‘Longtermism’

This post dis­cusses the in­tro­duc­tion and defi­ni­tion of the term ‘longter­mism’. Thanks to Toby Ord, Matthew van der Merwe and Hilary Greaves for dis­cus­sion.

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.