Actual utility is an agent-relative experience. It’s not comparable across parties since it’s entirely subjective. Although there are certain extremes that wouldn’t fall into anyone’s utility function, society can’t make public policy based on subjective states. Pursuit of utility is within the domain of the individual and their close personal relationships.
This argument instead makes the case for maximizing certain objective factors, or primary goods, that tend to put people in a better position to maximize their own subjective well-being. These should be increased regardless of one’s subjective preference for them.
No one would agree to maximizing someone else’s happiness because who knows what makes them happy. But people would agree to maximizing a collective pot of generally useful resources.
On interpersonal utility comparisons, consider gift giving. Gifts other than, for example, cash transfer payments are often derided by economists are producing deadweight loss. A great gift, however, is one which signals that the gift-giver has a refined sense of the gift-getter's preferences. Particularly given uncertainty and agency costs, it easy to tell a story in which this signal would be more valuable in absolute terms than the deadweight loss of the gift.
The best point is the risk idea. But I am inclined to push back.
My initial, perhaps quite biased and motivated-reasoning response, is that it is classic “the world has to be this way, because I can’t think of a model that doesn’t work like that.” It is very handy to analyze circumstances in a particular way, but this handiness doesn’t mean that this is what everyone consciously does. If we do it unconsciously, that is an interesting fact about evolution. If we don’t actually do it consistently, consciously or otherwise, what does the model tell us? Is it positive or normative?
If it is positive, supposedly we can test it and possibly discover it is wrong. (The behavioral economists think they have done so, though maybe they are mistaken.)
If it is normative, we are back where we started. Having a model of how people ought to respond to risk would say nothing about whether utility is interpersonally comparable; it would say only that we have to treat circumstances that way to use this neat model.
The risk model is very appropriate when the subject knows all the variables and they are all cardinal values that can be estimated or measured. The fact that different circumstances would invalidate the model does not mean that circumstances cannot be like that.
The risk model assumes we are at an extreme end of the explore/exploit continuum, where the only unknowns are the actual values of random variables. And utility ideally would be one of those known or estimable quantities. But that is just looking for the keys where the light is best. How do we know that is what we are looking at, rather than what we wish to see? It doesn’t convince me that utility is measurable as a cardinal quantity or interpersonally comparable in a determinate way.
Assuming you are a benevolent dictator, or even a economic advisor with a reasonable degree of certainty that your advice will be implemented. Would you try to compensate the losers of a economic improvement, say through transfers in wealth or a reallocation of rights. Assuming the deadweight loss from taxation and such was minimal. From what I can tell this seems to be something Robin Hanson would advise.
It's always been obvious to me that utility is interpersonally comparable. Setting a person on fire is worse than poking someone else with a stick is.
Actual utility is an agent-relative experience. It’s not comparable across parties since it’s entirely subjective. Although there are certain extremes that wouldn’t fall into anyone’s utility function, society can’t make public policy based on subjective states. Pursuit of utility is within the domain of the individual and their close personal relationships.
This argument instead makes the case for maximizing certain objective factors, or primary goods, that tend to put people in a better position to maximize their own subjective well-being. These should be increased regardless of one’s subjective preference for them.
No one would agree to maximizing someone else’s happiness because who knows what makes them happy. But people would agree to maximizing a collective pot of generally useful resources.
On interpersonal utility comparisons, consider gift giving. Gifts other than, for example, cash transfer payments are often derided by economists are producing deadweight loss. A great gift, however, is one which signals that the gift-giver has a refined sense of the gift-getter's preferences. Particularly given uncertainty and agency costs, it easy to tell a story in which this signal would be more valuable in absolute terms than the deadweight loss of the gift.
The best point is the risk idea. But I am inclined to push back.
My initial, perhaps quite biased and motivated-reasoning response, is that it is classic “the world has to be this way, because I can’t think of a model that doesn’t work like that.” It is very handy to analyze circumstances in a particular way, but this handiness doesn’t mean that this is what everyone consciously does. If we do it unconsciously, that is an interesting fact about evolution. If we don’t actually do it consistently, consciously or otherwise, what does the model tell us? Is it positive or normative?
If it is positive, supposedly we can test it and possibly discover it is wrong. (The behavioral economists think they have done so, though maybe they are mistaken.)
If it is normative, we are back where we started. Having a model of how people ought to respond to risk would say nothing about whether utility is interpersonally comparable; it would say only that we have to treat circumstances that way to use this neat model.
The risk model is very appropriate when the subject knows all the variables and they are all cardinal values that can be estimated or measured. The fact that different circumstances would invalidate the model does not mean that circumstances cannot be like that.
The risk model assumes we are at an extreme end of the explore/exploit continuum, where the only unknowns are the actual values of random variables. And utility ideally would be one of those known or estimable quantities. But that is just looking for the keys where the light is best. How do we know that is what we are looking at, rather than what we wish to see? It doesn’t convince me that utility is measurable as a cardinal quantity or interpersonally comparable in a determinate way.
Assuming you are a benevolent dictator, or even a economic advisor with a reasonable degree of certainty that your advice will be implemented. Would you try to compensate the losers of a economic improvement, say through transfers in wealth or a reallocation of rights. Assuming the deadweight loss from taxation and such was minimal. From what I can tell this seems to be something Robin Hanson would advise.