Some years ago I got into a series of exchanges with a group self-identified as Bleeding Heart Libertarians (hereafter BHL), starting on a Cato Unbound forum, continued on our blogs.
I agree with your post, though I think there are further problems with the "original position." It seems to me the thought experiment involves asking "what kind of society would you like if you didn't know anything about yourself?" to which there is no answer because if I don't know anything about myself, I don't have any preferences and so have no basis to make the choice. Rawls simply posits certain universal values ad hoc (like "freedom") which we're supposed to hold "behind the veil of ignorance" which amounts to begging the question. More generally, any aggregation of utility functions involves a *value judgement*, it is not an objectively well-defined operation. Pretending utility functions can be objectively aggregated amounts to assuming that there's really only one individual, and so it's no surprise all social questions are "solved" -- they're solved because the problem has been assumed away.
I'm trying to imagine what Rawls may have been thinking (based on your summary, not having actually read Rawls myself), and I can come up with a few reasons that one might use the utility of the poorest member of society rather than the "average-utility" member of society as a stand-in for the desirability of the whole society.
1) The person asking the question (e.g. Rawls) and the person being asked (with the education and leisure to read about economic philosophy) are probably both well above "average utility" in the current society. And people's expectations about utility are anchored to their present situations. So the unattractive prospect of being in a worse state in the hypothetical society probably outweighs the attractive prospect of being in a better state in that hypothetical society. This doesn't imply using the literal poorest member of society as a representative, but it does imply a bias below the mean.
2) Most societies in human history have been pyramidal, with a very few people near the top and an enormous number near the bottom. In such a society, "the poorest member of society" is a reasonable approximation of the average-utility member, and a whole lot easier to compute. Particularly since utility is easier to compare than to add, subtract, multiply, or divide.
1. Are you saying that Rawls and his reader are misjudging what utilities of various incomes are, and the rule somehow compensates for that? Utility already takes account of the declining marginal utility of income.
In any case, Rawls claim is that any rational person in the initial situation will make his choice, so it should not depend on the person's income.
2. That does not, however, describe the distribution in modern societies, which is the context Rawls is writing in.
Well, part of it is about declining marginal utility (of not only income, but housing, food, health, transportation, living relatives ... for almost _any_ objective measure of well-being, people tend to fear loss more than they anticipate gain, especially if they're reasonably comfortable like Rawls and his readers). And yes, "utility" technically already accounts for that.
But I was thinking more of the fact that people -- no matter how rational and educated -- tend to project their own feelings and experiences onto others. Rawls and most of his readers would be worse off than they presently are if reborn at random into (say) their own society; losing some utility is more likely than gaining some utility, so it's entirely rational for them to pay more attention to possible losses than possible gains. And entirely understandable, if not "rational" in the economist's sense, to assume that any rational person would do the same.
Most "modern" societies, in "developed" nations, are indeed less pyramidal than most societies in human history, but they still have a median (income, wealth, whatever) substantially below their mean. For example, themotleyfool.com says "The national average U.S. income in 2021 was $97,962. The median U.S. income in 2021 was $69,717", a ratio of almost 3:2. Wealth paints a more dramatic picture: businessinsider.com says "The average American family has a $748,000 net worth, according to Federal Reserve data. But the median net worth is $121,700", a ratio of over 6.
Which brings us back to the question of arithmetic operations on utility. Finding a median only requires comparison, so the median-utility member of a society is reasonably well-defined (if we ignore the fact that different people have different preferences between the same alternatives). But economically "rational" behavior tends to be defined in terms of expected value, which is a (weighted) mean, which requires addition, multiplication, and division, which are not so obviously meaningful for utility. (Similarly, talking about "declining marginal utility" requires comparing not two absolute utilities, but two utility _deltas_, which is more difficult and abstract.)
In particular, to judge a society's desirability rationally by the utility of being born into it at random requires a weighted mean of utilities, which is dubiously meaningful. Looking at median utility instead is easier and more meaningful, and (in almost all societies) produces a substantially lower score. Which wouldn't matter if all societies were equally pyramidal -- it would merely apply a constant factor, not changing the relative desirability of societies -- but they're not, so it does matter. One could argue that the "mean utility" approach is overly kind to pyramidal societies, and "median utility" corrects this error. "Minimum utility" errs in the opposite direction, overly penalizing pyramidal societies, but it's even easier to find and visualize than median utility.
In addition, I suppose "minimum utility" is an easy sell socially and politically because of its support from such well-known economists as Jesus of Nazareth: “whatsoever you do to the least of your brothers and sisters, you do unto me."
I don't find median utility plausible, for the obvious reason. Given a choice between two gambles, one of which has equal probabilities of 10, 20, 30 and one of 18,19, 100 I think I would obviously prefer the latter.
The response might be that either would be better than 1,1,200, which has a higher average. But the definition of VN utility already takes declining marginal utility into account, meaning that the 200 is 200 precisely because it is such an attractive outcome that 1/3 chance of it would be worth giving up a lot of certain utility for. If there are no outcomes that attractive, that just means that there are no outcomes with utility 200.
I think you are blurring the distinction between declining marginal utility and something like the endowment effect, valuing what you have much more than what you could have because you have it.
[Pausing to go read about the VN-M utility theorem, which largely but not entirely addresses my concerns about using operations other than comparison on utility. Need to ponder that some more.]
Yes, declining marginal utility is a different phenomenon from "anchoring" to present conditions, the observation that people's utility curves vary with their current circumstances in consistent, predictable ways. I was thinking mostly of the latter phenomenon. IIRC, utility curves tend to be steeper near the current state than farther from it, and they tend to be steeper to the immediate left (potential small loss) than to the immediate right (potential small gain).
A reader considering Rawls's thought experiment is in a particular state in the current society, and will inevitably make judgments based on that state, the current society, and hypothetical societies reasonably similar to it. If such readers are mostly selected to be above the median (or even the mean) of their current society, they'll see more downside than upside in being reborn at random into a society, and therefore weigh the low outcomes more heavily, and the high outcomes more lightly, than their objective probabilities would imply. As a result, even if the reader were trying to use a "mean utility" criterion, the reader's estimate of mean utility would be lower than the actual mean utility in that society.
All reasonable, but it doesn't lead to Rawls' result, merely to a modified version of Harsanyi's, with the modification arguably a mistake that should be ignored.
I'm still waiting for someone to provide a defense of Rawls. It seems unlikely that one of the most prominent political philosophers of the past century was writing, and got famous for writing, complete nonsense. If true, it has implications not only for modern academic political philosophy but for the modern academic world. And the implications go farther back than the current ideological takeover of large parts of the academy.
Great post. I was a regular reader of BHL back in the day. My post mortem for the site is that the comments and rebuttals were often better, more insightful and more nuanced than most of the original posts.
I have not read Ayn Rand's books, but have seen her lush Italian melodramatic movie of a personable heroine escaping soviet Russia. I remember the bewhiskered Old Bolshevik talking of how he stuffed middle-class people in a railroad boiler, knowing that was his future. Good stuff in a Gone With The Wind sort of way.
The hero of her books looked like a Bad Baronet doing his job, raising fires of wrath in the personable heroine's heart, fires that later reach her bosom and special places. Don't have the plumbing to feel this, but a fellow who sees a young lady reading Ayn Rand novels, best case moving her thighs a little, my Brother, go say hello.
I think it would make more sense to evaluate Rawls verses the median utility instead of mean. I think that's more relevant to what Rawls was saying, and also gets us away from a potentially really bad society (by most of our estimations) where one person has infinity utility and everyone else has zero. Or, more realistically, where a very small aristocracy has essentially all the wealth and the vast majority of people are in danger of starvation or other miserable conditions.
Another thought about Rawls' position is if he's looking at median instead of mean, and asking if the median level is an acceptable level of utility in that society. If the median falls below a certain threshold (however that society judges such things), then it may be worse than just looking at the total utility of that society. You can also look at the very bottom and determine that a society should not tolerate the bottom being below some subjective level of experience.
Scott at ACX has written about a theoretical society in which everyone is always super happy, but one child is tortured forever to make this society work (it's a strained hypothetical, but helps us if we intuitively reject such a society as unacceptably unjust). In a more realistic society, maybe the bottom 10% live grueling lives of constant struggle so that the 90% above them live comfortably. We may reject such a society even if the average or median utility is higher than an alternative that doesn't torture the bottom X percent.
I think Rawls makes more sense with these thoughts in mind, rather than the assumption that we always have to consider the bottom X percent no matter how well off they are. Rawls seems to be assuming a society where the bottom exists below an acceptable level, such that even if the remaining population is doing well we may still reject such an arrangement. If the bottom 10% are also comfortable, or at least above some minimally acceptable level, then Rawls arguments make much less sense - as you have outlined.
The problem for Rawls and those who agree with him, is to define what that "unacceptable" level would be. I think this may be an impossible task, or at least a constantly changing definition as you noted in your first section where the standards of 1700 would imply misery to those living in 2000. My solution is to use the standards of our long ago ancestors as much as possible, and thereby increase my own internal sense of wellbeing (I'm doing great by that standard!). The alternative that many seem to select is to compare themselves to those with the most, and therefore make themselves feel worse for the comparison, since they then feel they lack what others have.
Your small aristocracy society would be unlikely to have high average utility, due to declining marginal utility of income. Von Neumann utility is defined in such a way that someone choosing a gamble prefers the gamble with the highest average utility, so in the initial situation someone with an equal probability for all lives would necessarily choose the option with the highest average utility.
The hypothetical you are talking about is a short story by Ursula LeGuin, "Those Who Walk Away from Omelas."
Rawls quite explicitly says to maximize the welfare of the worst off person, so I'm not sure of the relevance of your points which imagine him making a different argument than he did.
"it came on the backs of millions (again, including millions of slaves)." If that's the case, how is it that slave owning societies are consistently poorer than non-slave owning societies? The American South was poorer than the American North, the Arab and African worlds remained poor despite extensive slave trading, Britain banned the slave trade in 1807 and it was followed by a century of rapid growth etc. What you're claiming is simply a myth that is asserted and never supported with evidence.
Interesting post! I had the same reaction reading Rawls.
Good to know that Harsanyi came up with the "original position" 20 years earlier. I'll aim to cite him instead, in the future. I suspect that Rawls' very visual phrase "veil of ignorance" is one big reason why he gets the public credit for that concept -- a testament to the power of branding/catchphrases.
My ungenerous guess is that the reason for Rawls' success is that he was offering arguments for things that many academic philosophers wanted to believe. My view, subject to someone showing me that his argument is more defensible than it appears to be, is that that is evidence against modern political philosophy deserving to be taken seriously.
I never got any substantive response from the BHL people on the Rawls issue, still don't know if they agree that his argument is nonsense but are unwilling to say so or really believe, correctly or otherwise, that it is defensible if not correct.
I agree with your post, though I think there are further problems with the "original position." It seems to me the thought experiment involves asking "what kind of society would you like if you didn't know anything about yourself?" to which there is no answer because if I don't know anything about myself, I don't have any preferences and so have no basis to make the choice. Rawls simply posits certain universal values ad hoc (like "freedom") which we're supposed to hold "behind the veil of ignorance" which amounts to begging the question. More generally, any aggregation of utility functions involves a *value judgement*, it is not an objectively well-defined operation. Pretending utility functions can be objectively aggregated amounts to assuming that there's really only one individual, and so it's no surprise all social questions are "solved" -- they're solved because the problem has been assumed away.
I'm trying to imagine what Rawls may have been thinking (based on your summary, not having actually read Rawls myself), and I can come up with a few reasons that one might use the utility of the poorest member of society rather than the "average-utility" member of society as a stand-in for the desirability of the whole society.
1) The person asking the question (e.g. Rawls) and the person being asked (with the education and leisure to read about economic philosophy) are probably both well above "average utility" in the current society. And people's expectations about utility are anchored to their present situations. So the unattractive prospect of being in a worse state in the hypothetical society probably outweighs the attractive prospect of being in a better state in that hypothetical society. This doesn't imply using the literal poorest member of society as a representative, but it does imply a bias below the mean.
2) Most societies in human history have been pyramidal, with a very few people near the top and an enormous number near the bottom. In such a society, "the poorest member of society" is a reasonable approximation of the average-utility member, and a whole lot easier to compute. Particularly since utility is easier to compare than to add, subtract, multiply, or divide.
1. Are you saying that Rawls and his reader are misjudging what utilities of various incomes are, and the rule somehow compensates for that? Utility already takes account of the declining marginal utility of income.
In any case, Rawls claim is that any rational person in the initial situation will make his choice, so it should not depend on the person's income.
2. That does not, however, describe the distribution in modern societies, which is the context Rawls is writing in.
Well, part of it is about declining marginal utility (of not only income, but housing, food, health, transportation, living relatives ... for almost _any_ objective measure of well-being, people tend to fear loss more than they anticipate gain, especially if they're reasonably comfortable like Rawls and his readers). And yes, "utility" technically already accounts for that.
But I was thinking more of the fact that people -- no matter how rational and educated -- tend to project their own feelings and experiences onto others. Rawls and most of his readers would be worse off than they presently are if reborn at random into (say) their own society; losing some utility is more likely than gaining some utility, so it's entirely rational for them to pay more attention to possible losses than possible gains. And entirely understandable, if not "rational" in the economist's sense, to assume that any rational person would do the same.
Most "modern" societies, in "developed" nations, are indeed less pyramidal than most societies in human history, but they still have a median (income, wealth, whatever) substantially below their mean. For example, themotleyfool.com says "The national average U.S. income in 2021 was $97,962. The median U.S. income in 2021 was $69,717", a ratio of almost 3:2. Wealth paints a more dramatic picture: businessinsider.com says "The average American family has a $748,000 net worth, according to Federal Reserve data. But the median net worth is $121,700", a ratio of over 6.
Which brings us back to the question of arithmetic operations on utility. Finding a median only requires comparison, so the median-utility member of a society is reasonably well-defined (if we ignore the fact that different people have different preferences between the same alternatives). But economically "rational" behavior tends to be defined in terms of expected value, which is a (weighted) mean, which requires addition, multiplication, and division, which are not so obviously meaningful for utility. (Similarly, talking about "declining marginal utility" requires comparing not two absolute utilities, but two utility _deltas_, which is more difficult and abstract.)
In particular, to judge a society's desirability rationally by the utility of being born into it at random requires a weighted mean of utilities, which is dubiously meaningful. Looking at median utility instead is easier and more meaningful, and (in almost all societies) produces a substantially lower score. Which wouldn't matter if all societies were equally pyramidal -- it would merely apply a constant factor, not changing the relative desirability of societies -- but they're not, so it does matter. One could argue that the "mean utility" approach is overly kind to pyramidal societies, and "median utility" corrects this error. "Minimum utility" errs in the opposite direction, overly penalizing pyramidal societies, but it's even easier to find and visualize than median utility.
In addition, I suppose "minimum utility" is an easy sell socially and politically because of its support from such well-known economists as Jesus of Nazareth: “whatsoever you do to the least of your brothers and sisters, you do unto me."
I don't find median utility plausible, for the obvious reason. Given a choice between two gambles, one of which has equal probabilities of 10, 20, 30 and one of 18,19, 100 I think I would obviously prefer the latter.
The response might be that either would be better than 1,1,200, which has a higher average. But the definition of VN utility already takes declining marginal utility into account, meaning that the 200 is 200 precisely because it is such an attractive outcome that 1/3 chance of it would be worth giving up a lot of certain utility for. If there are no outcomes that attractive, that just means that there are no outcomes with utility 200.
I think you are blurring the distinction between declining marginal utility and something like the endowment effect, valuing what you have much more than what you could have because you have it.
[Pausing to go read about the VN-M utility theorem, which largely but not entirely addresses my concerns about using operations other than comparison on utility. Need to ponder that some more.]
Yes, declining marginal utility is a different phenomenon from "anchoring" to present conditions, the observation that people's utility curves vary with their current circumstances in consistent, predictable ways. I was thinking mostly of the latter phenomenon. IIRC, utility curves tend to be steeper near the current state than farther from it, and they tend to be steeper to the immediate left (potential small loss) than to the immediate right (potential small gain).
A reader considering Rawls's thought experiment is in a particular state in the current society, and will inevitably make judgments based on that state, the current society, and hypothetical societies reasonably similar to it. If such readers are mostly selected to be above the median (or even the mean) of their current society, they'll see more downside than upside in being reborn at random into a society, and therefore weigh the low outcomes more heavily, and the high outcomes more lightly, than their objective probabilities would imply. As a result, even if the reader were trying to use a "mean utility" criterion, the reader's estimate of mean utility would be lower than the actual mean utility in that society.
All reasonable, but it doesn't lead to Rawls' result, merely to a modified version of Harsanyi's, with the modification arguably a mistake that should be ignored.
I'm still waiting for someone to provide a defense of Rawls. It seems unlikely that one of the most prominent political philosophers of the past century was writing, and got famous for writing, complete nonsense. If true, it has implications not only for modern academic political philosophy but for the modern academic world. And the implications go farther back than the current ideological takeover of large parts of the academy.
Great post. I was a regular reader of BHL back in the day. My post mortem for the site is that the comments and rebuttals were often better, more insightful and more nuanced than most of the original posts.
I have not read Ayn Rand's books, but have seen her lush Italian melodramatic movie of a personable heroine escaping soviet Russia. I remember the bewhiskered Old Bolshevik talking of how he stuffed middle-class people in a railroad boiler, knowing that was his future. Good stuff in a Gone With The Wind sort of way.
The hero of her books looked like a Bad Baronet doing his job, raising fires of wrath in the personable heroine's heart, fires that later reach her bosom and special places. Don't have the plumbing to feel this, but a fellow who sees a young lady reading Ayn Rand novels, best case moving her thighs a little, my Brother, go say hello.
I think it would make more sense to evaluate Rawls verses the median utility instead of mean. I think that's more relevant to what Rawls was saying, and also gets us away from a potentially really bad society (by most of our estimations) where one person has infinity utility and everyone else has zero. Or, more realistically, where a very small aristocracy has essentially all the wealth and the vast majority of people are in danger of starvation or other miserable conditions.
Another thought about Rawls' position is if he's looking at median instead of mean, and asking if the median level is an acceptable level of utility in that society. If the median falls below a certain threshold (however that society judges such things), then it may be worse than just looking at the total utility of that society. You can also look at the very bottom and determine that a society should not tolerate the bottom being below some subjective level of experience.
Scott at ACX has written about a theoretical society in which everyone is always super happy, but one child is tortured forever to make this society work (it's a strained hypothetical, but helps us if we intuitively reject such a society as unacceptably unjust). In a more realistic society, maybe the bottom 10% live grueling lives of constant struggle so that the 90% above them live comfortably. We may reject such a society even if the average or median utility is higher than an alternative that doesn't torture the bottom X percent.
I think Rawls makes more sense with these thoughts in mind, rather than the assumption that we always have to consider the bottom X percent no matter how well off they are. Rawls seems to be assuming a society where the bottom exists below an acceptable level, such that even if the remaining population is doing well we may still reject such an arrangement. If the bottom 10% are also comfortable, or at least above some minimally acceptable level, then Rawls arguments make much less sense - as you have outlined.
The problem for Rawls and those who agree with him, is to define what that "unacceptable" level would be. I think this may be an impossible task, or at least a constantly changing definition as you noted in your first section where the standards of 1700 would imply misery to those living in 2000. My solution is to use the standards of our long ago ancestors as much as possible, and thereby increase my own internal sense of wellbeing (I'm doing great by that standard!). The alternative that many seem to select is to compare themselves to those with the most, and therefore make themselves feel worse for the comparison, since they then feel they lack what others have.
Your small aristocracy society would be unlikely to have high average utility, due to declining marginal utility of income. Von Neumann utility is defined in such a way that someone choosing a gamble prefers the gamble with the highest average utility, so in the initial situation someone with an equal probability for all lives would necessarily choose the option with the highest average utility.
The hypothetical you are talking about is a short story by Ursula LeGuin, "Those Who Walk Away from Omelas."
Rawls quite explicitly says to maximize the welfare of the worst off person, so I'm not sure of the relevance of your points which imagine him making a different argument than he did.
>could never see why anyone took it seriously
https://www.mattball.org/2022/11/libertarian-or-grown-up.html
And I have not, each time you post this, been able to see that why you think an insult substitutes for an argument.
"it came on the backs of millions (again, including millions of slaves)." If that's the case, how is it that slave owning societies are consistently poorer than non-slave owning societies? The American South was poorer than the American North, the Arab and African worlds remained poor despite extensive slave trading, Britain banned the slave trade in 1807 and it was followed by a century of rapid growth etc. What you're claiming is simply a myth that is asserted and never supported with evidence.
Interesting post! I had the same reaction reading Rawls.
Good to know that Harsanyi came up with the "original position" 20 years earlier. I'll aim to cite him instead, in the future. I suspect that Rawls' very visual phrase "veil of ignorance" is one big reason why he gets the public credit for that concept -- a testament to the power of branding/catchphrases.
My ungenerous guess is that the reason for Rawls' success is that he was offering arguments for things that many academic philosophers wanted to believe. My view, subject to someone showing me that his argument is more defensible than it appears to be, is that that is evidence against modern political philosophy deserving to be taken seriously.
I never got any substantive response from the BHL people on the Rawls issue, still don't know if they agree that his argument is nonsense but are unwilling to say so or really believe, correctly or otherwise, that it is defensible if not correct.