No Soul Suggests IUCs

Glen Whitman has responded to our interpersonal utility comparison debate in a way that I found unconvincing but illuminating. Early on, he said:

Making interpersonal utility comparisons is a kind of category error, akin to trying to add degrees and pounds. I don’t have a logical proof of this point; it just seems intuitively right to me. You can’t add my utility to yours, because my utility and your utility are just different sorts of thing. And appealing to objective measures, like dopamine levels or endorphins in the brain, seems to me to miss the point. We’ve got different brains, and an endorphin in my brain just ain’t the same as an endorphin in yours.


People make IUCs, but they make lots of other mistakes, too. They say things like, “Mary is fatter than John is tall,” a statement that makes less and less sense the more your scrutinize it.

I take issue with "Mary is fatter than John is tall" and "trying to add degrees and pounds" as analogies for IUC. Remember that the thing we are trying to measure in IUC's is the same concept (utility), just applied to two different people. To assert that this makes it two separate concepts is to assume an answer to the question under debate.

A more appropriate analogy is that it's like asking whether Mary is as funny as John is funny. Funniness is not externally measurable, and different people see it differently. Actually, this analogy isn't quite fair, because we can get around the subjectivity by asking lots of people to compare M & J's funniness intrapersonally. So let's say instead that it's like asking whether Mary is as funny to Brian as she is to Susan.

Now we have a good analogy - we're asking about the same concept, but in two different people's brains, just like utility. Substitute "Eating an apple" for "Mary" and "enjoyable" (or, if you want to be anal about ordinality, "ranked as highly among preferences") for "funny" and we have an IUC. Note how different this is than the examples Glen presented. Most people will intuitively agree that you can't add degrees and pounds, but it is far from clear that we can't figure out whether Brian or Susan finds more humour in Mary's antics.

But this is just quibbling over examples, I have a much deeper problem with the argument that different brains are just plain different, and there ain't no way around that. As an atheist and a computer scientist, I firmly believe that my brain is just a computer made out of meat. A computer with some different properties than our silicon versions - less reliable memory, more redundancy, parallel processing, small numbers of complex operations rather than large numbers of simple ones, and a distressing tendency to shut down processing when given visual input like this - but a computer nonetheless. I don't believe in a soul or a "ghost in the machine", or any other kind of metaphysical glue imbuing my meat with a spirit.

Which, uncomfortably but inexorably, suggests at least the theoretical possibility that my brain could be modeled inside some other computer. That it could be experimented on, analyzed, its preferences probed and understood. As could yours. As could anyone's. And while this doesn't conclusively prove that if we thoroughly understood two brains, we could figure out how to compare their utility, their happiness, their preferences, doesn't it seem awfully likely?

I can see why those who believe in the spirit glue might see the utility of two separate souls to be two unique things, ineffable and thus incomparable. But for those of us with more prosaic leanings, what reason is there to doubt that we could theoretically understand and compare the patterns produced by electrical signals and connections inside our meat-filled skulls?

Related posts:

Why IUCs?
Cardinal Schmardinal, Ordinal Schmordinal
Encoding Happiness
IUCs and the Law of Large Numbers
Love and Intrapersonal Utility Comparison
What color does a submarine weigh? (True or False?)
Exploding IUCs on the roadside
Interpersonal Utility Comparisons
Pareto Efficiency and Justice
Can the Paradox of the Non-Comparability of Interpersonal Utility be Resolved?

Share this

And while this doesn’t

And while this doesn’t conclusively prove that if we thoroughly understood two brains, we could figure out how to compare their utility, their happiness, their preferences, doesn’t it seem awfully likely?

No, not in the slightest. Suppose that we had a machine that made choices when confronted with options. Even if we knew everything about that machine down to the smallest atom, it would not follow that we would know anything about any magnitude associated with the choices.

How might this machine be implemented? Suppose there was a vast lookup table encoded in the machine, with one entry for each choice. Even if we knew the composition of that lookup table down to the smallest atom, we would quite easily fail to find anything that gave us a magnitude of preference.

I am not saying we encode lookup tables. I am simply coming up with a clear example to demonstrate the point.

I have a post coming out on

I have a post coming out on Tuesday that demonstrates how you can do IUC even with ordinal utility. Ordinality is not a barrier, only comparison.

Constant, Let's say we


Let's say we programmed the machine with a very large table of ordinal preferences over sets of outcomes, which each differ *very* slightly from their neighbors. You could then count the number of ordinal rankings between any two compared outcomes, which certainly looks like a good candidate for magnitude of preference to me. Moreso the more orderings we have -- the discrete approximates the continuous at large N. We don't have any difficulty measuring the length of a nanowire just because it's made of discrete atoms.

Oh snap, didn't see that

Oh snap, didn't see that comment. Hope I just didn't step on your toes, Patri!

You could then count the

You could then count the number of ordinal rankings between any two compared outcomes, which certainly looks like a good candidate for magnitude of preference to me.

Of course you can do that, and I addressed this point several days ago (nobody to my knowledge had made it, I simply anticipated it and answered it). Sadly, I don't know how to search through back comments. Essentially I acknowledged that of course you can turn ordinals into cardinals. It's very easy if they're finite: just take the bottom one and call it "one unit", the second one call "two units" and so on. And so you can make cardinals out of ordinals. No biggie.

However, it's obviously just pulled out of your ---. And the obvious response is, what if we modify the lookup table to insert a whole bunch of new choices at the bottom, or middle, or top, or wherever we like. You can increase the magnitude of any different in utility by adding more choices in between those two choices you want to separate. Or, conversely, if you want to decrease the difference between them, just pull out every other choice between them to halve their distance. If you want to decrease it again, pull out every other choice of what remains.

By the way, having a large number of ordinals isn't necessary to apply the technique that turns ordinals into cardinals.

Aside from pulling it out of your ---, having a really large, even an infinite number of ordinals doesn't give you cardinality, even if your ordinals have the order of the rationals - i.e., even if you can map your ordinal set to the rational while preserving order. Because, suppose you've mapped your ordinals to the rationals via a one to one correspondence and then mapped the cardinality of the rationals back to your ordinal set through this one to one correspondence. Someone else could come along and re-map your ordinals to the rationals with very different magnitudes - for example, everything you mapped from zero to one, he could map from zero to a million, and everything you mapped from one to a million, he could map from a million to a million plus one. That would give him very different magnitudes.

I have a post coming out on

I have a post coming out on Tuesday that demonstrates how you can do IUC even with ordinal utility. Ordinality is not a barrier, only comparison.

Actually, cardinality by itself does not buy you comparison. Weight is cardinal and length is cardinal, but you can't compare weight with length (unless you want to do the mean-variance thing that you mentioned earlier). Similarly, my utility might be cardinal and your utility might be cardinal but that by itself does not allow a comparison to be made.

The real issue is comparison.

You can do self-utility comparisons. Suppose someone prefers A to B to C to D and you want to know whether he prefers A to B more than he prefers C to D. You can say to your subject, "You will wake up tomorrow, and there will be a 50/50 chance that you have B or D. But I will give you a choice right now: you can choose to upgrade your situation, either from B to A or from D to C, but not both. If you choose to upgrade from D to C, then when you wake up tomorrow there will be a 50/50 chance that you will have B or C. If you choose to upgrade from B to A, then when you wake up tomorrow there will be a 50/50 chance that you will have A or D."

If the subject prefers to upgrade from B to A, then the subject prefers a 50/50 chance of choosing B over A to a 50/50 chance of choosing C over D. And so we can say that the magnitude of his preference for A over B is greater than the magnitude of his preference for C over D.

That doesn't quite give it cardinality. But it does allow a comparison. It allows you to compare any pair of preferences to see which preference is greater. It does so by allowing a person to decide which of two preferences to have a chance to exercise.

However, it does not allow comparisons across people.

I thought of the following: imagine you have a 50/50 chance of waking up as one person or another, with the corresponding schedule of preferences.

But I don't think that thought experiment works. Think of werewolves. A werewolf might say, prior to his transformation, "I'm about to change, so chain me up." Obviously such a person chooses to override his wolf-half preferences, and similarly someone imagining himself into another person might simply choose to override that other person's preferences. But there are other issues as well, such as that the meta-preference is still the subject's own meta-preference over the preferences of the two people he could wake up as. It's not really an objective balancing of the two preferences.

"A distressing tendency to

"A distressing tendency to shut down processing when given visual input like this"

Well then your brain is different than mine. My mind doesn't so much shut down as it seems to spin out of control with it's own agendas. To make this extra disturbing, I'm in my late forties.

Perhaps the burka isn't such a bad idea after all. It will keep me from having a heart attack when I'm eighty so I can live into my ninties. What am I saying? Just kidding, I didn't really mean that. Go Lebannon! Go Bolivia!

For some reason I really like the Lebanese flag right now. Anyone else have that reaction to those photos? I mean in addition to other reactions. Is that really an attractive flag or is it that perhaps advertisers are on to something.

Yeah but is John as funny to

Yeah but is John as funny to Sarah, as Ice Cream is tasty to Joe? That's I think counts as a category error. Both may be purchased with money so you might think them comparable by that measure. That doesn't work either however because prices also depend on supply, and supply may depend odn preferences. John may not being on stage and cream may be in short supply.

There are other factors that disturb this notion of using money as a proxy for utility. Does my utility for ice cream go down just because I make more money? If you really believe that a poor person gets more utility out of a sandwich than Bill Gates does then perhaps diners should be charging hobos more. :)

Bill Gates comes in an has to pay ten cents for the sandwich because that is about his level of enjoyment of it while the hobo has to pay a quarter days wages to match his level of enjoyment.

I find the idea of IUCs very intuitively appealing. After all it is clear to me that a starving guy is going to enjoy the sandwich more than the well fed one. I can do that experiment myself. I know that the sandwich is just not appealing when I have just finished a full course meal. I don't offer my guests a sandwich after they have just finished thanksgiving dinner and lean back saying, "Boy, I'm full".

One problem I have with IUCs is that there are certain circumstances where they are clearly valid but the vast majority of cases are unclear or obviously inapplicable.

Another is when IUCs are used to build theories of justice. Is it just to take something from one person just because another would enjoy it more? I can say definately not. My example with the nun and pirates on a island makes that clear.

I am not one of those Libertarians that think rights are totally sancrosanct. I think that certain rights can be violated in cases of emergency. I would not however build this upon the idea of IUCs. I think it perfectly reasonable for a lost and starving hiker to break into my cabin for food and shelter. He is violating my property rights but not for reasons that violate the principles those rights are based on. He isn't trying to cheat me and is not doing this to get a competitive advantage over me. Everyone makes mistakes and it is possible I might also be in such a situation someday.

Furthermore if he leaves a note offering to repay me for goods used and damages done then I don't see any problem. The fact that I did not give explicit permission does not matter.

The way I get to this position is because I believe that people honor the concept of rights for what are ultimately self serving purposes. I don't steal from you in the expectation that you will do the reverse and we can both capture the productivity increase this understanding produces. It's quite clear that the hiker is not going to be able to participate in any of this enhanced productivity if he is dead. Since he is no longer capable, by no fault of his own, to collect on this bargain then it's hard to expect him to live up to his end.

In fact, I'll go further than what I said before. Even if I had posted a sign on the door saying, "I explicitly want starving hikers to keep the F__ out", it would be within the hikers rights to ignore the sign. For the prior reasons given, plus the fact that all men are fallible and may find themselves in such a position. Even if I was there with a shotgun and refused him help then gunned him down, I would be in the wrong. Why? Well unless I was in danger of starving myself at that point there is no reason he could not compensate me via the increased production that would occur with him being alive.

In other words, I cannot assert my property rights if respecting those rights is based the grounds of reciprocity when the basis of that reciprocity is dissolved. As an example of this, in racially discrimatory societies if the privileged can own property and the non-privileged cannot then it is an unreasonable expectation to expect those discrimanated against to honor those property rights with any conviction. Why should they?

I've not fully expressed all the implications here but it also extends into my "baby in the bucket" issue.

These are some of the issues I was contemplating when I said that libertarianism has holes.

Matt/Constant - my argument

Matt/Constant - my argument for IUC's is different. Basically, with ordinal preferencs, we can make ordinal IUC's like "I will like an apple more than you will like a banana". ordinal individual preferences lead to ordinal combined preferences, not to the impossibility of combined preferences. and for the purposes IUC's are used ("If I do X, which I like, and you do Y, which you don't, are we on net better of?") this turns out to be good enough.

more on the 4th.

Oops, didn't see Constant's

Oops, didn't see Constant's later post. Yes, I completely agree with you that the issues is comparison, not cardinal/ordinal - that is the point of my next post. I sort of jumped the gun and posted this one about why I think comparison is possible. But its a weird topic and it's fine to disagree on that. I just get annoyed when people talk about cardinal/ordinal because mathematically it turns out not to matter very much. What matters is whether you can compare at all.