This post discusses the introduction and definition of the term ‘longtermism’. Thanks to Toby Ord, Matthew van der Merwe and Hilary Greaves for discussion.
[Edit, Nov 2021: After many discussions, I've settled on the following informal definition:
- Longtermism is the view that positively influencing the longterm future is a key moral priority of our time.
This is what I'm going with for What We Owe The Future.
With this in hand, I call strong longtermism is the view that positively influencing the longterm future is *the* key moral priority of our time. It turns out to be suprisingly difficult to define this precisely, but Hilary and I give it our best shot in our paper.]
Up until recently, there was no name for the cluster of views that involved concern about ensuring the long-run future goes as well as possible. The most common language to refer to this cluster of views was just to say something like ‘people interested in x-risk reduction’. There are a few reasons why this terminology isn’t ideal:
- It’s cumbersome and somewhat jargony
- It’s a double negative; whereas focusing on the positive (‘ensuring the long-run future goes well’) is more inspiring and captures more accurately what we ultimately care about
- People tend to understand ‘existential risk’ as referring only to extinction risk, which is a strictly narrower concept
- You could care a lot about reducing existential risk even though you don’t care particularly about the long term if, for example, you think that extinction risk is high this century and there’s a lot we can do to reduce it, such that it’s a very effective thing even by the lights of the present generation’s interests.
- Similarly, you can care a lot about the long-run future without focusing on existential risk reduction, because existential risk is just about drastic reductions in the value of the future. (‘Existential risk’ is defined as a risk where an adverse outcome would either annihilate Earth-originating intelligent life or permanently and drastically curtail its potential.) But, conceptually at least (and I think in practice, too) smaller improvements in the expected value of the long-run future could be among the things we want to focus on, such as changing people’s values, or changing political institutions (like the design of a world government) before some lock-in event occurs. You might also think (as Tyler Cowen does) that speeding up economic and technological progress is one of the best ways of improving the long-run future.
For these reasons, and with Toby Ord’s in-progress book on existential risk providing urgency, Toby and Joe Carlsmith started leading discussions about whether there were better terms to use. In October 2017, I proposed the term ‘longtermism’, with the following definition:
“Longtermism =df the view that the most important determinant of the value of our actions today is how those actions affect the very long-run future.”
Since then, the term ‘longtermism’ seems to have taken off organically. I think it’s here to stay. Unlike ‘existential risk reduction’, the idea behind ‘longtermism’ is that it is compatible with any empirical view about the best way of improving the long-run future and, I hope, helps immediately convey the sentiment behind the philosophical position, in the same way that ‘environmentalism’ or ‘liberalism’ or ‘cosmopolitanism’ does.
But getting a good definition of the term is important. As Ben Kuhn notes, the term could currently be understood to refer to a mishmash of different views. I think that’s not good, and we should try to develop some standardisation before the term is locked in to something suboptimal.
I think that there are three natural concepts in this area, which we should distinguish. My proposal is that we should name them as follows (stating the concepts imprecisely for now):
(i) longtermism, which designates an ethical view that is particularly concerned with ensuring long-run outcomes go well;
(ii) strong longtermism, which, like my original proposed definition, is the view that long-run outcomes are the thing we should be most concerned about;
(iii) very strong longtermism, the view on which long-run outcomes are of overwhelming importance. [1]
My initial proposal was that ‘longtermism’ (with no modifier) should refer to (ii), whereas now I think it should refer to (i). This is primarily because:
- The first concept is intuitively attractive to a significant proportion of the wider public (including key decision-makers like policymakers and business leaders); my guess is that most people would find it intuitively attractive. In contrast, the second concept is widely regarded as unintuitive, including even by proponents of the view.
- At the same time, it seems that we’d achieve most of what we want to achieve if the wider public came to believe that ensuring the long-run future goes well is one important priority for the world, and took action on that basis, even if they didn’t regard it as the most important priority.
In general, if I imagine ‘longtermism’ taking off as a term, I imagine it getting a lot of support if it designates the first concept, and a lot of pushback if it designates the second concept. It’s also more in line with moral ideas and social philosophies that have been successful in the past: environmentalism claims that protecting the environment is important, not that protecting the environment is (always) the most important thing; feminism claims that upholding women’s rights is important, not that doing so is (always) the most important thing. I struggle to think of examples where the philosophy makes claims about something being the most important thing, and insofar as I do (totalitarian marxism and fascism are examples that leap to mind), they aren’t the sort of philosophies I want to emulate.
Let’s now consider definitions of the variants of longtermism.
Longtermism
I think we have two paths forward for the definition of longtermism. The first is the ‘no definition’ approach, suggested to me by Toby Ord:
Longtermism is a philosophy that is especially concerned with improving the long-term future.
This is roughly analogous to terms like 'environmentalism’ and ‘feminism.’
The second approach is to have some minimal definition. For example:
Longtermism is the view that:
(i) Those who live at future times matter just as much, morally, as those who live today;
(ii) Society currently privileges those who live today above those who will live in the future; and
(iii) We should take action to rectify that, and help ensure the long-run future goes well.
I’m not confident at all about this precise definition, but I prefer the minimal definition approach over the no-definition approach for a few reasons:
- When I look at other -isms, there is often a lot of confusion around what the concept denotes, and this hinders those who want to encourage others to take action in line with the -ism. Some examples:
- Effective altruism is still widely conflated with utilitarianism, or with earning to give, or with the randomista movement. I’ve suggested a definition and I think that having this definition will both help with responses to critics and lessen the amount by which people in the first place misunderstand what effective altruism is about. I wish we’d had the existing definition much earlier.
- Liberalism means two different things in the US and UK: in the US a liberal is a social progressive whereas in the UK a liberal is a proponent of free markets.
- Anecdotally, I see a lot of confusion and resultant fighting over the term ‘feminism’, where it seems to me that a precise definition could have helped mitigate this at least somewhat.
- In particular, I worry that without the minimal definition, ‘longtermism’ would end up referring to strong longtermism, or even to very strong longtermism. The analogy here would be ‘effective altruism’ referring simply to applied utilitarianism in many people’s minds. Or, alternatively, it might refer to an unattractive mishmash of concepts, with Ben Kuhn’s suggestion about what ‘longtermism’ currently refers to being an example of that.
I also just don’t see much of a case against having a minimal definition. If the precise definition turns out to be unhelpful in the future, we can quietly drop it. Or the precise definition might be something we don’t often highlight, but is just something we can refer to if people are grossly misrepresenting the position. And the minimal definition is compatible with people using the ‘no definition’ version too.
The strongest case for the no-definition approach, in my view, is that it could enable the term to evolve so as to better fit future times, and any current definition could be myopic. Perhaps that flexibility helped explain why terms like ‘environmentalism’ and ‘liberalism’ took off. But my proposed definition is so minimal that I find it hard to see that there would be much benefit from even greater flexibility.
An alternative minimal definition, suggested by Hilary Greaves (though the precise wording is my own), is that we could define longtermism as the view that the (intrinsic) value of an outcome is the same no matter what time it occurs. This rules out views on which we should discount the future or that we should ignore the long-run indirect effects of our actions, but would not rule out views on which it’s just empirically intractable to try to improve the long-term future. Part of the idea is that this definition would open the way to a debate about the relevant empirical issues, in particular on the tractability of affecting the long run. This definition makes ‘longtermism’ somewhat more like the terms ‘cosmopolitanism’ or ‘antispeciesism’, and less like ‘neoliberalism’ or ‘feminism’ or ‘environmentalism’.
In my view, this definition would be too broad. I think the distinctive idea that we should be trying to capture is the idea of trying to promote good long-term outcomes. I see the term 'longtermism' creating value if it results in more people taking action to help ensure that the long-run future goes well. But if one can endorse longtermism without thinking that we should, at least to some extent, try to promote good long-term outcomes, then it seems like we lose much of that value. And, insofar as the term has taken off so far, it has been used to refer to people who think that we should be trying to make the long-run future go better.
One implication of my definition, which one might object to, is that if, in the future, society starts to care about the long-term future exactly to the extent it should (or more than it should), then longtermism is no longer true. In my view, that seems like a good implication. Suppose that society started caring too much about the long term and was neglecting the interests of the present generation: then there would be no need for ‘longtermism’ as an idea; indeed, we would want to promote shorttermism instead! On my definition, longtermism stops being true exactly when it is no longer needed.
Strong Longtermism
The definition I initially proposed for longtermism was an attempt to capture the idea of strong longtermism. Here’s a stylistically modified version:
Strong Longtermism is the view that the primary determinant of the value of our actions today is how those actions affect the very long-term future.
I think this definition is good enough for general use, but is technically not correctly capturing what we want. Perhaps most of the value of our actions comes from their long-run effects, but most of the differences in value between actions comes from their short-run effects. If so, then we should spend our time trying to figure out which actions best improve the short run; this is not the spirit of longtermism.
Recently, Hilary Greaves and I have been working on a paper on the core case for longtermism and propose the more unwieldy but more philosophically precise:
Axiological strong longtermism =df In a wide class of decision situations, the option that is ex ante best is contained in a fairly small subset of options whose ex ante effects on the very long-run future are best.
Deontic strong longtermism =df In a wide class of decision situations, the option one ought, ex ante, to choose is contained in a fairly small subset of options whose ex ante effects on the very long-run future are best.
Where by “the option whose effects on the very long-run future are best”, we mean “the option whose effects on the future from time t onwards are best”, where t is a surprisingly long time from now (say, 1000 years). My view is that we should choose the smallest t such that, any larger choice of t makes little difference to what we would prioritise.
The key idea behind both the informal definition and the more precise definition is that, in order to assess the value (or normative status) of a particular action we can in the first instance just look at the long-run effects of that action (that is, those after 1000 years), and then look at the short-run effects just to decide among those actions whose long-run effects are among the very best.
Hyphenation
People tend to naturally use both ‘long-termism’ and ‘longtermism’. I think it makes sense to decide on one as canonical, and I think the right choice is the un-hyphenated ‘longtermism’. There are a few reasons for this.
First, grammatically, either would be fine. ‘Long-term’ is a compound adjective (e.g. “She cares about the long-term future”), ‘long term’ is an adjective-noun pair (e.g. “She cares about the long term.”) And, in general, as long as a word is unambiguous, you don’t need to include a hyphen even in cases where it’s permissible to do so: so, for example, it’s ‘post-structuralism’ but ‘postfeminism’. [2] As the style manual of the Oxford University Press comments: “If you take hyphens seriously, you will surely go mad.”
Second, if you can make a term shorter and quicker to write without sacrificing much, you should do so. So, for example, “Neoliberalism” is clearly a better term than “Neo-liberalism” and either is grammatically permissible.
Third, hyphenated words tend to lose their hyphen over time as they become increasingly familiar. Examples: to-morrow, to-day, co-operative, pigeon-hole, e-mail, etc. In 2007, the sixth edition of the Shorter Oxford English Dictionary removed the hyphens from 16,000 entries. So even if we adopted ‘long-termism’ it would probably change to ‘longtermism’ over time.
Fourth, the hyphenation makes the term ambiguous. Consider some other hyphenated -isms: ‘anarcho-capitalism’, or ‘post-structuralism’. Here the hyphenated prefix modifies an existing -ism. So the natural reading of ‘long-termism’ would be that ‘long’ modifies some other concept, ‘termism’. But of course that’s not what this term is supposed to convey. Insofar as ‘termism’ isn’t a concept, I don’t expect this to cause confusion, but it’s still a mild reason to prefer the unhyphenated version.
The best counterargument I know is that, on this view, the opposite of longtermism would be ‘shorttermism,’ which has a strange-looking double ‘t’. But there are many compound words with double consonants that we’ve gotten used to, like ‘bookkeeping’, ‘earring’, and ‘newsstand,’ including at least one with a double ‘t’, namely ‘posttraumatic’ (though this is also written ‘post-traumatic’), and even some with double vowels as a result of hyphen loss, like ‘cooperation’. And I’m not sure how often ‘shorttermism’ will get used. So I don’t see this as a strong counterargument.
[1] Nick Beckstead’s Main Thesis in his dissertation makes a claim similar to strong longtermism: “Main Thesis: From a global perspective, what matters most (in expectation) is that we do what is best (in expectation) for the general trajectory along which our descendants develop over the coming millions, billions, and trillions of years.” But the title of his thesis — ‘On the Overwhelming Importance of Shaping the Far Future’ — suggests an endorsement of very strong longtermism.
[2] Note that, for the compound adjective form, it’s grammatically preferred to say ‘the long-term future’, fine to say ‘the long term future’ (because there’s no ambiguity caused by dropping the hyphen), but currently not grammatical to say ‘the longterm future’. We could try using ‘longterm' with the aim of changing usage; my view is to stick with current grammar here, though, as we’re not using ‘long-term’ as a term of art or aiming to change its meaning.
Thanks for writing this; I thought it was good.
I would wonder if we might consider weakening this a little:
Anecdotally, it seems that many people - even people I've spoken to at EA events! - consider future generations to have zero value. Caring any amount about future people at all is already a significant divergence, and I would instinctively say that someone who cared about the indefinite future, but applied a modest discount factor, was also longtermist, in the colloquial-EA sense of the word.
I second weakening the definition. As someone who cares deeply about future generations, I think it is infeasible to value them equally to people today in terms of actual actions. I sketched out an optimal mitigation path for asteroid/comet impact. Just valuing the present generation in one country, we should do alternate foods. Valuing the present world, we should do asteroid detection/deflection. Once you value hundreds of future generations, we should add in food storage and comet detection/deflection, costing many trillions of dollars. But if you value even further in the future, we should take even more extreme measures, like many redundancies. And this is for a very small risk compared to things like nuclear winter and AGI. Furthermore, even if one does discount future generations, if you think we could have many computer consciousnesses in only a century or so, again we should be donating huge amount of resources for reducing even small risks. I guess one way of valuing future generations equally to the present generation is to value each generation an infinitesimal amount, but that doesn't seem right.
Is the argument here something along the lines of; I find that I don't want to struggle to do what these values would demand, so they must not be my values?
I hope I'm not seeing an aversion to surprising conclusions in moral reasoning. Science surprises us often, but it keeps getting closer to the truth. Technology surprises us all of the time, but it keeps getting more effective. If you wont accept any sort of surprise in the domain of applied morality, your praxis is not going to end up being very good.
Thanks for your comment. I think my concern is basically addressed by Will's comment below. That is it is good to value everyone equally. However, it is not required in our daily actions to value a random person alive today is much as ourselves or a random person in the future as much as ourselves. That is, it is permissible to have some special relationships and have some personal prerogatives.
Thanks for this! Wasn’t expecting pushback on this aspect, so that’s helpful.
I’ll start with a clarification. What I mean by that clause is: For any populations of people p, and any permutation with respect to time of that population that keeps the wellbeing levels of all individuals the same, p’, p and p’ are equally good.
I.e. if you take some world, and move everyone around in time but keep their levels of wellbeing the same, then the new world is just as good as the old world.
(Caveat, I’m sure that there will be technical problems with this principle, so I’m just suggesting it as a first pass. This is in analogy with how I’d define the ideas that race, gender, etc are morally irrelevant: imagine two worlds where everything is the same except that people’s races or genders are different; these are equally good.)
That *does* rule out pure time-discounting, which Haydn suggests. But I’d need quite a lot of convincing to allow that into a definition of longtermism. (The strongest case I could see would be if spatiotemporal discounting is the best solution to problems of infinite ethics.)
But it doesn’t rule out the following:
If you also add the normative assumptions of agent-neutral consequentialism and expected utility theory, and the empirical assumptions that the future is big and affectable, then you do get led to strong or very strong longtermism. But, if you accept those assumptions, then strong or very strong longtermism seems correct.
I’m worried that a weakening where we just claim that future people matter, to some degree, would create too broad a church. In particular: Economists typically suppose something like a 1.5% pure rate of time preference. On this view, even people in a billion years’ time matter. But the amount by which they matter is tiny and, in practice, any effects of our actions beyond a few centuries are irrelevant. I want a definition of longtermism - even a minimal definition - to rule out that view. But then I can’t think of a non-arbitrary stopping point in between that view and my version. And I do think that there are some benefits in terms of inspiringness of my version, too — it’s clear and robust.
What would most move me is if I thought that the weaker version would capture a lot more people. And it’s interesting that Larks has had the experiences they have had.
But it currently seems to me that’s not the case. Aron Vallinder and I ran a survey on people’s attitudes on this issue: here were the results for how much people agree or disagree with the claim that ‘people in the distant future matter just as much as those alive today’:
Given this, I’m inclined to stick with the stronger version — it already has broad appeal, and has some advantages over the weaker version.
Hi Will,
I would expect the item "People in the distant future matter just as much as those alive today" to produce somewhat inflated levels of agreement. One reason is that I expect that any formulation along the lines of "X people matter just as much as Y people" will encourage agreement, because people don't want to be seen as explicitly saying that any people matter less than others and agreement seems pretty clearly to be the socially desirable option. Another is that acquiescence bias will increase agreement levels, particularly where the proposition is one that people haven't really considered before and/or don't have clearly defined attitudes towards.
Me and Sanjay found pretty different results to this and, as I think he mentioned, we'll be sharing a writeup of the results soon.
Interesting! Can't wait. :)
(And agree that this would produce inflated levels of agreement, but feel like "do you endorse this statement?" is the relevant criterion for a definition even if that endorsement is inflated relative to action).
It seems quite plausible to me (based on intuitions from algorithmic complexity theory) that spatiotemporal discounting is the best solution to problems of infinite ethics. (See Anatomy of Multiversal Utility Functions: Tegmark Level IV for a specific proposal in this vein.)
I think the kinds of discounting suggested by algorithmic information theory is mild enough in practice to be compatible with our intuitive notions of longtermism (e.g., the discount factors for current spacetime and a billion years from now are almost the same), and would prefer a definition that doesn't rule them out, in case we later determine that the correct solution to infinite ethics does indeed lie in that direction.
Why not include this in the definition of strong longtermism, but not weak longtermism?
Having longtermism just mean "caring a lot about the long-term future" seems the most natural and least likely to cause confusion. I think for it to mean anything other than that, you're going to have to keep beating people over the head with the definition (analogous to the sorry state of the phrase, "begs the question").
When most people first hear the term longtermism, they're going to hear it in conversation or see it in writing without the definition attached to it. And they are going to assume it means caring a lot about the long-term future. So why define it to mean anything other than that?
On the other hand, anyone who comes across strong longtermism, is much more likely to realize that it's a very specific technical term, so it seems much more natural to attach a very specific definition to it.
IMHO the most natural name for "people at any time have equal value" should be something like temporal indifference, which more directly suggests that meaning.
Edit: I retract temporal indifference in favor of Holly Elmore's suggestion of temporal cosmopolitanism.
I agree with the sentiment that clause (i) is stronger than it needs to be. I don't really think this is because it would be good to include other well-specified positions like exponential discounting, though. It's more that it's taking a strong position, and that position isn't necessary for the work we want the term to do. On the other hand I also agree that "nonzero" is too weak. Maybe there's a middle ground using something like the word "significant"?
[For my own part intellectual honesty might make me hesitate before saying "I agree with longtermism" with the given definition — I think it may well be correct, but I'm noticeably less confident than I am in some related claims.]
"Aron Vallinder and I ran a survey on people’s attitudes on this issue…"
Hey Will — who was this a survey of?
I think Positly and mTurk, for people with university degrees in the US. We'll share a proper write-up soon of the wider survey, we haven't really looked at the data yet.
I wonder if it might be helpful to modify your claim (i) to be more similar to Hilary’s definition by referring to intrinsic value rather than mattering morally. Eg. Something like:
(i) Lives situated in the future have just as much intrinsic value as lives situated in the present
I think that wording could be improved but to me it seems like it does a better job of conveying:
“For any populations of people p, and any permutation with respect to time of that population that keeps the wellbeing levels of all individuals the same, p’, p and p’ are equally good."
As well as making allowance for special relationships and personal prerogatives, this also allows for the idea that the current generation holds some additional instrumental value (in enabling/affecting future generations) in addition to our intrinsic value. To me this instrumental value would have some impact on the extent to which people matter morally.
I think if you acknowledge that current populations may have some greater value (eg by virtue of instrumental value) then you would need to make claim (ii) stronger, eg. "society currently over-privileges those who live today above those who will live in the future".
I appreciate that “matter just as much, morally” is a stronger statement (and perhaps carries some important meaning in philosophy of which I’m ignorant?). I think it also sounds nicer, which seems important for an idea that you want to have broad appeal. But perhaps its ambiguity (as I perceive it) leaves it more open to objections.
Also, FWIW I disagree with the idea that (i) could be replaced with “Those who live at future times matter morally”. It doesn’t seem strong enough and I don’t think (iii) would flow from that and (ii) as it is. So I think if you did change to this weaker version of (i) it would be even more important to make (ii) stronger.
Well, we should probably distinguish between:
The minimal definition should probably include something like 2, but not 1. I think the current definition leaves it somewhat ambiguous, although I'm inclined to interpret it as my 2). I'd be surprised if you think 2 is controversial.
It is not that easy to distinguish between these two theories! Consider three worlds:
If you don't value creating positive people, you end up being indifferent between the first and second worlds, and between the second and third worlds... But then by 2), you want to prefer the third to the first, suggesting a violation of transitivity.
The (normal) person-affecting response here is to say that options 1 and 3 and incomparable in value to 2 - existence is neither better than, worse than, or equally good as, non-existence for someone. However, if Sam exists necessarily, then 2 isn't a option, so then we say 3 is better than 1. Hence, no issues with transitivity.
Well that doesn't show it's hard to distinguish between the views, it just shows a major problem for person-affecting views that want to hold 2) but not 1).
You mean, shows a major finding, no? :)