Note: The Global Priorities Institute (GPI) has started to create summaries of some working papers by GPI researchers with the aim to make our research more accessible to people outside of academic philosophy (e.g. interested people in the effective altruism community). We welcome any feedback on the usefulness of these summaries.
Summary: Staking our future: deontic long-termism and the non-identity problem (Andreas Mogensen)
In “The case for strong longtermism”, Greaves and MacAskill (2021) argue that potential far-future effects are the most important determinant of the value of our options. This is “axiological strong longtermism”. On some views, we can achieve astronomical value by making the future population of worthwhile lives much greater than it would otherwise have been. The question of whether it is intrinsically good to add lives worth living to the population is controversial, however. Greaves and MacAskill argue that the case for strong longtermism can also be made by focusing on the possibility of improving expected future well-being conditional on the existence of a large and roughly fixed-sized future population, that is, by focusing on prospects for generating large amounts of expected value by improvements to the average well-being of future people. In this form, the argument is said to be robust across plausible variations in population-ethical assumptions.
They also argue for “deontic strong longtermism”—the view that increasing the value of the far future is what we are morally required to do.
However, it may be that bringing about the most valuable outcomes is not always morally required. For this reason, axiological strong longtermism being right does not guarantee its deontic analogue is too. Greaves and MacAskill need an additional argument to lead us from the former to the latter.
In “Staking our future: deontic long-termism and the non-identity problem”, Mogensen argues no such argument will succeed, at least not insofar as the focus is on raising the average welfare level of future people to greater heights. This is because of an inevitable feature of improving the average well-being of the far future: it does not increase the well-being of anyone in particular.
The Non-Identity Problem
This feature is due to what Parfit (1984) called the “Non-Identity Problem”. This problem results from the nature of personal identity and reproduction. For mammals and many other living organisms, the combination of particular gametes (reproductive cells) determines the identities of those who will exist. Any particular gamete combination is a coincidence of timing that most likely could not have occurred if anything in the surrounding environment were different. Walking across the street and causing a stranger to pause can alter who will be born even in the near future since this is likely to change the timing of many conceptions. Interventions to improve the world many centuries from now affect the identities of everyone who will exist in the targeted time. Suppose we succeed in making the future happier. We are then not only responsible for a higher happiness average, but also for the individual makeup of the population enjoying extraordinary lives in the future.
Yet, for any individual, having an extraordinary life is not an improvement in well-being over never existing, since those who never live lack well-being altogether. Assuming the actions we take now lead to future people having lives that are much better than the lives that would have existed otherwise, these future people will not be better off than if we had done nothing. That is because the alternative would have been their non-existence rather than their having some lesser amount of well-being.
And so, increasing average future well-being does not increase the well-being of any particular person. Mogensen’s challenge for deontic strong longtermism is that increasing well-being in this impersonal way is not morally obligatory even though it may be the best thing we can do.
An imaginary woman named Hiroko helps make his case.
Suppose Hiroko can have either a mildly happy child or no child at all and that she will be equally content whichever she chooses. It intuitively seems Hiroko is morally permitted to have the mildly happy child, even if this child would be less happy than the world’s average (but still had a life worth living).
But what if Hiroko’s choice were instead between having a gloriously happy child and having no child at all, and that she would again be equally content either way? It intuitively seems that Hiroko is permitted to have no child, even though this is worse for well-being in the world.
Now suppose Hiroko is choosing between having a mildly happy child, having a different gloriously happy child, or having no child at all. Again, she will be equally content with any choice she makes. Is it still permissible for her to have a mildly happy child? Given our previous judgements, it is hard to see why not. If it is permissible for her to have no child instead of the gloriously happy child, and permissible for her to have a mildly happy child instead of no child, how could it be impermissible for her to have the mildly happy child instead of the gloriously happy child?
The judgement need not deny it is better for Hiroko to have the gloriously happy child. It does not even mean denying that she has a moral reason to have the gloriously happy child. What she lacks is an obligation to have the gloriously happy child instead of the mildly happy one. Perhaps this is because having a gloriously happy child instead of a (different) mildly happy one is not better for anyone in particular, and having a mildly happy child instead of a gloriously happy one is not worse for anyone in particular. While it is better to create the gloriously happy child rather than the mildly happy one or no child at all, adding to the welfare of the world in this impersonal way is morally good but not obligatory.
Justifications, requirements and the permissibility of a mildly happy future
Gert (2000, 2003) makes a relevant distinction in the roles and strengths of moral reasons. Moral reasons can justify certain responses and they can require certain responses, and both of these roles can vary in strength. For instance, we are required to keep a promise to a friend, but if it is a minor promise that the friend does not even remember, the requirement to keep it might be a weak one. This promise has weak “requiring strength.” An even weaker requirement, arguably, is saving a stranger’s life at a huge risk to oneself. Yet saving a stranger has so much “justifying strength” that it is permissible to break an important promise in order to save a stranger—even though keeping the promise would otherwise have been strongly required and saving the stranger was not required at all.
The permissibility of all three of Hiroko’s childbearing options might be because creating a gloriously happy life has a lot of justifying strength, and creating a mildly happy life has some but less justifying strength—while both have zero requiring strength. Hiroko is never required to have a child, but she can be justified in doing so; depending on the strength of this justification, it can outweigh other considerations, including moral requirements. If having a child would be a massive inconvenience to Hiroko and others and involve breaking many promises, yet she considers having one anyway, she may need to have the gloriously happy child to justify the costs to herself and others. However, if nothing weighs against Hiroko having a child, it could be permissible for her to have the mildly happy child instead of the gloriously happy one, because she does not need a strong justification.
How is this relevant for deontic strong longtermism? Imagine Hiroko chooses to have the mildly happy child instead of the gloriously happy child, and there is not enough weighing against her having a child to render this impermissible. So now she has one mildly happy child and is again deciding whether to have no child, a mildly happy child or a gloriously happy child—and she chooses to have another mildly happy child. It seems that so long as nothing much ever counts against her having a child, it is permissible for her to keep having mildly happy children instead of gloriously happy children until the end of time. This is so even though it would lead to a mildly happy future instead of a gloriously happy one. Furthermore, she would also be permitted to have no children whatsoever, even if this led to a far lower total of positive well-being in the world.
Conclusion
The implication for deontic strong longtermism is this: Increasing the far future’s well-being has a vast amount of justifying strength, perhaps even enough for us to neglect our moral duties such as improving the welfare of those who exist now. Yet because it does not improve well-being for anyone in particular, we are not required to help bring about the best possible future. So there is no sound argument leading from axiological strong longtermism to deontic strong longtermism, at least not insofar as the focus is on raising the average welfare level of future people to greater heights.
If we nevertheless dedicate ourselves to improving the far future, we will be acting in service to the highest good. However, if we choose not to, it does not mean we can just play video games instead. Without the justifying power of the far future, we may have no excuse to neglect our moral requirements to the here and now.
Paper summary: Staking our future: deontic long-termism and the non-identity problem (Andreas Mogensen)
Link post
Note: The Global Priorities Institute (GPI) has started to create summaries of some working papers by GPI researchers with the aim to make our research more accessible to people outside of academic philosophy (e.g. interested people in the effective altruism community). We welcome any feedback on the usefulness of these summaries.
Summary: Staking our future: deontic long-termism and the non-identity problem (Andreas Mogensen)
This is a summary of the GPI Working Paper “Staking our future: deontic long-termism and the non-identity problem” by Andreas Mogensen. The summary was written by Rhys Southan.
In “The case for strong longtermism”, Greaves and MacAskill (2021) argue that potential far-future effects are the most important determinant of the value of our options. This is “axiological strong longtermism”. On some views, we can achieve astronomical value by making the future population of worthwhile lives much greater than it would otherwise have been. The question of whether it is intrinsically good to add lives worth living to the population is controversial, however. Greaves and MacAskill argue that the case for strong longtermism can also be made by focusing on the possibility of improving expected future well-being conditional on the existence of a large and roughly fixed-sized future population, that is, by focusing on prospects for generating large amounts of expected value by improvements to the average well-being of future people. In this form, the argument is said to be robust across plausible variations in population-ethical assumptions.
They also argue for “deontic strong longtermism”—the view that increasing the value of the far future is what we are morally required to do.
However, it may be that bringing about the most valuable outcomes is not always morally required. For this reason, axiological strong longtermism being right does not guarantee its deontic analogue is too. Greaves and MacAskill need an additional argument to lead us from the former to the latter.
In “Staking our future: deontic long-termism and the non-identity problem”, Mogensen argues no such argument will succeed, at least not insofar as the focus is on raising the average welfare level of future people to greater heights. This is because of an inevitable feature of improving the average well-being of the far future: it does not increase the well-being of anyone in particular.
The Non-Identity Problem
This feature is due to what Parfit (1984) called the “Non-Identity Problem”. This problem results from the nature of personal identity and reproduction. For mammals and many other living organisms, the combination of particular gametes (reproductive cells) determines the identities of those who will exist. Any particular gamete combination is a coincidence of timing that most likely could not have occurred if anything in the surrounding environment were different. Walking across the street and causing a stranger to pause can alter who will be born even in the near future since this is likely to change the timing of many conceptions. Interventions to improve the world many centuries from now affect the identities of everyone who will exist in the targeted time. Suppose we succeed in making the future happier. We are then not only responsible for a higher happiness average, but also for the individual makeup of the population enjoying extraordinary lives in the future.
Yet, for any individual, having an extraordinary life is not an improvement in well-being over never existing, since those who never live lack well-being altogether. Assuming the actions we take now lead to future people having lives that are much better than the lives that would have existed otherwise, these future people will not be better off than if we had done nothing. That is because the alternative would have been their non-existence rather than their having some lesser amount of well-being.
And so, increasing average future well-being does not increase the well-being of any particular person. Mogensen’s challenge for deontic strong longtermism is that increasing well-being in this impersonal way is not morally obligatory even though it may be the best thing we can do.
An imaginary woman named Hiroko helps make his case.
Suppose Hiroko can have either a mildly happy child or no child at all and that she will be equally content whichever she chooses. It intuitively seems Hiroko is morally permitted to have the mildly happy child, even if this child would be less happy than the world’s average (but still had a life worth living).
But what if Hiroko’s choice were instead between having a gloriously happy child and having no child at all, and that she would again be equally content either way? It intuitively seems that Hiroko is permitted to have no child, even though this is worse for well-being in the world.
Now suppose Hiroko is choosing between having a mildly happy child, having a different gloriously happy child, or having no child at all. Again, she will be equally content with any choice she makes. Is it still permissible for her to have a mildly happy child? Given our previous judgements, it is hard to see why not. If it is permissible for her to have no child instead of the gloriously happy child, and permissible for her to have a mildly happy child instead of no child, how could it be impermissible for her to have the mildly happy child instead of the gloriously happy child?
The judgement need not deny it is better for Hiroko to have the gloriously happy child. It does not even mean denying that she has a moral reason to have the gloriously happy child. What she lacks is an obligation to have the gloriously happy child instead of the mildly happy one. Perhaps this is because having a gloriously happy child instead of a (different) mildly happy one is not better for anyone in particular, and having a mildly happy child instead of a gloriously happy one is not worse for anyone in particular. While it is better to create the gloriously happy child rather than the mildly happy one or no child at all, adding to the welfare of the world in this impersonal way is morally good but not obligatory.
Justifications, requirements and the permissibility of a mildly happy future
Gert (2000, 2003) makes a relevant distinction in the roles and strengths of moral reasons. Moral reasons can justify certain responses and they can require certain responses, and both of these roles can vary in strength. For instance, we are required to keep a promise to a friend, but if it is a minor promise that the friend does not even remember, the requirement to keep it might be a weak one. This promise has weak “requiring strength.” An even weaker requirement, arguably, is saving a stranger’s life at a huge risk to oneself. Yet saving a stranger has so much “justifying strength” that it is permissible to break an important promise in order to save a stranger—even though keeping the promise would otherwise have been strongly required and saving the stranger was not required at all.
The permissibility of all three of Hiroko’s childbearing options might be because creating a gloriously happy life has a lot of justifying strength, and creating a mildly happy life has some but less justifying strength—while both have zero requiring strength. Hiroko is never required to have a child, but she can be justified in doing so; depending on the strength of this justification, it can outweigh other considerations, including moral requirements. If having a child would be a massive inconvenience to Hiroko and others and involve breaking many promises, yet she considers having one anyway, she may need to have the gloriously happy child to justify the costs to herself and others. However, if nothing weighs against Hiroko having a child, it could be permissible for her to have the mildly happy child instead of the gloriously happy one, because she does not need a strong justification.
How is this relevant for deontic strong longtermism? Imagine Hiroko chooses to have the mildly happy child instead of the gloriously happy child, and there is not enough weighing against her having a child to render this impermissible. So now she has one mildly happy child and is again deciding whether to have no child, a mildly happy child or a gloriously happy child—and she chooses to have another mildly happy child. It seems that so long as nothing much ever counts against her having a child, it is permissible for her to keep having mildly happy children instead of gloriously happy children until the end of time. This is so even though it would lead to a mildly happy future instead of a gloriously happy one. Furthermore, she would also be permitted to have no children whatsoever, even if this led to a far lower total of positive well-being in the world.
Conclusion
The implication for deontic strong longtermism is this: Increasing the far future’s well-being has a vast amount of justifying strength, perhaps even enough for us to neglect our moral duties such as improving the welfare of those who exist now. Yet because it does not improve well-being for anyone in particular, we are not required to help bring about the best possible future. So there is no sound argument leading from axiological strong longtermism to deontic strong longtermism, at least not insofar as the focus is on raising the average welfare level of future people to greater heights.
If we nevertheless dedicate ourselves to improving the far future, we will be acting in service to the highest good. However, if we choose not to, it does not mean we can just play video games instead. Without the justifying power of the far future, we may have no excuse to neglect our moral requirements to the here and now.
References
Joshua Gert (2000). Practical rationality, morality, and purely justificatory reasons. American Philosophical Quarterly, 37, 227–243.
Joshua Gert (2003). Requiring and justifying: two dimensions of normative strength. Erkenntnis, 59, 5–36.
Hilary Greaves and William MacAskill, W. (2021). The case for strong longtermism. GPI Working Paper (No. 5-2021).
Andreas Mogensen (2019). Staking our future: deontic long-termism and the non-identity problem. GPI Working Paper (No. 9-2019).
Derek Parfit (1984). Reasons and persons. Oxford: Oxford University Press.