Longtermism: a call to protect future generations

Making moral decisions always involves tradeoffs. We have limited resources, so spending on one issue means we have less to spend on another. And there are many deserving causes we could devote our efforts to. If we focus on helping future generations, we will necessarily not prioritise as highly many of the urgent needs in the present.

But we don’t think this is as troubling an objection to longtermism as it may initially sound, for at least three reasons:

1. Most importantly, many longtermist priorities, especially reducing extinction risk, are also incredibly important for people alive today. For example, we believe preventing an AI-related catastrophe or a cataclysmic pandemic are two of the top priorities, in large part because of their implications for future generations. But these risks could materialise in the coming decades, so if our efforts succeed most people alive today would benefit. Some argue that preventing global catastrophes could actually be the single most effective way to save the lives of people in the present.

2. If we all took moral impartiality more seriously, there would be a lot more resources going to help the worst-off today — not just the far future. Impartiality is the idea that we should care about the interests of individuals equally, regardless of their nationality, gender, race, or other characteristics that are morally irrelevant. This impartiality is part of what motivates longtermism — we think the interests of future individuals are often unjustifiably undervalued.

We think if impartiality were taken more seriously in general, we’d live in a much better world that would commit many more resources than it currently does toward alleviating all kinds of suffering, including for the present generation. For example, we’d love to see more resources go toward fighting diseases, improving mental health, reducing poverty, and protecting the interests of animals.

3. Advocating for any moral priority means time and resources are not going to another cause that may also be quite worthy of attention. Advocates for farmed animals’ or prisoners’ rights are in effect deprioritising the interests of alternative potential beneficiaries, such as the global poor. So this is not just an objection to longtermism — it’s an objection to any kind of prioritisation.

Ultimately, this objection hinges on the question of whether future generations are really worth caring about — which is what the rest of this article is about.

Some people, especially those trained in economics, claim that we shouldn’t treat individual lives in the future equally to lives today. Instead, they argue, we should systematically discount the value of future lives and generations by a fixed percentage.

(We’re not talking here about discounting the future due to uncertainty, which we cover below.)

When economists compare benefits in the future to benefits in the present, they typically reduce the value of the future benefits by some amount called the “discount factor.” A typical rate might be 1% per year, which means that benefits in 100 years are only worth 36% as much as benefits today, and benefits in 1,000 years are worth almost nothing.

This may seem like an appealing way to preserve the basic intuition we began with — that we have strong reasons to care about the wellbeing of future generations — while avoiding the more counterintuitive longtermist claims that arise from considering the potentially astronomical amounts of value that our universe might one day hold. On this view, we would care about future generations, but not as much as the present generation, and mostly only the generations that will come soon after us.

We agree there are good reasons to discount economic benefits. One reason is that if you receive money now, you can invest it, and earn a return each year. This means it’s better to receive money now rather than later. People in the future might also be wealthier, which means that money is less valuable to them.

However, these reasons don’t seem to apply to welfare — people having good lives. You can’t directly ‘invest’ welfare today and get more welfare later, like you can with money. The same seems true for other intrinsic values, such as justice. And longtermism is about reasons to care about the interests of future generations, rather than wealth.

As far as we know, most philosophers who have worked on the issue don’t think we should discount the intrinsic value of future lives — even while they strongly disagree about other questions in population ethics. It’s a simple principle that is easy to accept: one person’s happiness is worth just the same amount no matter when it occurs.

Indeed, if you suppose we can discount lives in the far future, we can easily end up with conclusions that sound absurd. For instance, a 3% discount rate would imply that the suffering of one person today is morally equal to the suffering of 16 trillion people in 1,000 years. This seems like a truly horrific conclusion to accept.

And any discount rate will mean that, if we found some reliable way to save 1 million lives from intense suffering in either 1,000 years or 10,000 years, it would be astronomically more important to choose the sooner option. This, too, seems very hard to accept.²⁰

If we reject the discounting of the value of future lives, then the many potential generations that could come after us are still worthy of moral concern. And this doesn’t stand in tension with the economic practice of discounting monetary benefits.

If you’d like to see a more technical discussion of these issues, see Discounting for Climate Change by Hilary Graves. There is a more accessible discussion at 1h00m50s in our podcast with Toby Ord and in Chapter 4 of Stubborn Attachments by Tyler Cowen.

There are some practical, rather than intrinsic, reasons to discount the value of the future. In particular, our uncertainty about how the future will unfold makes it much harder to influence than the present, and even more near-term actions can be exceedingly difficult to forecast.

And because of the possibility of extinction, we can’t even be confident that the future lives we think are so potentially valuable will come into existence. As we’ve argued, that gives us reason to reduce extinction risks when it’s feasible — but it also gives us reason to be less confident these lives will exist and thus to weight them somewhat less in our deliberations.

In the same way, a doctor performing triage may choose to prioritise caring for a patient who had a good chance of surviving their injuries over one who has much less clear likelihood of survival regardless of the medical care they receive.

This uncertainty — along with the extreme level of difficulty in trying to predict the long-term impacts of our actions — certainly makes it much harder to help future generations, all else equal. And in effect, this point lowers the value of working to benefit future generations.

So even if we can affect how things unfold for future generations, we’re generally going to be very far from certain that we are actually making things better. And arguably, the further away in time the outcomes of our actions are, the less sure we can be that they will come about. Trying to improve the future will never be straightforward.

Still, even given the difficulty and uncertainty, we think the potential value at stake for the future means that many uncertain projects are still well worth the effort.

You might disagree with this conclusion if you believe that human extinction is so likely and practically unavoidable in the future that the chance that our descendants will still be around rapidly declines as we look a few centuries down the line. We don’t think it’s that likely — though we are worried about it.

Journalist Kelsey Piper critiqued MacAskill’s argument for longtermist interventions focused on positive trajectory changes (as opposed to extinction risks) in Asterisk, writing:

What share of people who tried to affect the long-term future succeeded, and what share failed? How many others successfully founded institutions that outlived them — but which developed values that had little to do with their own?
…
Most well-intentioned, well-conceived plans falter on contact with reality. Every simple problem splinters, on closer examination, into dozens of sub-problems with their own complexities. It has taken exhaustive trial and error and volumes of empirical research to establish even the most basic things about what works and what doesn’t to improve peoples’ lives.

Piper does still endorse working on extinction reduction, which she thinks is a more tractable course of action. Her doubts about the possibility of reliably anticipating our impact on the trajectory of the future, outside of extinction scenarios, are worth taking very seriously.

You might have a worry about longtermism that goes deeper than just uncertainty. We act under conditions of uncertainty all the time, and we find ways to manage it.

There is a deeper problem known as cluelessness. While uncertainty is about having incomplete knowledge, cluelessness refers to the state of having essentially no basis of knowledge at all.

Some people believe we’re essentially clueless about the long-term effects of our actions. This is because virtually every action we take may have extremely far-reaching unpredictable consequences. In time travel stories, this is sometimes referred to as the “butterfly effect” — because something as small as a butterfly flapping its wings might influence air currents just enough to cause a monsoon on the other side of the world (at least for illustrative purposes).

If you think your decision of whether to go to the grocery store on Thursday or Friday might determine whether the next Gandhi or Stalin is born, you might conclude that actively trying to make the future go well is a hopeless task.

Like some other important issues discussed here, cluelessness remains an active area of philosophical debate, so we don’t think there’s necessarily a decisive answer to these worries. But there is a plausible argument, advanced philosopher and advisor to 80,000 Hours Hilary Greaves that longtermism is, in fact, the best response to the issue of cluelessness.

This is because cluelessness hangs over the impact of all of our actions. Work trying to improve the lives of current generations, such as direct cash transfers, may predictably benefit a family in the foreseeable future. But the long-term consequences of the transfer are a complete mystery.

Successful longtermist interventions, though, may not have this quality — particularly interventions to prevent human extinction. If we, say, divert an asteroid that would otherwise have caused the extinction of humanity, we are not clueless about the long-term consequences. Humanity will at least have the chance to continue existing into the far future, which it wouldn’t have otherwise had.

There’s still uncertainty, of course, in preventing extinction. The long-term consequences of such an action aren’t fully knowable. But we’re not clueless about them either.

If it’s correct that the problem of cluelessness bites harder for some near-term interventions than longtermist ones, and perhaps least of all for preventing extinction, then this apparent objection doesn’t actually count against longtermism.

For an alternative perspective, though, check out The 80,000 Hours Podcast interview with Alexander Berger.

Because of the nature of human reproduction, the identity of who gets to be born is highly contingent. Any individual is the result of the combination of one sperm and one egg, and a different combination of sperm and egg would’ve created a different person. Delaying the act of conception at all — for example, by getting stuck at a red light on your way home — can easily result in a different sperm fertilising the egg, which means another person with a different combination of genes will be born.

This means — somewhat surprisingly — that pretty much all our actions have the potential to impact the future by changing which individuals get born in the future.

If you care about affecting the future in a positive way, this creates a perplexing problem. Many actions undertaken to improve the future, such as trying to reduce the harmful effects of climate change or developing a new technology to improve people’s lives, may deliver the vast majority of their benefits to people who wouldn’t have existed had the course of action never been taken.

So while it seems obviously good to improve the world in this way, it may be impossible to ever point to specific people in the future and say they were made better off by these actions. You can make the future better overall, but you may not make it better for anyone in particular.

Of course, the reverse is true: you may take some action that makes the future much worse, but all the people who experience the consequences of your actions may never have existed had you chosen a different course of action.

This is known as the ‘non-identity problem.’ Even when you can make the far future better with a particular course of action, you will almost certainly never make any particular individuals in the far future better off than they otherwise would be.

Should this problem cause us to abandon longtermism? We don’t think so.

While the issue is perplexing, accepting it as a refutation of longtermism would prove too much. It would, for example, undermine much of the very plausible case that policymakers should in the past have taken significant steps to limit the effects of climate change (since those policy changes can be expected to, in the long run, lead to different people being born).

Or consider a hypothetical case of a society that is deciding what to do with its nuclear waste. Suppose there are two ways of storing it: one way is cheap, but it means that in 200 years time, the waste will overheat and expose 10,000,000 people to sickening radiation that dramatically shortens their lives. The other storage method guarantees it will never hurt anyone, but it is significantly more expensive, and it means currently living people will have to pay marginally higher taxes.

Assuming this tax policy alters behaviour just enough to start changing the identities of the children being born, it’s entirely plausible that, in 200 years time, no one would exist who would’ve existed if the cheap, dangerous policy had been implemented. This means that none of the 10,000,000 people who have their lives cut short can say they would have been better off had their ancestors chosen the safer storage method.²¹

Still, it seems intuitively and philosophically unacceptable to believe that a society wouldn’t have very strong reasons to adopt the safe policy over the cheap, dangerous policy. If you agree with this conclusion, then you agree that the non-identity problem does not mean we should abandon longtermism. (You may still object to longtermism on other grounds!)

Nevertheless, this puzzle raises pressing philosophical questions that continue to generate debate, and we think better understanding these issues is an important project.

We said that we thought it would be very bad if humanity was extinguished, in part because future individuals who might have otherwise been able to live full and flourishing lives wouldn’t ever get the chance.

But this raises some issues related to the ‘non-identity problem.’ Should we actually care whether future generations come into existence, rather than not?

Some people argue that perhaps we don’t actually have moral reasons to do things that affect whether individuals exist — in which case ensuring that future generations get to exist, or increasing the chance that humanity’s future is long and expansive or would be morally neutral in itself.

This issue is very tricky from a philosophical perspective; indeed, a minor subfield of moral philosophy called population ethics sets out to answer this and related questions.

So we can’t expect to fully address the question here. But we can give a sense of why we think working to ensure humanity survives and that the future is filled with flourishing lives is a high moral priority.

Consider first a scenario in which you, while travelling the galaxy in a spaceship, come across a planet filled with an intelligent species leading happy, moral, fulfilled lives. They haven’t achieved spaceflight, and may never do so, but they appear likely to have a long future ahead of them on their planet.

Would it not seem like a major tragedy if, say, an asteroid were on course to destroy their civilization? Of course, any plausible moral view would advise saving the species for their own sakes. But it also seems like it’s an unalloyed good that, if you divert the asteroid, this flourishing species will be able to continue on for many future generations, flourishing in their corner of the universe.

If we have that view about that hypothetical alien world, we should probably have the same view of our own planet. Humans, of course, aren’t necessarily that happy, moral, and fulfilled for their lives. But the vast majority of us want to keep living — and it seems at least possible that our descendants could have lives many times more flourishing than we have. They might even ensure that all other sentient beings have joyous lives well-worth living. This seems to give us strong reasons to make this potential a reality.

For a different kind of argument along these lines, you can read Joe Carlsmith’s “Against neutrality about creating happy lives.”

Some people advocate a ‘person-affecting’ view of ethics. This view is sometimes summed up with the quip: “ethics is about helping make people happy, not making happy people.”

In practice, this means we only have moral obligations to help those who are already alive²² — not to enable more people to exist with good lives. For people who hold such views, it may be permissible to create a happy person, but doing so is morally neutral.

This view has some plausibility, and we don’t think it can be totally ignored. However, philosophers have uncovered a number of problems with it.

Suppose you have the choice to bring into existence one person with an amazing life, or another person whose life is barely worth living, but still more good than bad. Clearly, it seems better to bring about the amazing life.

But if creating a happy life is neither good nor bad, then we have to conclude that both options are neither good nor bad. This implies the options are equal, and you have no reason to do one or the other, which seems bizarre.

And if we accepted a person-affecting view, it might be hard to make sense of many of our common moral beliefs around issues like climate change. For example, it would imply that policymakers in the 20th century might have had little reason to mitigate the impact of CO₂ emissions on the atmosphere if the negative effects would only affect people who would be born several decades in the future. (This issue is discussed more above.)

This is a complex debate, and rejecting the person-affecting view also has counterintuitive conclusions. In particular, Parfit showed that if you agree that it’s good to create people whose lives are more good than bad, there is a strong argument for the conclusion that we could have a better world filled with a huge number of people whose lives are just barely worth living. He called this the “repugnant conclusion”.

Both sides make important points in this debate. You can see a summary of the arguments in this public lecture by Hilary Greaves (based on this paper). It’s also discussed in our podcast with Toby Ord.

We’re uncertain about what the right position is, but we’re inclined to reject person-affecting views. Since many people hold something like the person-affecting view, though, we think it deserves some weight, and that means we should act as if we have somewhat greater obligations to help someone who’s already alive compared to someone who doesn’t exist yet. (This is an application of moral uncertainty).

One note however: even people who otherwise embrace a person-affecting view often think that is morally bad to do something that brings someone into existence who has a life full of suffering and who wishes they’d never been born. If that’s right, you should still think that we have strong moral reasons to care about the far future, because there’s the possibility it could be horrendously bad as well as very good for a large number of individuals. On any plausible view, there’s a forceful case to be made for working to avert astronomical amounts of suffering. So even someone who believes strongly in a person-affecting view of ethics might have reason to embrace a form of longtermism that prioritises averting large-scale suffering in the future.

Trying to weigh this up, we think society should have far greater concern for the future than it does now, and that as with climate change, it often makes sense to prioritise making things go well for future individuals. In the case of climate change, for example, it was likely the case that society should have long ago taken on the non-trivial costs of financing efforts to develop highly reliable clean energy and navigating away from a carbon-intensive economy.

Because of moral uncertainty, though, we care more about the present generation than we would if we naively weighed up the numbers.

Yes, it would be arrogant. But longtermism doesn’t require us to know the future.

Instead, the practical implication of longtermism is that we take steps that are likely to be good over the wide range of possible futures. We think it’s likely better for the future if, as we said above, we avoid extinction, we manage our resources carefully, we foster institutions that promote cooperation rather than violent conflict, and we responsibly develop powerful technology. None of these strategies requires us knowing what the future will look like.

We talk more about the importance of all this uncertainty in the sections above.

This isn’t exactly an objection, but one response to longtermism asserts not that the view is badly off track but that it’s superfluous.

This may seem plausible if longtermism primarily inspires us to prioritise reducing extinction risks. As discussed above, doing so could benefit existing people — so why even bother talking about the benefits to future generations?

One reply is: we agree that you don’t need to embrace longtermism to support these causes! And we’re happy if people do good work whether or not they agree with us on the philosophy.

But we still think the argument for longtermism is true, and we think it’s worth talking about.

Firstly, when we actually try to compare the importance of work in certain cause areas — such as global health or mitigating the risk of extinction from nuclear war — whether and how much we weigh the interests of future generations may play a decisive role in our conclusions about prioritisation.

Moreover, some longtermist priorities, such as ensuring that we avoid the lock-in of bad values or developing a promising framework for space governance, may be entirely ignored if we don’t consider the interests of future generations.

Finally, if it’s right that future generations deserve much more moral concern than they currently get, it just seems good for people to know that. Maybe issues will come up in the future that aren’t extinction threats but which could still predictably affect the long-run future – we’d want people to take those issues seriously.

In short, no. Total utilitarianism is the view that we are obligated to maximise the total amount of positive experiences over negative experiences, typically by weighting for intensity and duration.

This is one specific moral view, and many of its proponents and sympathisers advocate for longtermism. But you can easily reject utilitarianism of any kind and still embrace longtermism.

For example, you might believe in ‘side constraints’ — moral rules about what kinds of actions are impermissible, regardless of the consequences. So you might believe that you have strong reasons to promote the wellbeing of individuals in the far future, so long as doing so doesn’t require violating anyone’s moral rights. This would be one kind of non-utilitarian longtermist view.

You might also be a pluralist about value, in contrast to utilitarians who think a singular notion of wellbeing is the sole true value. A non-utilitarian might intrinsically value, for instance, art, beauty, achievement, good character, knowledge, and personal relationships, quite separately from their impact on wellbeing.

(See our definition of social impact for how we incorporate these moral values into our worldview.)

So you might be a longtermist precisely because you believe the future is likely to contain vast amounts of all the many things you value, so it’s really important that we protect this potential.

You could also think we have an obligation to improve the world for future generations because we owe it to humanity to “pass the torch”, rather than squander everything people have done to build up civilisation. This would be another way of understanding moral longtermism that doesn’t rely on total utilitarianism.²³

Finally, you can reject the “total” part of utilitarianism and still believe longtermism. That is, you might believe it’s important to make sure the future goes well in a generally utilitarian sense without thinking that means we’ll need to keep increasing the population size in order to maximise total wellbeing. You can read more about different kinds of views in population ethics here.

As we discussed above, people who don’t think it’s morally good to bring a flourishing population into existence usually think it’s still important to prevent future suffering — in which case you might support a longtermism focused on guarding against the worst outcomes for future generations.

No.

We believe, for instance, that you shouldn’t have a harmful career just because you think you can do more good than bad with the money you’ll earn. There are practical, epistemic, and moral reasons that justify this stance.

And as a general matter, we think it’s highly unlikely to be the case that working in a harmful career will be the path that has the best consequences overall.

Some critics of longtermism say the view can be used to justify all kinds of egregious acts in the name of a glorious future. We do not believe this, in part because there are plenty of plausible intrinsic reasons to object to egregious acts on their own, even if you think they’ll have good consequences. As we explained in our article on the definition of ‘social impact’:

We don’t think social impact is all that matters. Rather, we think people should aim to have a greater social impact within the constraints of not sacrificing other important values – in particular, while building good character, respecting rights and attending to other important personal values. We don’t endorse doing something that seems very wrong from a commonsense perspective in order to have a greater social impact.

Perhaps even more importantly, it’s bizarrely pessimistic to believe that the best way to make the future go well is to do horrible things now. This is very likely false, and there’s little reason anyone should be tempted by this view.

Some of the claims in this article may sound like science fiction. We’re aware this can be off-putting to some readers, but we think it’s important to be upfront about our thinking.

And the fact that a claim sounds like science fiction is not, on its own, a good reason to dismiss it. Many speculative claims about the future have sounded like science fiction until technological developments made them a reality.

From Eunice Newton Foote’s perspective in the 19th century, the idea that the global climate would actually be transformed based on a principle she discovered in a glass cylinder may have sounded like science fiction. But climate change is now our reality.

Similarly, the idea of the “atomic bomb” had literally been science fiction before Leo Szilard discovered the possibility of the nuclear chain reaction in 1933. Szilard first read about such weapons in H.G. Wells’ The World Set Free. As W. Warren Wager explained in The Virginia Quarterly:

Unlike most scientists then doing research into radioactivity, Szilard perceived at once that a nuclear chain reaction could produce weapons as well as engines. After further research, he took his ideas for a chain reaction to the British War Office and later the Admiralty, assigning his patent to the Admiralty to keep the news from reaching the notice of the scientific community at large. “Knowing what this [a chain reaction] would mean,” he wrote, “—and I knew it because I had read H.G. Wells—I did not want this patent to become public.”

This doesn’t mean we should accept any idea without criticism. And indeed, you can reject many of the more ‘sci-fi’ claims of some people who are concerned with future generations — such as the possibility of space settlement or the risks from artificial intelligence — and still find longtermism compelling.

One worry about longtermism some people have is that it seems to rely on having a very small chance of achieving a very good outcome.

Some people think this sounds suspiciously like Pascal’s wager, a highly contentious argument for believing in God — or a variant of this idea, “Pascal’s mugging.” The concern is that this type of argument may be used to imply an apparent obligation to do absurd or objectionable things. It’s based on a thought experiment, as we described in a different article:

A random mugger stops you on the street and says, “Give me your wallet or I’ll cast a spell of torture on you and everyone who has ever lived.” You can’t rule out with 100% probability that he won’t — after all, nothing’s 100% for sure. And torturing everyone who’s ever lived is so bad that surely even avoiding a tiny, tiny probability of that is worth the $40 in your wallet? But intuitively, it seems like you shouldn’t give your wallet to someone just because they threaten you with something completely implausible.

This deceptively simple problem raises tricky issues in expected value theory, and it’s not clear how they should be resolved — but it’s typically assumed that we should reject arguments that rely on this type of reasoning.

The argument for longtermism given above may look like a form of this argument because it relies in part on the premise that the number of individuals in the future could be so large. Since it’s a relatively novel, unconventional argument, it may sound suspiciously like the mugger’s (presumably hollow) threat in the thought experiment.

But there are some key differences. To start, the risks to the long-term future may be far from negligible. Toby Ord estimated the chance of an existential catastrophe that effectively curtails the potential of future generations in the next century at 1 in 6.²⁴

Now, it may be true that any individual’s chance of meaningfully reducing these kinds of threats is much, much smaller. But we accept small chances of doing good all the time — that’s why you might wear a seatbelt in a car, even though in any given drive your chances of being in a serious accident are miniscule. Many people buy life insurance to guarantee that their family members will have financial support in the unlikely scenario that they die young.

And while an individual is unlikely to be solely responsible for driving down the risk of human extinction by any significant amount (in the same way no one individual could stop climate change), it does seem plausible that a large group of people working diligently and carefully might be able to do it. And if the large group of people can achieve this laudable end, then taking part in this collective action isn’t comparable to Pascal’s mugging.

But if we did conclude the chance to reduce the risks humanity faces is truly negligible, then we would want to look much more seriously into other priorities, especially since there are so many other pressing problems. As long as it’s true, though, that there are genuine opportunities to have a significant impact on improving the prospects for the future, then longtermism does not rely on suspect and extreme expected value reasoning.