Thanks so much for your comment! It’s great to hear your perspective, and I’m glad my post could be a helpful resource.
30 year + planning is impractically difficult in most cases [...] So, if political systems were making the best decisions over a 30+ year time horizon [...] I think this would cover roughly 95%+ of the policies that a strong longtermist would want to see in the world that are not already happening.
Very interesting, I hadn’t thought much about this. One hesitation I have is that, as I understand it, strong longtermism can be acted on only when the very long-term effects of our actions are somewhat predictable. So I’m confused by the premise that 30+ year effects aren’t predictable—don’t we want longtermist policy making precisely when that premise doesn’t hold?
Maybe I’ve misunderstood, and what you’re saying is the following?
We can sometimes predict the 30+ year effects of an action, but only when these effects are so obvious that they’re similar to the <30 year effects.
E.g. The things that make brutal totalitarianism terrible for the very long-term future also make it terrible for the next few decades
Therefore there’s a lot of overlap between the policies that (as best as we can tell) have the best effects on the world in 30+ years, and those that have the best effects on the world in 0-30 years.
Therefore we can achieve most longtermist policy goals by getting policies that are best for the world over the next 0-30 years.
This seems mostly right to me, although I’d still be worried that shifting the focus entirely to current generations would have some important limitations:
Advocates of the long-term interests of current generations might not prioritize the most important very long-term issues
E.g. a 1% risk of extinction in a century might be really bad if you’re thinking about future generations, but not so important (potentially: not worth the political capital, or the costs of mitigation) if you’re thinking about current generations
Advocates of the long-term interests of current generations would pay no attention to benefits that take >30 years to realize, which might be very important
Future governments might have the option to fund long-term projects that wouldn’t pay off for several decades (e.g. sending people to Mars or more distant planets?), and much of the potential future value of humanity might depend on governments’ abilities to make such long-term investments
On the other hand, if improvements in nutrition and medicine continue to extend people’s lifespans, then the time horizon of serving current generations’ long-term interests might expand a lot by default
That said, any political change that’s feasible in the near term will have significant limitations, and I’m generally pretty optimistic about your strategic proposal of focusing more on long-term benefits to current generations. This seems like a really big takeaway that I totally missed in my initial writeup, so I’m adding a link to your comment on the main post.
Hi, The point isn’t about what is “predictable” it is about what is “plannable”. Predictions are only useful insofar as they let us decide how to act. What we want is to be able robustly positively affect the world in the future.
So the adapted version of your version of my argument would be:
We can sometimes take actions that robustly positively affect the world with a timescale of 30+ years, BUT as the future is so uncertain most such long-term plans involve being flexible and adaptable to changes and in practice they look very similar to if you planned for <30 year effects (with the additional caveat that in 30 years you want to be in as good a position as possible to keep making long-term positive changes going forward).
E.g. The global technologies and trends that could lead to brutal totalitarianism over the very long-term future are so uncertain that addressing the trends and technologies that might lead to totalitarianism in the next 30 years whilst also keeping an ongoing watchful eye on emerging trends and technologies and adapting to concerning changes and/or to opportunities to strengthen democracy, is likely the best plan you can make.
Therefore there’s a lot of overlap between the policies that (as best as we can tell) have the best effects on the world in 30+ years (or 60+ years ), those that have the best effects on the world in 0-30 years (and also leave the world in 30 years time ready to the next 30 years).
Therefore we can achieve most longtermist policy goals by getting policies that are best for the world over the next 0-30 years (and also leave the world in 30 years time ready to the next 30 years).
I think this mostly (although not quite 100%) addresses the two concerns that you raise
NOTES:
I would note that 30 years is not some magic number. Much of policy, including some long-term policy is time-independent. Where plans are made they might be over the next 1, 3, 5, 10, 20, 25, 30 or 50 years, as appropriate given the topic at hand. Over each length of time the aim should not be solely to maximise benefit over the planned time period but to leave the world in a good end state so that it can continue to maximise benefit going forward (eg your 1 year budgeting shouldn’t say lets spend all the money this year).
There are plans that go beyond 30 years but according to Roman Kaznaric’s book The Good Ancestor making plans for more than 30 years are very rare. And my own experience suggests 25 years is the maximum in most (UK) government work, and even at that length of time it is often poorly done. Hence I tend to settle on 30 years as reasonable maximum. There are of course some plans that go beyond 30 years. They tend to be on issues where long-term thinking is both necessary and simple (eg tree planting) or to use adopt adaptive planning techniques to allow for various changes in circumstances (eg the Thames Estuary 2100 flood planning).
I think this mostly (although not quite 100%) addresses the two concerns that you raise
Could you expand on this? I see how policy makers could realize very long-term value with ~30 yr planning horizons (through continual adaptation), but it’s not very clear to me why they would do so, if they’re mainly thinking about the long-term interests of future [edit: current] generations. For example, my intuition is that risks of totalitarianism or very long-term investment opportunities can be decently addressed with decades-long time horizons for making plans (for the reasons you give), but will only be prioritized if policy makers use even longer time horizons for evaluating plans. Am I missing something?
Yes sorry my misunderstanding. You are correct that this would still be non-ideal.
I don’t think in most cases it would be a big problem but yes it would be problem.
Also another very clear problem will all of this is that humans do not naturally plan in their own long-term self interest. So for example enfranchising the young would not necessarily lead to less short-termism just because they have longer to live. The policies would have to be more nuanced and complex than that.
Either way I think I am drawing a lesson to lean more towards strategies that focus a bit more on policies that empower creating a good world for the next generation rather than for all future generations, although of course both matter.
Thanks so much for your comment! It’s great to hear your perspective, and I’m glad my post could be a helpful resource.
Very interesting, I hadn’t thought much about this. One hesitation I have is that, as I understand it, strong longtermism can be acted on only when the very long-term effects of our actions are somewhat predictable. So I’m confused by the premise that 30+ year effects aren’t predictable—don’t we want longtermist policy making precisely when that premise doesn’t hold?
Maybe I’ve misunderstood, and what you’re saying is the following?
We can sometimes predict the 30+ year effects of an action, but only when these effects are so obvious that they’re similar to the <30 year effects.
E.g. The things that make brutal totalitarianism terrible for the very long-term future also make it terrible for the next few decades
Therefore there’s a lot of overlap between the policies that (as best as we can tell) have the best effects on the world in 30+ years, and those that have the best effects on the world in 0-30 years.
Therefore we can achieve most longtermist policy goals by getting policies that are best for the world over the next 0-30 years.
This seems mostly right to me, although I’d still be worried that shifting the focus entirely to current generations would have some important limitations:
Advocates of the long-term interests of current generations might not prioritize the most important very long-term issues
E.g. a 1% risk of extinction in a century might be really bad if you’re thinking about future generations, but not so important (potentially: not worth the political capital, or the costs of mitigation) if you’re thinking about current generations
Advocates of the long-term interests of current generations would pay no attention to benefits that take >30 years to realize, which might be very important
Future governments might have the option to fund long-term projects that wouldn’t pay off for several decades (e.g. sending people to Mars or more distant planets?), and much of the potential future value of humanity might depend on governments’ abilities to make such long-term investments
On the other hand, if improvements in nutrition and medicine continue to extend people’s lifespans, then the time horizon of serving current generations’ long-term interests might expand a lot by default
That said, any political change that’s feasible in the near term will have significant limitations, and I’m generally pretty optimistic about your strategic proposal of focusing more on long-term benefits to current generations. This seems like a really big takeaway that I totally missed in my initial writeup, so I’m adding a link to your comment on the main post.
Hi, The point isn’t about what is “predictable” it is about what is “plannable”. Predictions are only useful insofar as they let us decide how to act. What we want is to be able robustly positively affect the world in the future.
So the adapted version of your version of my argument would be:
We can sometimes take actions that robustly positively affect the world with a timescale of 30+ years, BUT as the future is so uncertain most such long-term plans involve being flexible and adaptable to changes and in practice they look very similar to if you planned for <30 year effects (with the additional caveat that in 30 years you want to be in as good a position as possible to keep making long-term positive changes going forward).
E.g. The global technologies and trends that could lead to brutal totalitarianism over the very long-term future are so uncertain that addressing the trends and technologies that might lead to totalitarianism in the next 30 years whilst also keeping an ongoing watchful eye on emerging trends and technologies and adapting to concerning changes and/or to opportunities to strengthen democracy, is likely the best plan you can make.
Therefore there’s a lot of overlap between the policies that (as best as we can tell) have the best effects on the world in 30+ years (or 60+ years ), those that have the best effects on the world in 0-30 years (and also leave the world in 30 years time ready to the next 30 years).
Therefore we can achieve most longtermist policy goals by getting policies that are best for the world over the next 0-30 years (and also leave the world in 30 years time ready to the next 30 years).
I think this mostly (although not quite 100%) addresses the two concerns that you raise
NOTES:
I would note that 30 years is not some magic number. Much of policy, including some long-term policy is time-independent. Where plans are made they might be over the next 1, 3, 5, 10, 20, 25, 30 or 50 years, as appropriate given the topic at hand. Over each length of time the aim should not be solely to maximise benefit over the planned time period but to leave the world in a good end state so that it can continue to maximise benefit going forward (eg your 1 year budgeting shouldn’t say lets spend all the money this year).
There are plans that go beyond 30 years but according to Roman Kaznaric’s book The Good Ancestor making plans for more than 30 years are very rare. And my own experience suggests 25 years is the maximum in most (UK) government work, and even at that length of time it is often poorly done. Hence I tend to settle on 30 years as reasonable maximum. There are of course some plans that go beyond 30 years. They tend to be on issues where long-term thinking is both necessary and simple (eg tree planting) or to use adopt adaptive planning techniques to allow for various changes in circumstances (eg the Thames Estuary 2100 flood planning).
I see, thanks for clarifying these points.
Could you expand on this? I see how policy makers could realize very long-term value with ~30 yr planning horizons (through continual adaptation), but it’s not very clear to me why they would do so, if they’re mainly thinking about the long-term interests of
future[edit: current] generations. For example, my intuition is that risks of totalitarianism or very long-term investment opportunities can be decently addressed with decades-long time horizons for making plans (for the reasons you give), but will only be prioritized if policy makers use even longer time horizons for evaluating plans. Am I missing something?Yes sorry my misunderstanding. You are correct that this would still be non-ideal.
I don’t think in most cases it would be a big problem but yes it would be problem.
Also another very clear problem will all of this is that humans do not naturally plan in their own long-term self interest. So for example enfranchising the young would not necessarily lead to less short-termism just because they have longer to live. The policies would have to be more nuanced and complex than that.
Either way I think I am drawing a lesson to lean more towards strategies that focus a bit more on policies that empower creating a good world for the next generation rather than for all future generations, although of course both matter.