I’ve had a thought about a possible replacement for ‘hyperbolic discounting’ of future gains: What if, instead of using a simple time-series, the discount used a metric based on how similar your future self is to your present-self? As your future self develops different interests and goals, your present goals would tend to be less fulfilled the further your future self changed; and thus the less invested you would be in helping that future iteration achieve its goals.
Given a minimal level of identification for ‘completely different people’, then this could even be expanded to ems who can make copies of themselves, and edit those copies, to provide a more coherent set of values about which future selves to value more than others.
(I’m going to guess that Robin Hanson has already come up with this idea, and either worked out all its details or thoroughly debunked it, but I haven’t come across any references to that. I wonder if I should start reading that draft /before/ I finish my current long-term project...)
DataPacRat’s comment together with your observation strike me as the most interesting thing I’ve seen in an open thread in a while. I’m not convinced that the idea is in any sense correct or even a good idea but the originality is striking. It is easy to come up with obviously wrong ideas that are original, but coming up with an original (I think) idea that is that plausible is more impressive, and your observation makes it more striking.
Shane Frederick had the idea that hyperbolic discounting might be because people identify less with their future self. He actually wrote his dissertation on this topic, using Parfit’s theory of personal identity (based on psychological continuity & connectedness). He did a few empirical studies to test it, but I think the results weren’t all that consistent with his predictions and he moved on to other research topics.
Thank you /very/ much for that link. The first two sections do a much better job explaining the general background and existing thought around my thought than I’d be able to write on my own.
I am, however, less confident that the study described in the third section does a very good job of disproving the correlation between amount of selfhood and future discounting. Among other reasons, the paper itself posits that most people likely subscribe to the “simple” theory of identity instead of the “complex” one under discussion.
As a third thought, reading the paper has suggested a new variation of my original thought. Perhaps the correlation exists, but I have causation backwards: future discounting could be, in fact, an expression of how much people consider their future selves to be dissimilar to their present selves. At present, I’m not sure what it would take to figure out which version of this idea comes closer to being true, and that’s even assuming that the correlation exists in the first place; but it seems worth further consideration, at least.
Sub-part one: Ask participants various questions to determine the minimum value of x for them to agree to the former option in, “Would you rather we give $100+x to a perfect stranger, or $100 to you?”. (Initial prediction: values will vary widely, from particularly generous people with an x of -$99.99, to particularly selfish people with an x of infinity.)
Sub-part two: Ask participants various questions to determine the minimum value of y for them to agree to the former option in, “Would you rather we give $100+y to you in 5 years, or $100 to you now?”.
Initial prediction: x and y will be closely correlated; the more a person is willing to give money to perfect strangers, the more they’ll be willing to give money to their future selves.
Possible variation: Change ‘perfect stranger’ in sub-part one to people with varying levels of closeness to the participant: distant acquaintance, close friend, family member.
Possible variation: Change ‘5 years’ in sub-part two to different time-scales.
If both variations are included: Then, possibly, the data may converge into a simple shape. Possible complications for that shape may arise from how likely the participant feels they’re likely to still be alive in n years; and in how strongly they trust the experimenters to actually distribute the money.
Additional complications: I am completely unaffiliated with any university or other educational institution, have never performed a psychological experiment, and have no budget to perform any such experiment.
I’ve had a thought about a possible replacement for ‘hyperbolic discounting’ of future gains: What if, instead of using a simple time-series, the discount used a metric based on how similar your future self is to your present-self? As your future self develops different interests and goals, your present goals would tend to be less fulfilled the further your future self changed; and thus the less invested you would be in helping that future iteration achieve its goals.
Given a minimal level of identification for ‘completely different people’, then this could even be expanded to ems who can make copies of themselves, and edit those copies, to provide a more coherent set of values about which future selves to value more than others.
(I’m going to guess that Robin Hanson has already come up with this idea, and either worked out all its details or thoroughly debunked it, but I haven’t come across any references to that. I wonder if I should start reading that draft /before/ I finish my current long-term project...)
Consistent with why children care so little about their future.
DataPacRat’s comment together with your observation strike me as the most interesting thing I’ve seen in an open thread in a while. I’m not convinced that the idea is in any sense correct or even a good idea but the originality is striking. It is easy to come up with obviously wrong ideas that are original, but coming up with an original (I think) idea that is that plausible is more impressive, and your observation makes it more striking.
This echoes my thoughts.
Shane Frederick had the idea that hyperbolic discounting might be because people identify less with their future self. He actually wrote his dissertation on this topic, using Parfit’s theory of personal identity (based on psychological continuity & connectedness). He did a few empirical studies to test it, but I think the results weren’t all that consistent with his predictions and he moved on to other research topics.
Thank you /very/ much for that link. The first two sections do a much better job explaining the general background and existing thought around my thought than I’d be able to write on my own.
I am, however, less confident that the study described in the third section does a very good job of disproving the correlation between amount of selfhood and future discounting. Among other reasons, the paper itself posits that most people likely subscribe to the “simple” theory of identity instead of the “complex” one under discussion.
As a third thought, reading the paper has suggested a new variation of my original thought. Perhaps the correlation exists, but I have causation backwards: future discounting could be, in fact, an expression of how much people consider their future selves to be dissimilar to their present selves. At present, I’m not sure what it would take to figure out which version of this idea comes closer to being true, and that’s even assuming that the correlation exists in the first place; but it seems worth further consideration, at least.
An idea for an experiment:
Sub-part one: Ask participants various questions to determine the minimum value of x for them to agree to the former option in, “Would you rather we give $100+x to a perfect stranger, or $100 to you?”. (Initial prediction: values will vary widely, from particularly generous people with an x of -$99.99, to particularly selfish people with an x of infinity.)
Sub-part two: Ask participants various questions to determine the minimum value of y for them to agree to the former option in, “Would you rather we give $100+y to you in 5 years, or $100 to you now?”.
Initial prediction: x and y will be closely correlated; the more a person is willing to give money to perfect strangers, the more they’ll be willing to give money to their future selves.
Possible variation: Change ‘perfect stranger’ in sub-part one to people with varying levels of closeness to the participant: distant acquaintance, close friend, family member.
Possible variation: Change ‘5 years’ in sub-part two to different time-scales.
If both variations are included: Then, possibly, the data may converge into a simple shape. Possible complications for that shape may arise from how likely the participant feels they’re likely to still be alive in n years; and in how strongly they trust the experimenters to actually distribute the money.
Additional complications: I am completely unaffiliated with any university or other educational institution, have never performed a psychological experiment, and have no budget to perform any such experiment.
Interesting to think that empathy for others might be the same mechanism by which we make long term decisions.
Closely related to this whole discussion is Self-empathy as a source of “willpower”.
How sure are you that the outcome would be as predicted? My Y is about 10, but my X is undefined because I’d rather they didn’t get it.
That smells like an actual insight to me, so please do work it up into a full theory if no-one comes up with an example of how it’s been done already.