Of course hyperbolic discounting is a useful heuristic. The paradigm I subscribe to is not just bias. We have these heuristics because they’re the best that evolution or our developing minds could do. That is, they’re pretty good in some other environment (ancestral or childhood), which might be very different. You singled out hyperbolic discounting, among all the biases, but it seems to me much more likely to be maladapted to the present than the other standard biases.
Most of your comment argues that it’s a good heuristic, but your first paragraph (“bite the bullet and accept hyperbolic discounting”) seems to make a stronger claim.
Usually the points of time where you can “change your mind” correspond to events which introduce all the new kinds of risks and transaction costs and are not neutral.
That is a different heuristic than I would call hyperbolic discounting. You can certainly produce situations in the lab where people apply a worse heuristic than that. I expect the two heuristics were more similar in EEA than today.
If you want to make a quick decision, go with your gut and trust hyperbolic discounting. But trust it for a decision on an action, not its intermediate output of utility. It mixes up probability and utility. “If you’re building complex interconnected structures of beliefs” then you have to separate the two and you can’t trust your gut model of yourself because of hyperbolic discounting. People screw up long term planning all the time because of hyperbolic discounting.
By biting the bullet I meant using hyperbolic time as the first default approximation, instead of exponential time. I think exponential time is usually much more wrong in practice than hyperbolic time.
Can you give a concrete example of someone screwing up due to hyperbolic accounting in a case where there’s an objective measure of utility to compare the person’s estimates against ?
There are no objective measures of utility. But just about everyone who has failed a diet or exercise schedule could be seen as failing beause of hyperbolic discounting.
I don’t know. What I was referring to was that people’s estimates of their future utility of some course of action are not constant. And they often vary in such a way that one choice (dieting, exercising, saving...) appear rational when you are planning for it, and when you evaluate it in retrospect, but is unappealing at the time that you actually do it.
Think about our evolutionary history. Presumably, life was less stable, deals less predictable than they are today. In that case it would have been better to have a strong hyperbolic discount rate, while now, when outcomes are increasingly reliable, then that rate should be dropped but it (presumably) hasn’t.
Of course, our intuitive discount rate should never reach the exponential that a model would predict, because there are always new unforeseen factors, but I would contend that the uncertainties have dropped substantially. This would make the particular hyperbolic rate that we intuitively discount payoffs at today biased, while in our evolutionary past it presumably would have been a better approximation of a suitable discount rate.
Of course hyperbolic discounting is a useful heuristic. The paradigm I subscribe to is not just bias. We have these heuristics because they’re the best that evolution or our developing minds could do. That is, they’re pretty good in some other environment (ancestral or childhood), which might be very different. You singled out hyperbolic discounting, among all the biases, but it seems to me much more likely to be maladapted to the present than the other standard biases.
Most of your comment argues that it’s a good heuristic, but your first paragraph (“bite the bullet and accept hyperbolic discounting”) seems to make a stronger claim.
That is a different heuristic than I would call hyperbolic discounting. You can certainly produce situations in the lab where people apply a worse heuristic than that. I expect the two heuristics were more similar in EEA than today.
If you want to make a quick decision, go with your gut and trust hyperbolic discounting. But trust it for a decision on an action, not its intermediate output of utility. It mixes up probability and utility. “If you’re building complex interconnected structures of beliefs” then you have to separate the two and you can’t trust your gut model of yourself because of hyperbolic discounting. People screw up long term planning all the time because of hyperbolic discounting.
By biting the bullet I meant using hyperbolic time as the first default approximation, instead of exponential time. I think exponential time is usually much more wrong in practice than hyperbolic time.
Can you give a concrete example of someone screwing up due to hyperbolic accounting in a case where there’s an objective measure of utility to compare the person’s estimates against ?
There are no objective measures of utility. But just about everyone who has failed a diet or exercise schedule could be seen as failing beause of hyperbolic discounting.
Without objective measures of utility, what could it even mean to speak of someone’s utility judgements as being biased or wrong ?
I don’t know. What I was referring to was that people’s estimates of their future utility of some course of action are not constant. And they often vary in such a way that one choice (dieting, exercising, saving...) appear rational when you are planning for it, and when you evaluate it in retrospect, but is unappealing at the time that you actually do it.
Think about our evolutionary history. Presumably, life was less stable, deals less predictable than they are today. In that case it would have been better to have a strong hyperbolic discount rate, while now, when outcomes are increasingly reliable, then that rate should be dropped but it (presumably) hasn’t.
Of course, our intuitive discount rate should never reach the exponential that a model would predict, because there are always new unforeseen factors, but I would contend that the uncertainties have dropped substantially. This would make the particular hyperbolic rate that we intuitively discount payoffs at today biased, while in our evolutionary past it presumably would have been a better approximation of a suitable discount rate.