Bryan Caplan  

Are Grotesque Hypotheticals Cheap Shots?

PRINT
Hayek Interviewed... Short Papers on the Financial ...
In my opening statement for last night's Caplan-Hanson Debate, I relied heavily on a couple of grostesque hypotheticals:

...Robin endorses an endless list of bizarre moral claims.  For example, he recently told me that "the main problem" with the Holocaust was that there weren't enough Nazis!  After all, if there had been six trillion Nazis willing to pay $1 each to make the Holocaust happen, and a mere six million Jews willing to pay $100,000 each to prevent it, the Holocaust would have generated $5.4 trillion worth of consumers surplus.

Let's consider another example.  Suppose the only people in the world are Hannibal the millionaire, a slave trader, and 10,000 penniless orphan slaves.  The slave trader has no direct use for his slaves, but likes money; Hannibal, on the other hand, is a ravenous cannibal.  According to Robin, the "optimal outcome" is for Hannibal to get all 10,000 orphans and eat them.

If you've ever taken a class in moral philosophy, you've probably heard weird examples like these before.  Normally, these examples lead people to back away from their pet moral theories.  Robin's devotion to efficiency is so strong, however, that he will bite any bullet you present.  The most he'll say in his own defense is that he is "Merely serving as an advisor to help people get what they want," but the fact remains that faced with the preceding examples, he'd advise genocide and cannibalism.

I suspect that many attendees saw these examples as "cheap shots."  But when I pressed Robin, he predictably bit both bullets.  His main objection was just that my examples aren't very relevant to actual policy analysis.  What's the point of bringing up genocide and cannibalism?

My response: Weird hypotheticals are philosophers' equivalent of controlled experiments.  When a scientist wants to test a physical theory, he sets up weird laboratory conditions that make it easy to find an exception to the theory.  Similarly, when a philosopher wants to test a moral theory, he sets up weird examples that make it easy to find an exception to the theory.  Philosopher Michael Huemer explains it best:
How, one might ask, can we draw conclusions about how the world really is from purely hypothetical premises, i.e., premises about some imagined but not actual situations?

This objection is vaguely felt by many who object to thought experiments in philosophy in general, but it is a logical error. You can validly deduce a categorical proposition from hypothetical premises. For example:

A -> (B -> C).

B -> not-C.

Therefore, not-A.

is a valid form of inference, where the "->" stands for the "if ... then" relation (i.e. "If x were true, y would be true") ...And this form of inference is relevant to the way hypotheticals are used in philosophy to test moral principles. The typical form of thought-experiment-based arguments in moral philosophy is as follows: "If moral theory T were true, then in situation S, it would be right to do A. But in situation S, it would surely be wrong to do A. Therefore, T is false." Notice that this form of argument is perfectly valid: the conclusion deductively follows from its premises (it's a variant on modus tollens). Notice also that both premises are hypothetical - i.e., both are about what would be right if so-and-so were the case. But the conclusion is categorical. (emphasis mine)
When I point out that Robin's theory implies that the rightness of the Holocaust depends on the number of Nazis, then, I'm not trying to smear him as a Nazi sympathizer.  I'm saying that his theory has crazy implications, so his theory is false and he should revise it.  End of story.

P.S. Here's my complete debate archive.


Comments and Sharing





TRACKBACKS (1 to date)
TrackBack URL: http://econlog.econlib.org/mt/mt-tb.cgi/1732
The author at Brad Taylor's Blog in a related article titled Wise Words writes:
    From Bryan Caplan: Weird hypotheticals are philosophers’ equivalent of controlled experiments.  When a scientist wants to test a physical theory, he sets up weird laboratory conditions that make it easy to find an exception to the theory.  Simila... [Tracked on April 15, 2009 7:43 PM]
COMMENTS (28 to date)
Greg Ransom writes:

Kuhn and Wittgenstein and others point out that we learn generalizations from particular instances, from particular examples, from particular instance of training using particular examples. It often then takes further training to learn where are "generalization" applies and where it doesn't.

It looks like Hansen is overlooking this core insight of psychology, neuroscience, and philosophy of the last 100 years.

Neal W. writes:

Interesting that you bring this up as I have been thinking about it lately. Something I've noticed is that people untrained in philosophy do not except weird hypotheticals as valid reasoning even if deductively valid.

dWj writes:

I think people have trouble with the basic logic of existence proofs and so in general, especially when they involve Nazis. I'm not clear why this confuses people as much as it does.

Zac writes:

For the record, I cast my vote for "liberty" before and after the debate.

From talking to others in attendance, the impression was that Robin did not "bite the bullets" on the grotesque hypotheticals but rather deflected them. Grotesque hypotheticals, I think, are simply not persuasive in debate if they are too abstract and fanciful compared to real world issues. They are a good starting point, but you must move directly into a relevant policy issue where one model is superior to the other. Bryan did bring up the argument that censoring atheists is likely (or at least possible) to be an efficient policy, but is clearly a wrong policy. Unfortunately, discussion of that issue was cut short. Bryan would have served his argument better by providing more moderate examples where Robin's model would fail to provide sensible "advice."

That said I still think Bryan had the stronger argument. Robin provided no examples (during the debate) of situations where he would recommend overriding the "liberty heuristic" and advise a clearly more efficient but liberty-limiting policy. My view is that there are few if any examples of this in the real world, and when there is conflict, it probably makes sense to recommend against efficiency.

Otoh I think it makes sense to focus on efficiency analysis. One danger of the liberty heuristic is that you may be too quick to denounce a policy as liberty-limiting when you should first denounce it as inefficient. If you cannot provide an efficiency argument for why your position is correct, you will convince few people. For example: few people agree that taxation is theft. It is, but that's quite beside the point. The best way to convince someone that we should reduce or eliminate taxation is not to talk about taxation as theft. As economists, we can show that taxation is inefficient in addition to being immoral. As Bryan said during the during the debate, economists can show that we should not pursue limity-liberting policies for "less than no reason at all."

Philo writes:

Huemer's example is *not valid*. If '->' stands for the material conditional, let A be true and B false; then the premises are true and the conclusion false. Even if '->' stands for implication (entailment), just add that A and B are incompatible; again, true premises and a false conclusion.

ao writes:

Was this debate recorded or is there a transcript?

Bryan Caplan writes:
Philo writes:

Huemer's example is *not valid*. If '->' stands for the material conditional, let A be true and B false; then the premises are true and the conclusion false. Even if '->' stands for implication (entailment), just add that A and B are incompatible; again, true premises and a false conclusion.


Actually, Huemer has a statement on exactly this point that I omitted in the ellipses. Unedited:
the "->" stands for the "if ... then" relation (i.e. "If x were true, y would be true") (N.B. not the so-called 'material conditional' of first-order formal logic).
Please don't blame Huemer for my omissions. :-)
Ian Dunois writes:

AO,
It was recorded. Not sure when it will be posted on the web. We'll have to wait to hear from the GMU Econ Society.
http://gmueconsociety.blogspot.com

During the Q&A period, a GMU undergraduate, Dallas Stauffer, had brought up the notion that all value is subjective.

As value is subjective, any advice given would be skewered with the advice givers own value rather than the clients. This does not deny the use of efficiency within making our arguments, only that efficiency can not be the sole support for the argument.

Ignoto Fiorentino writes:

I agree that Hanson's moral theory is unpersuasive [I don't think you want to say "false"], but disagree that the best way to show it is with weird hypotheticals.

With regard to Huemer's argument, either you or he or both are misapplying it. Weird hypotheticals can be used to show that a theory is illogical or internally inconsistent, but you are not claiming that Hanson's argument is inconsistent, only that it is normatively unpersuasive in the real world. And weird hypotheticals do not work to show this [as Kuhn and Wittgenstein would agree].

Specifically: what is the basis for your assumption that your own moral theory extends to the crazy worlds of your hypotheticals, worlds in which you have no experiential intution?

Jeremy, Alabama writes:

If your "weird hypotheticals" on genocide and cannibalism are controlled experiments, then they are like a thought experiment of the inside of a Black Hole - they are behind the event horizon so you can't observe them. And if you do, you get crushed to death.

David Friedman writes:

I didn't hear the debate, but the issue you are raising is one I have often discussed, both in class and in writing and Robin's position on your second hypothetical, as you describe it, is wrong.

The problem is that economic efficiency--ultimately derived from Marshall's discussion--measures utility by willingness to pay. There are lots of good reasons to do that, as I have discussed elsewhere. But it means that "improve efficiency" doesn't correspond to "increase total utility," since there is no reason to assume that everyone has the same marginal utility of income. An example like yours leverages that problem by comparing a dollar measured gain to a rich person with low marginal utility of income to a dollar measured loss to poor people with high marginal utility of income.

Tom Crispin writes:

In discussing consumer surplus, Nazis, Jews, and the Holocaust, aren't there other parties involved, if only as obervers? After all, in Robin's example, if there are 24 trillion non-Nazis willing to pay 25 cents each to prevent the Holocaust, only one Jew needs to object to make it inefficient!

eccdogg writes:

Bryan, when I first read your post I agreed totally but the more I thought about it I can't say that Robin is wrong.

I think the values you have for the willingness to pay for someones life are incredibly low.

Wouldn't a Jew or Orphan be wiling to pay an almost infinite amount for their life. So under a more realistic assumtion would there be any efficiency increasing transaction of the type you mention?

Granted if you give the Nazi or Slavers property right over their lives, they may not have the ability to pay. But isn't that different than what they are willing to pay? And isn't the willing to pay standard the one that matters.

T Student writes:

The problem with the Nazi hypothetical is that you are ignoring who owns the Jews' lives. If they own their own lives and value them as a whole above what the Nazis are willing to pay, then the Holocaust is not generating surplus.

The framing you have above implies that the Jews' lives are owned by some sort of third party and that the Jews are in a bidding war with the Nazis.

Zac writes:

@eccdogg, "willingness to pay" presupposes ability to pay. You cannot be willing to pay money that you do not have. If you think about it for a moment, you should conclude that this is all "willingness to pay" really can mean, and that "hypothetical willingness" is nonsensical.

Tim Fowler writes:

Zac, hypothetical willingness to pay (or more accurately willingness to forgo payment) is relevant in light of T Student's point. Consider (properly in my opinion) the Jew's to own their own lives. Then the Nazi's would have to have a willingness to pay that exceeds the Jew's willingness to forgo payment (which is essentially their hypothetical willingness to pay).

Bryan Caplan writes:
T Student writes:

The problem with the Nazi hypothetical is that you are ignoring who owns the Jews' lives. If they own their own lives and value them as a whole above what the Nazis are willing to pay, then the Holocaust is not generating surplus.

The framing you have above implies that the Jews' lives are owned by some sort of third party and that the Jews are in a bidding war with the Nazis.

Right. But as Robin explicitly admits, efficiency analysis begins with a status quo - any status quo - and then assessing changes by summing willingness to pay. So he has no reason to disqualify my hypothetical.
Bryan Caplan writes:
David Friedman writes:

I didn't hear the debate, but the issue you are raising is one I have often discussed, both in class and in writing and Robin's position on your second hypothetical, as you describe it, is wrong.

The problem is that economic efficiency--ultimately derived from Marshall's discussion--measures utility by willingness to pay. There are lots of good reasons to do that, as I have discussed elsewhere. But it means that "improve efficiency" doesn't correspond to "increase total utility," since there is no reason to assume that everyone has the same marginal utility of income. An example like yours leverages that problem by comparing a dollar measured gain to a rich person with low marginal utility of income to a dollar measured loss to poor people with high marginal utility of income.

I'm not sure if I'm understanding you, David. Robin explicitly defends economic efficiency as opposed to utilitarianism, so the cannibalism deduction follows, doesn't it?
Phil writes:

It seems to me that the problem is the use of "willingness to pay" (which presumes the ability to pay). That may work pretty well normally -- for almost any consumer good, my willingness to pay is less than my ability to pay (especially if I can borrow). But, at the extremes, willingness to pay can be more than ability to pay, as in my life.

If Bill Gates is willing to pay $1 billion to kill me, I would certainly be willing to pay $1 billion plus $1 to stop him. Unfortunately, that's beyond my ability to pay, even if I sell myself into slavery.

I'm not an economist, but my understanding is that willingness to pay $1 billion plus $1 doesn't qualify as "willing to pay" if the willingness is not backed up by the means, and I think that's what's causing the problem. Isn't revealed preference the evidence of "willingness?" How can I reveal my preference when I don't have the means to do so?

Garrett Harmon writes:

I would have loved to have been there!

David Jinkins writes:

Phil: I haven't thought deeply about this, but I think the issue is that if willingness to pay isn't constrained by some sort of budget, then I could be willing to pay an infinite amount for whatever I wanted. We need some sort of criterion to assess the relative importance of goods to people.

BlackSheep writes:

Phil: you can only assess of how much people value something by seeing what else they're willing to give up to get it.
As Friedman explained it, it only makes sense to use money as a value meter when you're evaluating a diverse population, because they should value money similarly. In the case of Hannibal and the orphans, this assumption doesn't hold. It would make sense for the atheist-vs-theist censorship scenario mentioned by Zac.

US writes:

I wonder if Hanson would reach the same conclusions if it was his own life that was on the line in the hypothetical.

Let's try a little game: Right at this moment, I would be willing to pay 2 dollars to the person who kidnapped and sold Robin Hanson to a nazi slavetrader cannibal. Anyone else feel like pitching in? Let's presume the government applies Hansonian reasoning, so participating in the specific transaction is perfectly legal.

Phil writes:

David Jinkins/BlackSheep: Agreed that *what people actually do pay* is the correct of willingness to pay in almost all circumstances. I'm just arguing that it breaks down in this particular case, and for obvious reasons.

T Student writes:

Bryan Caplan writes: Right. But as Robin explicitly admits, efficiency analysis begins with a status quo - any status quo - and then assessing changes by summing willingness to pay. So he has no reason to disqualify my hypothetical.


So provided that this is made clear I don't see the problem with accepting the hypothetical outcome. It's as though I advocate building a bridge strong enough to support the weight that would be applied to it when full of the heaviest cars possible and then you come along and say "But what if gravity increased by tenfold?"

Robin Hanson writes:

I think my position has been misrepresented. I endorse the efficiency criteria as part of large deals to consistently use the criterion across many topics for a long time, in which case it should approach a Pareto improvement in the limit. I do not claim this matches moral intuitions or improves total utility in each case, taken separately.

John Sabotta writes:

Here's the bottom line. Robin Hanson is simply an evil person. So are most of the Hu-Man Ro-bots over at that center of pestilence known as "Overcoming Bias".

And the rest of you here are fools who think that evil is just a quaint, outdated, superstitious notion. No. Evil is real, and playing games with evil leads to moral corruption and real-world consequences of the most horrific kind.

I used to think nerds were just disgusting and stupid. My opinion of them is rapidly changing for the worse. It's no longer just innocent stupidity that characterizes your average nerd these days but something genuinely malign.

Wei Dai writes:

Robin, do you have a reference for the "should approach a Pareto improvement in the limit" claim? What I've found so far seem to contradict it. From http://cepa.newschool.edu/het/essays/paretian/paretosocial.htm:

The second defense, pursued by Hicks (1941), was that even if the losers do not get compensated in the move, they might still benefit in the "long-run" if the criteria were followed consistently by society. This argument is similar to that of "trickle-down" theory and in arguments for free trade: some people may be worse off in the short-run, but in the long run, everyone will be better off. The underlying assumption, of course, is that at some point, those who lost utility initially will come across a possible move in which they benefit and a society which follows the Kaldorian rule will move to it and thus they will gain in the end. Of course, as Little (1950) notes, this is completely hypothetical. There is nothing to guarantee that there will eventually be a move in which the initial losers will be the ultimate winners.
(end quote)

Intuitively, the space of Kaldor-Hicks improvements is so much larger than the space of Pareto improvements (see this graph and imagine an n-dimensional space with n=population), that it seems very unlikely that a random sequence of Kaldor-Hicks improvements will happen to constitute a Pareto improvement.

Comments for this entry have been closed
Return to top