll notice that it also resembles the Kantian universalizability tests. Once you start talking about which generally applicable rules are utility-maximizing to follow you've begun to smuggle deontology in by the back door. Which is natural, I suppose, since the normative prescriptions of deontological systems -- even if these systems themselves don't stand up to thorough scrutiny -- bear at least a passing resemblance to what we call "ethics" in normal life.
All of this has been well hashed out. As Jess said, perhaps OB readers would do well by reading moral philosophy textbooks.

Robin, When you asked if anyone else had a similar idea, I slightly misunderstood the example you were referring to. Now that I see the quotation above, I do think that Peter Singer had a similar observation about vegans (but I don't recall whether he drew from that example the sort of broad generalization that I did.) Regarding Jess's example of the beautiful woman, that is exactly what I was referring to---anti-utilitarians seem to want to use these far-fetched thought experiments, rather than consider real world cases. If utilitarianism is so bad as a policy guide, why not use real world policy examples. I'll use one that (very slightly) shocks my moral sensibilities. Mandatory seat-belt laws that do save many people from dying and being paralyzed, but treat the public like children. But if paternalistic legislation is the worst counterexample that I can find in the real world, and if most people wouldn't even be shocked the way I am, then maybe utilitarian policy evaluation isn't that bad.
Regarding David's last point, maybe policy rules are sneaking deontological principles in by the back door (although I have some doubts.) But even if that is so, my response would be that in any case utilitarianism is only interesting for policy evaluation, not as a personal guide to life. I don't expect people to care more about the welfare of 3 children on the other side of the world, than their own two children. I'm a pragmatist.

Jess, if I were to hold her to the same standards that I hold myself, and there's nothing else more useful that she can do with a given hour, then yes, she is a bad person for not alleviating the misery that she alone can alleviate more effectively than anyone else, with a comparative advantage greater than any other comparative advantage she holds. But in fact it would be inappropriate for others to hold her to such a standard, unless she endorsed that standard herself, or she asked them for strict advice.

Utilitarian policy evaluation in the real world?

Let's see:

How about the use of scorched earth tactics and collective punishment in counterinsurgency warfare? This article describes simple and effective tactics by which a conventional army can defeat a guerrilla force. In the recent past, Russia used them very effectively to quell the insurrection in Chechnya. A few good massacres, such as the Hama massacre committed by the Syrian government, could easily save more lives in the long run by ending drawn-out conflicts before they start.

Additionally, I propose that we deter suicide bombers by executing the immediate family of people who commit such attacks. A suicide bomber is clearly willing to sacrifice his or her own life for the cause, but their relatives might not be willing to share that sacrifice.

Doug S., I have suggested the same policy regarding the families of terrorists. You're the only other person I can think of who agrees.

While I also become frustrated at this way of doing philosophy, I think there is a really powerful objection lurking that is rarely given its proper treatment. The objection is that we don't ever really know the consequences of our actions. The objection is doubly powerful because not only do all theories of well-being have serious problems, but we also just cannot know or predict the future. As a result, we cannot know with any reasonable degree certainty the consequences of our actions. There is a decent paper by James Lenman about this called "Consequentialism and Cluelessness". Lenman also has a good response to Parfit's new book if anyone is interested.

I do have grave misgivings about such a policy (one contradicts the received wisdom about ends and means at one's peril), but I also think it might very well be an effective policy for a country enduring repeated suicide attacks. The biggest downside that I can see is that it could easily be a public relations disaster for any government that tries it.

Interestingly, under the Geneva Conventions, the use of civilians as human shields is a war crime, and countries at war have every right to go ahead and kill them if they interfere with an attack on military targets.

(In other words, under international law, Hezbollah commits a war crime every time it builds a school and puts an arsenal in its basement, and the civilian casualties inflicted by Israel during its recent conflict in Lebanon were completely legal.)

preference utilitarianism solves these sorts of thought experiments IMO. In the case of the beautiful woman, the classic utilitarian argument would be to say that her preference to not sleep with unattractive men is outweighed by the preferences of those men to sleep with her. But what if instead of trying to create "aggregate preferences" we only accounted for individual preference? Each individual preference of an unattractive man to sleep with her is perfectly counter balanced by her preference not to sleep with that particular man.

You are accusing consequentialism of curve fitting. But this is really the problem of induction restated. Since we can never prove that induction isn't working what alternative do we have? Strict deontological maxims?

I like Jess Riedel's example, although as others have suggested, utilitarianism seems to address more what one shouldn't do than what one should. For a more straightforward and realistic example, it wouldn't pain many of us to give a dollar (or ten) to every single charity we come across every time we come across it. Are we obligated to do so?

But suppose there was a network of voyerist photographers and fetishists who only photograph Amish women in the shower, and only from the neck down, so that these women will never see/recognize the photos of themselves being passed around on the Internet and thus will never experience any pain from it. Maybe they even do it with remote hidden cams so that getting caught is not a possibility. Utilitarianism seems to suggest this is OK, I suggest it is not.

"...reaching for the Holocaust as your example of evil every time strikes me as a kind of intellectual laziness."

Paul, I agree with those who criticize Godwin's Law. "Intellectual laziness" is useful shorthand when people want to discuss evil in some context without discussing endlessly what evil is:

1st guy: "Suppose you had this evil guy. An abortionist".
2nd guy: "Wait, I don't find abortionists evil."

1st guy: "Suppose you had this evil guy. A politician".
2nd guy: "Not all politicians are evil. Take for example my favorite..."

1st guy: "Suppose you had this evil guy. A murderer.".
2nd guy: "What do you mean by murder? Killing is sometimes defensible..."

1st guy: "Suppose you had this evil guy. Like Hitler".
2nd guy: "Ok. Go on."

As a utilitarian, i get annoyed by the rather constant attacks on my morality of choice(tm). So being the type to over think things, i generally feels there's three major complaints against utilitarianism.

1) Almost every moral philosophy [b]will[/b] make an "Appeal to utilitarianism", namely the idea that there philosophy will be somehow better for everyone if followed and if you pull on that string you can very likely reduce the philosophy into a form of rules based utilitarianism. Utilitarianism can subsume practically any other system into itself, people don't want there pet ideology being an aspect of a bigger ideology.
2) It reduces morality to a math problem, and people don't seem to want to accept the idea that all choice is really just a bunch of complex math problems. So morality is one of the places they more firmly sit on "This clearly isn't math."
3) Utilitarianism, is, hard. It judges based on consequence which means there's no easy answer of "Will this be right or wrong." you literally can't know until after the fact(And since that action causes other actions.....), it also means everyone literally everyone fails to live up to it, and most people don't want to accept they aren't just and moral people.

Behold my bias, assuming not being a utilitarian is born of your bias.

The comments to this entry are closed.

Less Wrong (sister site)

May 2009

Sun Mon Tue Wed Thu Fri Sat
          1 2
3 4 5 6 7 8 9
10 11 12 13 14 15 16
17 18 19 20 21 22 23
24 25 26 27 28 29 30