Hanson says regarding national metrics: 'Well, now we’re going to authorize this same sort of agency to estimate number like GDP except we’re going to tell them to put more things in the number. Bills before Congress would say, “Count more trees, and count leisure, and count international reputation.” They would just make a bigger formula that included all the stuff they cared about in this measure of national welfare.'
Here's the tricky part. It seems to me that the metrics which you don't choose to bundle into the aggregate measurement will get annihilated (optimized out) at the expense of those which you do choose to bundle. And the decisions as to which things you do bundle and with what weight represent ethical, moral, spiritual, aesthetic, and otherwise intangible judgments which are (unsurprisingly) difficult to quantify or even come to basic agreement on. He weasels a bit by using relatively uncontroversial examples ("more trees", "leisure") and by handwaving ("authorize": how?, "sort of agency": what sort?), but it's trivial to imagine metrics which produce wild disagreement regarding the magnitude or even sign of the weight to be applied.
(Shadows of paperclip maximization loom.)
At that point, I'm not sure where we've significantly improved things since we're still left with the problem of how to choose the people who decide which metrics go into the aggregate measure and at what weight. Can we make a market on that, too? At that point we're in some crazy recursion and I get lost. I'm deeply skeptical.
Here's the tricky part. It seems to me that the metrics which you don't choose to bundle into the aggregate measurement will get annihilated (optimized out) at the expense of those which you do choose to bundle. And the decisions as to which things you do bundle and with what weight represent ethical, moral, spiritual, aesthetic, and otherwise intangible judgments which are (unsurprisingly) difficult to quantify or even come to basic agreement on. He weasels a bit by using relatively uncontroversial examples ("more trees", "leisure") and by handwaving ("authorize": how?, "sort of agency": what sort?), but it's trivial to imagine metrics which produce wild disagreement regarding the magnitude or even sign of the weight to be applied.
(Shadows of paperclip maximization loom.)
At that point, I'm not sure where we've significantly improved things since we're still left with the problem of how to choose the people who decide which metrics go into the aggregate measure and at what weight. Can we make a market on that, too? At that point we're in some crazy recursion and I get lost. I'm deeply skeptical.