The main ways to get a handle on this are to use subjective well-being scores (which is what those graphs do, and is somewhat questionable as to whether it’s a natural unit), or to ask people about trade-offs or gambles they’d make (to elicit preferences as in a vN-M utility function). Both approaches lead to data saying it’s approximately logarithmic, and there are also some theoretical reasons to think this is roughly right.
The main ways to get a handle on this are to use subjective well-being scores (which is what those graphs do, and is somewhat questionable as to whether it’s a natural unit), or to ask people about trade-offs or gambles they’d make (to elicit preferences as in a vN-M utility function). Both approaches lead to data saying it’s approximately logarithmic, and there are also some theoretical reasons to think this is roughly right.