“Utilons” isn’t quite the right word: utilons are all I purchase. My utility function is a sum of components: I can decompose it into a local part to do with my happiness and the happiness of those close to me (and thus status, warm fuzzies and the like) and a global part to do with things like the lives of strangers and the future of humanity. I try to strongly mark the boundary between those two, so I don’t for example value the lives of people in the same country as me more than those in different countries.
You’re saying I can more optimally spend resources on efforts that clearly serve one or the other than on efforts that try to do both and do neither well, and I agree, I’d just phrase it differently: purchase big-picture utility and small-picture utility separately.
Perhaps Eliezer doesn’t directly value fuzzies or status, so when he is purchasing them he isn’t purchasing utilons directly. Rather, he is purchasing motivation to continue doing things which directly purchase utilons. In other words, he doesn’t really want fuzzies, but if he doesn’t buy any, he’ll lose his motivation to be altruistic altogether. So he buys fuzzies to keep his motivation up which allows him to keep directly purchasing utilons—the things he does actually value. That’s at least how I read it.
“Utilons” isn’t quite the right word: utilons are all I purchase.
I think Eliezer is using “utilon” to refer to the unit of value in utilitarianism (i.e., the theory of aggregating value linearly across individuals) whereas what you’re talking about is probably the unit of value in expected utility maximization (i.e., the theory of aggregating value linearly across possible worlds). To avoid confusion, I propose that we call the latter “utils” (which as dreeves pointed out is already standard for this usage). In other words, let’s use “utilons” when talking about utilitarianism, and “utils” when talking about expected utility maximization.
Would you agree that we’re all (in this thread) drawing the same distinction and just labelling it differently?
I agree that we should switch to the standard term “utils” here, but I think I wouldn’t go for also using “utilons” in the way you propose; enough people would continue to use the words in a different way that we wouldn’t succeed in ironing out the confusion. I’d prefer something like “global utils” and “fuzzy utils/fuzzies” which can’t be taken to mean something else.
It’s more about what you use the word “utilon” for; I see “utilon” as the measure of whatever consequences I value, but EY is using it here to refer to utility without weighting for proximity to the speaker, which I call “big picture utility” above.
“Utilons” isn’t quite the right word: utilons are all I purchase. My utility function is a sum of components: I can decompose it into a local part to do with my happiness and the happiness of those close to me (and thus status, warm fuzzies and the like) and a global part to do with things like the lives of strangers and the future of humanity. I try to strongly mark the boundary between those two, so I don’t for example value the lives of people in the same country as me more than those in different countries.
You’re saying I can more optimally spend resources on efforts that clearly serve one or the other than on efforts that try to do both and do neither well, and I agree, I’d just phrase it differently: purchase big-picture utility and small-picture utility separately.
Perhaps Eliezer doesn’t directly value fuzzies or status, so when he is purchasing them he isn’t purchasing utilons directly. Rather, he is purchasing motivation to continue doing things which directly purchase utilons. In other words, he doesn’t really want fuzzies, but if he doesn’t buy any, he’ll lose his motivation to be altruistic altogether. So he buys fuzzies to keep his motivation up which allows him to keep directly purchasing utilons—the things he does actually value. That’s at least how I read it.
edit: clarity
Sum of components, product, or more complex functions.
Just a sum, I think; my understanding of utility is that it’s part of the definition that it makes sense to sum it.
I think Eliezer is using “utilon” to refer to the unit of value in utilitarianism (i.e., the theory of aggregating value linearly across individuals) whereas what you’re talking about is probably the unit of value in expected utility maximization (i.e., the theory of aggregating value linearly across possible worlds). To avoid confusion, I propose that we call the latter “utils” (which as dreeves pointed out is already standard for this usage). In other words, let’s use “utilons” when talking about utilitarianism, and “utils” when talking about expected utility maximization.
Would you agree that we’re all (in this thread) drawing the same distinction and just labelling it differently?
I agree that we should switch to the standard term “utils” here, but I think I wouldn’t go for also using “utilons” in the way you propose; enough people would continue to use the words in a different way that we wouldn’t succeed in ironing out the confusion. I’d prefer something like “global utils” and “fuzzy utils/fuzzies” which can’t be taken to mean something else.
So you see warm fuzzies, status boost, and societal good as subtypes of the utilon output of altruistic activities? Interesting.
It’s more about what you use the word “utilon” for; I see “utilon” as the measure of whatever consequences I value, but EY is using it here to refer to utility without weighting for proximity to the speaker, which I call “big picture utility” above.