Because of this, the dicussion in the direction of “Further, it is intuitively much harder for ordinary people (i.e. non-mathematicians) to report how happy/satisfied they feel on a logarithmic scale than a linear one.” is mostly irrelevant—because the log() or other nonlinear transformation transformation is happening in the percepting stage, it is mostly hidden from the people themseves. (You can compare the arguments given with something like asking people to rate the star brightness on logarithmic scale)
On a positive note, I agree with 15. Having the right non-linear transformation would also probably explain away part of the results where people appear inconsistent or aggregating in wrong way. (ala Kahneman)(Their brains may be aggregating correctly the raw, untransformed quantity.)
Conclusion: IMO the possibility of nonlinear perception of happiness is a crucial consideration for this whole business and should be given research priority. I’m probably quite bad at explaining it (eg the above LW post which was probably incomprehensible for most readers)
I tend to agree with you that it is still unclear whether the relationship between ‘raw’ happiness (as mentioned in your post) and perceived happiness is linear or not. Log makes intuitive sense in some ways but the implication that happiness increases linearly with wealth is extremely counterintuitive to me.
I tend think something akin to prospect theory’s weighting happiness relative to some set point may be plausible. I also think there will be some odd effects near the boundaries (0&10).
For me the project we’re trying to push forward with this manifesto is in large part about trying to figure out satisfactory answers to questions like these, so I agree it’s a priority.
Despite these open questions I tend to think it’s useful to proceed as if it were linear for now as that seems to be status quo in the field and to me it seems more intuitive. Even if we’re wrong about linearity this measurement system should lead to more accurate approximations of ‘raw’ happiness than, say, QALYs
I don’t think I follow your post. One idea is that happiness functions like sound, where it takes twice as much of an increase in sound to cause the same increase in perceived loudness. This seems confused if we extend it to happiness. What are we supposed to say: it takes twice as much of increase in happiness to cause the same increase in perceived happiness? That would be crazy, because we only have one item, happiness, rather than two. If we labelled this on a graph, the axis would be the same.
An alternative is that it takes twice as much of an increase in brain activity (of a certain kind) to cause the same increase in perceived happiness. Okay. But people are reported their perceived happiness (or life satisfaction), not the brain states.
Imagine, for example, there exist some “S1 happiness” which is actually important for humans, and they are making decisions based on it, and some “S2 perception of happiness” which people report. People are incorrectly assuming (a) this is the same, and (b) you have good introspective access to it.
For more direct comparison than sound, imagine there isn’t such thing as money, and you ask people to report their percieved wealth on a scale 1 to 10. Now introduce money, and compare how the 1..10 rating relate to monetary wealth. I would bet you would get nonlinear relationship in this case.
Contrary to the post, I think non-linear reporting is quite likely. The problem is not so much in “reporting” but in “perception”. More about it here:
https://www.lesswrong.com/posts/7Kv5cik4JWoayHYPD/nonlinear-perception-of-happiness#TMPkeCnPe85zqdZqn
Because of this, the dicussion in the direction of “Further, it is intuitively much harder for ordinary people (i.e. non-mathematicians) to report how happy/satisfied they feel on a logarithmic scale than a linear one.” is mostly irrelevant—because the log() or other nonlinear transformation transformation is happening in the percepting stage, it is mostly hidden from the people themseves. (You can compare the arguments given with something like asking people to rate the star brightness on logarithmic scale)
On a positive note, I agree with 15. Having the right non-linear transformation would also probably explain away part of the results where people appear inconsistent or aggregating in wrong way. (ala Kahneman)(Their brains may be aggregating correctly the raw, untransformed quantity.)
Conclusion: IMO the possibility of nonlinear perception of happiness is a crucial consideration for this whole business and should be given research priority. I’m probably quite bad at explaining it (eg the above LW post which was probably incomprehensible for most readers)
I tend to agree with you that it is still unclear whether the relationship between ‘raw’ happiness (as mentioned in your post) and perceived happiness is linear or not. Log makes intuitive sense in some ways but the implication that happiness increases linearly with wealth is extremely counterintuitive to me.
I tend think something akin to prospect theory’s weighting happiness relative to some set point may be plausible. I also think there will be some odd effects near the boundaries (0&10).
For me the project we’re trying to push forward with this manifesto is in large part about trying to figure out satisfactory answers to questions like these, so I agree it’s a priority.
Despite these open questions I tend to think it’s useful to proceed as if it were linear for now as that seems to be status quo in the field and to me it seems more intuitive. Even if we’re wrong about linearity this measurement system should lead to more accurate approximations of ‘raw’ happiness than, say, QALYs
Hello Jan.
I don’t think I follow your post. One idea is that happiness functions like sound, where it takes twice as much of an increase in sound to cause the same increase in perceived loudness. This seems confused if we extend it to happiness. What are we supposed to say: it takes twice as much of increase in happiness to cause the same increase in perceived happiness? That would be crazy, because we only have one item, happiness, rather than two. If we labelled this on a graph, the axis would be the same.
An alternative is that it takes twice as much of an increase in brain activity (of a certain kind) to cause the same increase in perceived happiness. Okay. But people are reported their perceived happiness (or life satisfaction), not the brain states.
Imagine, for example, there exist some “S1 happiness” which is actually important for humans, and they are making decisions based on it, and some “S2 perception of happiness” which people report. People are incorrectly assuming (a) this is the same, and (b) you have good introspective access to it.
For more direct comparison than sound, imagine there isn’t such thing as money, and you ask people to report their percieved wealth on a scale 1 to 10. Now introduce money, and compare how the 1..10 rating relate to monetary wealth. I would bet you would get nonlinear relationship in this case.