The Billions Barely Alive: Parfit's Repugnant Conclusion and the Ethics of Future Generations

The Hook: Two Worlds and an Uncomfortable Arithmetic

Imagine two possible futures for humanity.

In the first world — call it A — ten billion people live. Every one of them leads a rich, fulfilling life: good health, deep relationships, meaningful work, leisure, art, love. No war, no hunger, no despair. It is, as far as one can picture it, a flourishing civilization.

In the second world — call it Z — a hundred trillion people live. But their lives are different. They are just barely worth living — that kind of existence in which the good moments narrowly outnumber the bad, in which you eat potatoes and occasionally see a sunset, but little beyond bare endurance. No leisure, no art, no great joys. Only enough wellbeing that these people, if you asked them, would still faintly affirm their lives.

Now the question: which world is better?

Almost everyone answers without hesitation: A, obviously. A smaller number of people who truly live well is preferable to an enormous mass of the barely-existing. The second world looks like a dystopia, an anthill of insignificance.

And yet: if one follows one of the most plausible and widely held moral accounting rules — that the better outcome is the one containing more good in total — then Z is better than A. For a hundred trillion lives, each with a small positive value, sum to a far greater total quantity of wellbeing than ten billion very good lives. Sheer number beats quality.

This conclusion carries a name that anticipates its effect: the Repugnant Conclusion. It was precisely formulated in 1984 by the British philosopher Derek Parfit in his work Reasons and Persons, and it has since plunged moral philosophy into a state of emergency that remains unresolved to this day. Its explosive force lies not in asserting a strange idea — but in the fact that it follows, compellingly, from apparently harmless steps that are individually almost impossible to deny. And, as will become clear at the end, it is no academic curiosity: it sits in the engine room of every decision that affects the number of future people — from climate policy, through the assessment of existential risk, to the question of what goals we should give a future superintelligence at all.

The Core Concept: Population Ethics and Totalism

To understand why the Repugnant Conclusion is a problem, one must first step onto the field where it grows: population ethics. It is the branch of moral philosophy that evaluates outcomes differing not only in how well off people are, but in how many people exist at all and who they are.

Classical ethics usually assumes, tacitly, a fixed population. Whether I keep or break my promise changes nothing about which people there are. Parfit called such cases "Same People Choices" — decisions where the same persons exist, however we act. But many of humanity's most consequential decisions are of a different kind. Whether and how strongly we protect the climate, whether we preserve or consume resources, what population policy a state pursues — all of this affects not only how well future people live, but which and how many people are born at all. Parfit called these cases "Different People Choices." And for these, the familiar morality fails.

The most obvious answer is what one calls totalism or the total view, the classical position of utilitarianism for variable populations: the best outcome is the one containing the greatest total sum of wellbeing. One adds up the wellbeing value of every single life and chooses the world with the highest sum.

Totalism has enormous virtues. It is simple, egalitarian (every life counts equally), and it takes the future seriously: people who do not yet exist but could exist do not simply drop out of the moral accounting. Yet precisely this last move — that additional happy lives make the outcome better — is the door through which the Repugnant Conclusion strolls in. For if more lives mean more value, then a small loss in quality of life can always be offset by a sufficient increase in number. And if you repeat that often enough, you land at the trillions at subsistence level.

A note on terminology, which often causes confusion: to speak of the "life barely worth living" presupposes a neutral level of wellbeing — a zero point above which a life is, for the person living it, good on balance, and below which it is bad. Exactly where this zero point lies is itself contested. But without the assumption that wellbeing has such a scale with a change of sign, one cannot do population ethics at all.

Part 1: The Mere Addition Paradox

Parfit's real achievement was not simply to assert the Repugnant Conclusion. It was to show that you cannot get rid of it by rejecting totalism. For it follows from a chain of individual judgments that almost everyone would sign — including those who reject totalism. This chain is called the Mere Addition Paradox.

It works with a handful of fictional populations. Picture them as bars whose width indicates the number of people and whose height indicates their wellbeing.

Population A: a relatively small group, everyone at a very high level of wellbeing, all equal. A narrow, tall bar. A happy, equal society.

Population A+: it arises from A by mere addition. All the people from A remain, at exactly the same high level. Added to them is a new, separate group of people whose lives are worth living — that is, positive — but at a lower level than that of the original group. Graphically: the tall bar of A, next to it a second, wider but lower bar. These two groups are, in Parfit's construction, separate from one another; they know nothing of each other, no wellbeing flows from one to the other.

First intuition: A+ is no worse than A. Why should it be? No one in A is worse off. And the added people lead lives that are good for them — it is better for them to exist than not to exist. Nothing has been taken away, only something positive added. How could a world become worse by having additional people with lives worth living come into it, without anyone else suffering for it? (Whoever denies this must hold the uncomfortable thesis that the mere existence of contented people makes the world worse.)

Population B: now take A+ and change it in two ways. First: equalize the wellbeing, so that everyone — the original and the added group — lives at the same level. Second: set this shared level so that the total sum and the average of wellbeing are higher than in A+. B is thus a single, wide bar, slightly lower than the tall A bar, but distinctly higher than the low added bar of A+ — and since it encompasses everyone, its area (the total welfare) is greater than that of A+.

Second intuition: B is better than A+. For B has more total wellbeing, a higher average, and is perfectly equal — it has abolished the inequality of A+. In every respect that counts, B stands better. Anyone who places even a little value on equality or on the total sum must prefer B.

And now the trap snaps shut. From "A+ no worse than A" and "B better than A+" it follows, by simple transitivity (if X is at least as good as Y and Y is better than Z, then X is better than Z): B is better than A. A larger population at a somewhat lower but equal level is better than the small, highly happy original society.

One can guess how it continues. B now plays the role of A. One adds a further, still larger group at a lower level (B+), equalizes and improves again (C), and C is better than B. Then D, then E … Step by step one descends a staircase: each step has more people and a slightly lower level, and each is better than the previous. At the bottom of this staircase stands Z — the vast population whose lives are just barely worth living. And Z is, step by step, better than A.

That is the Repugnant Conclusion, in Parfit's own canonical formulation:

"For any possible population of at least ten billion people, all with a very high quality of life, there must be some much larger imaginable population whose existence, if other things are equal, would be better even though its members have lives that are barely worth living." (Parfit, Reasons and Persons, 1984)

Part 2: Why This Is a Genuine Paradox

The reason this shook moral philosophy, rather than being a mere arithmetical curiosity, lies in its structure. Parfit had built not a mere argument but exposed an incompatibility. Four claims, each highly plausible on its own, cannot all be true at once:

1. Mere addition is not bad: adding additional people with lives worth living, without harming anyone, does not make a world worse.
2. Equality and total sum count: a more equal world with higher total and average welfare is better.
3. Transitivity of "better than": the better-than relation is transitive — if X is better than Y and Y better than Z, then X is better than Z.
4. Rejection of the Repugnant Conclusion: the small, highly happy world A is better than the vast, barely-worth-living world Z.

You cannot have all four. Whoever accepts 1, 2 and 3 must give up 4 and swallow the Repugnant Conclusion. Whoever wants to save 4 must sacrifice one of the first three — and giving up any of them carries a high price. This is exactly what makes the problem so stubborn: there is no cheap way out. Every "solution" is a surrender of something nearly as dear to us as what we are trying to save.

The rest of the history of population ethics is essentially the history of the attempt to turn exactly one of these four screws without the whole edifice collapsing. As we will see, it collapses every time — just in a different spot.

Part 3: The Escape Attempts and Their Prices

Escape route 1: Average utilitarianism

The most obvious way out: if summing drives us to Z, then do not sum — average. According to average utilitarianism (averagism), the best world is the one in which the average wellbeing is highest. Z, where everyone lives near subsistence, has a miserable average; A, with nothing but highly happy people, an excellent one. The Repugnant Conclusion seems defeated.

But the average leads into abysses that may be worse. First: it implies that it can be bad to bring a happy child into the world — namely, whenever that child's life, however good, slightly lowers the existing average. A world that books additional happiness as a loss deeply violates our sensibility.

Second, and even more devastating, the Sadistic Conclusion, which Gustaf Arrhenius worked out: under certain circumstances the average view rates adding a few people with negative wellbeing — that is, lives of suffering, not worth living — as better than adding very many people with positive, if low, wellbeing. The reason: the many moderately happy lives drag the average down more than the few suffering ones. A theory that under some circumstances prefers creating suffering to creating happiness has discredited itself. The average has traded the Repugnant Conclusion for something even more repugnant.

Escape route 2: Critical-level utilitarianism

A more refined attempt shifts the zero line. Critical-level utilitarianism does sum, but it counts toward the total value only what a life possesses in wellbeing above a critical threshold — a level clearly above bare subsistence. Lives that are worth living but do not reach this threshold contribute negatively to the balance. Thus billions of barely-worth-living lives no longer sum to a plus but to a minus — and Z is no longer better than A. Set the critical threshold high enough, and the Repugnant Conclusion vanishes.

The price: this approach too leads into its own version of the Sadistic Conclusion. If a positive critical threshold holds, then adding people whose lives are worth living but below the threshold counts as bad. Under some circumstances it then rates better to create a world with a few suffering people (negative wellbeing) than a world with very many people whose lives are faintly positive but below the threshold. Once again one has treated lives that are good for those who live them as though they were an evil. Refinements such as the critical-range theories (which posit, instead of a sharp threshold, a whole range of neutral evaluation) soften some of these consequences but take on others in exchange, such as incompleteness of the ranking.

Escape route 3: Person-affecting views and the Non-Identity Problem

Perhaps the most intuitive way out attacks the whole thing at its root. It appeals to a slogan that sounds so evident one can hardly deny it — its most famous version comes from Jan Narveson: "We are in favor of making people happy, but neutral about making happy people." Put differently: morality exists for the sake of existing (or anyway future) persons. Creating an additional life is in itself neither good nor bad; only what is good or bad for someone can be good or bad. This family of positions is called person-affecting views. Their core: "What is bad must be bad for someone."

If that is right, the Mere Addition Paradox collapses at the very first step: the added people in A+ do not make the world better, because a world cannot be better for them — without them, they would not exist at all. The Repugnant Conclusion loses its engine.

But here Parfit's second great discovery lurks: the Non-Identity Problem. It shows that the person-affecting view shatters on quite ordinary cases. Imagine a society pursuing a policy of resource depletion that lays the planet waste in three hundred years. One would say: that is bad, because it harms future people. But the Non-Identity Problem counters: a policy so far-reaching changes, over the centuries, who is born at all. Different people meet, different children are conceived (a shift of even seconds in the moment of conception yields an entirely different person — a butterfly effect of identity). The people who, in three hundred years, live in the devastated world are not the same who would have lived in a protected world. They exist only because the harmful policy was pursued. So long as their lives are still worth living, they have therefore not been harmed — without the depletion they would not exist. The person-affecting view must consequently judge: the devastation of the planet harms no one and is therefore not bad. That is plainly absurd.

Parfit drew a far-reaching conclusion from this: the moral theory we seek — he called it "Theory X" — "will not take a person-affecting form." We must admit impersonal values, that is, acknowledge that some outcomes are simply better or worse without being better or worse for a particular person. But as soon as we do that, we are back in the realm where the Repugnant Conclusion rules. Parfit himself never found Theory X. He died in 2017 without the solution he had worked on for over thirty years.

Part 4: Arrhenius's Impossibility Theorems

Perhaps the search for the one correct theory was doomed from the start. This disquieting possibility has been investigated with mathematical rigor by the Swedish philosopher Gustaf Arrhenius. His result is a series of six impossibility theorems that mark the current state of research.

An impossibility theorem has the same construction as famous results of social choice theory — for instance Arrow's impossibility theorem, which shows that no voting method can simultaneously satisfy a small number of self-evident fairness conditions. Arrhenius transfers this idea to population ethics. He formulates a handful of adequacy conditions — requirements that any acceptable population theory ought to satisfy, because each one is individually highly plausible. Roughly and simplified, these include conditions such as:

• Avoidance of the Repugnant Conclusion — A should be better than Z.
• Avoidance of the Sadistic Conclusion — it must never be better to create suffering than happiness.
• Avoidance of the "anti-egalitarian" conclusion — a more equal distribution must not fare worse than a more unequal one with the same sum.
• A dominance or addition condition — if every person in a population is better off, that population should be better.
• Transitivity — the ordering must not run in a circle.

Arrhenius's theorems now prove: these conditions are mutually incompatible. There is no population theory — none, however ingenious — that satisfies them all at once. Every possible position must violate at least one of the conditions and thereby accept at least one of the "counterintuitive conclusions." The Repugnant Conclusion is therefore not an isolated defect of totalism that could be fixed with the right theory. It is one of several cliffs, and the theorem says: on at least one of them you must be wrecked.

This changes the character of the whole debate. It is no longer about the question "which theory is correct?" but about the far more bitter one: "which of the unbearable consequences is the least unbearable?" Some philosophers draw radical conclusions from this. Some see in the impossibility theorems an argument for moral skepticism or anti-realism — the suspicion that there is simply no objective truth about which population is better, because our intuitions do not even possess a consistent core. Others counter that the impossibility only shows that one of our apparently self-evident conditions is after all false and must be given up — we just do not yet know which.

Part 5: Inescapability — and Why Parfit Came to Accept It

For a long time population ethics lived in the hope that at least the Repugnant Conclusion — the most offensive of all — could be specifically avoided, even if one had to swallow other toads. This hope has been formally shaken by the economists and philosophers Mark Budolfson and Dean Spears in a series of works.

Their finding: many theories commonly regarded as "escape routes" from the Repugnant Conclusion do not in truth escape it — they merely avoid the one narrowly defined standard version while implying an equally repugnant extended version. The trick of the previous "avoidances," Budolfson and Spears show, often consisted in drawing the formal definition of the Repugnant Conclusion so narrowly that it excludes precisely those cases in which one's own theory fails (for instance cases with an unaffected sub-population). Relax these restrictions — arbitrary from the standpoint of "repugnance" — and it follows: practically all leading welfarist axiologies imply a Repugnant Conclusion — average utilitarianism as much as the elaborate approaches, indeed even theories that are incomplete, intransitive, rank-dependent, person-affecting or pluralist. The Repugnant Conclusion is, as the title of one of their works puts it, inescapable.

If that is right, the perspective shifts once more. The question is no longer whether one accepts a repugnant consequence, but only which. And it is exactly at this point that some philosophers have begun to think the unthinkable: perhaps the Repugnant Conclusion is not as repugnant as its name suggests.

For what does "a life just barely worth living" actually mean? Our intuition immediately paints a picture of misery, grey and privation. But by definition such a life is one in which the good outweighs the bad — a life that the person living it affirms on balance and whose existence she prefers to non-existence. If one pictures Z as a hundred trillion such — modest but affirmed — lives, rather than a hundred trillion grey ants, the picture loses some of its terror. Possibly, some suspect, it is our imagination that deceives us, not our ethics: we involuntarily picture "barely worth living" as "barely bearable" and thus confuse a modest good with an evil. I am of the opinion that there is something correct in this suspicion — that part of the "repugnance" rests on a failure of imagination — but I also hold it to be dishonest to explain away the whole unease with it: even after correcting the picture, the core remains, that pure quantity is supposed to be able to override any quality, however high, and that many find, rightly, hard to bear.

Parfit himself wrestled with this problem to the end. In his late works he increasingly approached the view that the Repugnant Conclusion may have to be accepted — that our reluctance might be a weakness of our intuition rather than a fault of the theory. He went on searching for a "Theory X" that satisfies all reasonable conditions, but he died without having found it, and in the awareness that Arrhenius's theorems suggest it may not exist at all.

Part 6: Why This Is Not an Academic Game

One might take all of this for an elegant edifice of thought without bearing on reality. That would be a mistake. Population ethics has, in the last two decades, migrated from the seminar room into practice — above all through the movement of effective altruism and the thesis of longtermism.

Longtermism — the view that positively influencing the long-term future is a central moral priority of our time — takes seriously exactly those impersonal, aggregative values that lead to the Repugnant Conclusion. Its argument runs, roughly: the number of people who could still live in the future is astronomical — possibly trillions over millions of years. If these future lives count morally, then almost everything we do today is, compared with its effects on that immense future, of secondary importance. In particular, the reduction of existential risks — dangers that could wipe out humanity entirely or permanently curtail its potential — becomes the overriding duty, because their occurrence would destroy not merely the present but an immeasurable number of possible future lives.

Yet the same logic has a disturbing flip side, which has been called the Altruistic Repugnant Conclusion: if additional worthwhile lives make the universe better, then the most urgent moral task might consist in maximizing the number of future people — filling the galaxy with as many, as frugal existences as possible, rather than enabling a smaller number to live a truly rich life. Whoever takes longtermism seriously must ask whether they are not thereby readmitting the billions at subsistence through the back door. The debate over the Repugnant Conclusion is thus directly action-guiding: it helps decide whether "more people" is a value in itself or not.

Three concrete fields in which the same structure recurs:

Climate policy. Virtually every major climate-policy decision is a "Different People Choice": it changes not only how well future people live, but which and how many are born. The Non-Identity Problem sharply raises the question here of what our duty to the future is actually grounded in, if strictly speaking we do not harm the concrete future people, because without our policy they would not exist at all.

Population policy and distribution. Whether a smaller, wealthier or a larger, poorer world population is "better" is not a merely economic question — it hangs inextricably on exactly the axiology Parfit laid bare.

AI value alignment. Whoever wants to give a future, very powerful AI an objective function that "maximizes the good" must inevitably build a population axiology into it — implicitly or explicitly. A naively totalist objective function might drive a superintelligence to pave the universe with a maximal number of barely-worth-living existences. Arrhenius's impossibility theorems have therefore entered the debate on AI safety: they show that there can be no mathematically clean utility function, satisfying all intuitions, for the number and wellbeing of future beings — a warning to anyone who believes ethics can simply be delegated to a machine as an optimization problem. The same tension between contradictory yet individually plausible principles also meets us in Newcomb's Paradox, where dominance and expected utility collide.

The Central Takeaway

The real lesson of the Repugnant Conclusion is not that we should wish trillions of people to subsistence — no one does. Nor is it that ethics is impossible. The lesson is subtler and more practical at once: our moral intuitions do not form a consistent system. We carry with us a handful of convictions, each of which seems beyond doubt — that additional happiness cannot harm, that equality and total welfare count, that comparisons are transitive, that the small happy world is better than the vast barren one — and population ethics proves, with the rigor of a mathematical theorem, that they cannot all be true at once.

For anyone who deals professionally with optimization, metrics and objective functions — from system architecture through security assessment to AI — there is a tangible warning in this. As soon as one presses "the good" into a single quantity to be maximized, one smuggles in, tacitly, a whole theory of value, with all the consequences one would never sign elsewhere. Totalism in ethics is the sister of the naive metric in engineering: both reward what can be added up, and both risk maximizing the wrong thing the moment one trusts them blindly. The consequence is not to give up measuring, but to know that every metric carries an implicit value decision — and to make that decision consciously, rather than leaving it to the formula. Whoever has ever defined an objective function has, without knowing it, touched exactly the question Parfit worked on for thirty years.

Reflection Question

If you are honest with yourself: which of the four premises of the Mere Addition Paradox would you sacrifice to avoid the Repugnant Conclusion — and are you prepared to pay the price its abandonment demands? And one level deeper: should the mere number of contented people be a moral value in itself at all — or does only how well off those who will exist anyway matter in the end?

Cross-References in the Vault

• Moral Luck — as there, this is about the conflict between a clean moral principle and a practice of judgment that eludes it.
• The Trolley Problem and Autonomous Vehicles — the struggle between aggregative (utilitarian) and non-aggregative thinking in a concrete technical application.
• Newcomb's Paradox — another example of how individually plausible principles lead to incompatible conclusions.

Sources

• Parfit, Derek: Reasons and Persons, Oxford University Press 1984, Part IV ("Future Generations"), esp. chs. 17–19 and Appendix I.
• Zuber, S. et al. / Arrhenius, G.: "The Repugnant Conclusion", Stanford Encyclopedia of Philosophy — https://plato.stanford.edu/entries/repugnant-conclusion/
• "Mere addition paradox", Wikipedia — https://en.wikipedia.org/wiki/Mere_addition_paradox
• "Population Ethics", Utilitarianism.net — https://utilitarianism.net/population-ethics/
• "The Nonidentity Problem", Stanford Encyclopedia of Philosophy — https://plato.stanford.edu/entries/nonidentity-problem/
• Thomas, Teruji: "Reconstructing Arrhenius's Impossibility Theorems", University of Oxford — https://users.ox.ac.uk/~mert2060/webfiles/Reconstructing-Arrhenius-for-web.pdf
• Budolfson, Mark & Spears, Dean: "Repugnant Conclusions" / "Why the Repugnant Conclusion is Inescapable" — https://philpapers.org/rec/BUDWTR
• Moberger, Victor: "Impossible Ethics: Do Population Ethical Impossibility Results Support Moral Skepticism and/or Anti-Realism?", Pacific Philosophical Quarterly 2024 — https://onlinelibrary.wiley.com/doi/10.1111/papq.12462

Note: where claims were made about the "correct" way to handle the Repugnant Conclusion, these are positions contested in the scholarly literature; they are marked as such in the text, or flagged with "I am of the opinion that ..." as the author's own assessment.

← All articles