Population ethics · interactive

The Shape of a Good World axiologies over populations, and the paradoxes that break them

An axiology is a theory of betterness — a ranking of possible worlds by how good they are. When worlds can differ in who and how many people exist, every candidate ranking collides with an intuition we are loath to give up. This page lets you build populations, watch six axiologies disagree about them, and walk through the classic results — the Repugnant Conclusion, the Mere Addition Paradox, the Sadistic Conclusion — that culminate in Arrhenius's impossibility theorems.

0The welfare scale & the neutral level

Throughout, each person has a welfare level on a cardinal scale with a natural zero. Welfare 0 is the neutral level: a life neither good nor bad for the one living it. Positive welfare = a life worth living; negative welfare = a life worth not living (net suffering). A life "barely worth living" sits just above zero. In the diagrams below a world is drawn as blocks: width = number of people, height = welfare per person, so a block's signed area = that group's total welfare, and the dashed line is the neutral level.

Two modelling choices, stated up front so the maths is checkable: welfare is cardinal and interpersonally comparable (we add and average it), and the neutral level is fixed at 0. Both are substantive assumptions in this literature; every formula on the page respects them.

1Build two worlds — watch the axiologies disagree

Drag a block's right edge to change how many people are in that group, or its top edge to change their welfare. Add or remove groups with the buttons. Both worlds share the same axes, so equal areas mean equal total welfare. The table recomputes live: watch a single pair of worlds get ranked in opposite directions by different theories of value.

World A
World B
welfare > 0 (worth living) welfare < 0 (worth not living) neutral level (0) drag edges to edit

Axiology parameters

AxiologyValue · AValue · B Ranks betterReads the numbers as…

What to look for. The default worlds — a small élite at high welfare (A) versus a large crowd at low but positive welfare (B) — are ranked A-better by average, maximin, critical-level and (for enough dampening) variable-value, yet B-better by the total sum — and, at the default priority strength, by prioritarianism too (a concave transform still rewards B's far larger population). That disagreement is the seed of every paradox below. Push the critical level or γ up and watch views cross over.

2The Repugnant Conclusion

Parfit (1984). “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.” Slide from world A (few people, high welfare) toward world Z (astronomically many, each life barely worth living), holding the interpolation so that total welfare rises while average welfare falls. Total utilitarianism must call every step an improvement on A; average utilitarianism must call every step a loss.

The two worlds (A fixed · Z morphing)

Height = welfare per person (honest linear scale). Width uses a log scale because populations span 100 → 100,000; areas here are not to scale — read the totals from the readout and the charts.

How each theory scores the morph
Total welfare
Average welfare

Two panels, two honest y-axes — never a shared dual axis. The dashed line marks world A's level. Total only ever climbs above it; average only ever sinks below.

Why total-util is forced into it. Value is V = Σ wᵢ. Adding people with any positive welfare ε adds ε each, so a large enough population of barely-worth-living lives sums past any fixed A. Why average-util escapes it. Value is V = (Σ wᵢ)/n, which depends only on the welfare level, not the count — a lower level is simply worse. Section 4 shows the price average-util pays for that escape.

3The Mere Addition Paradox

Parfit (1984). The Repugnant Conclusion can be argued for without assuming total utilitarianism, using three worlds and a chain of hard-to-deny comparisons. Step through them.

The load-bearing step is A+ is not worse than A. A+ merely adds a group of people with lives worth living, off to one side, making no one who already existed worse off. To deny this you must say the extra inequality by itself makes A+ worse — i.e. be anti-egalitarian about mere addition. Grant the step, grant that the more-equal, higher-total B beats A+, and transitivity hands you B ≥ A. Iterate and you descend to Z — the Repugnant Conclusion, with no summation assumed. Average utilitarianism blocks the argument by denying the first step (A+ has a lower average than A), which is exactly why it is vulnerable in §4.

4The Sadistic Conclusion

Arrhenius (2000). The theories that dodge the Repugnant Conclusion tend to buy their escape with something worse: they can rank adding a few miserable lives as better than adding many good ones. Start from a fixed background population C, then choose an addition.

Add to C:
The mechanism. Average-util prefers the miserable addition because a few lives a little below the average drag it down less than many lives far below it. Critical-level util (here c = 5) prefers it because every added life scores wᵢ − c: fifty lives at 1 each score 50·(1−5) = −200, worse than ten lives at −1 each, 10·(−1−5) = −60. Total utilitarianism is immune — it scores the additions +50 and −10 and picks the happy lives — but total-util is the one stuck with §2. That trade is the whole subject.

5The Non-Identity Problem

Parfit (1984) (anticipated by Schwartz, Adams and Kavka). Many choices — social policies, when to conceive — change which people come to exist, not just how well off a fixed group is. A "depletion" policy might leave future people worse off than an alternative would have — yet those very people owe their existence to it; had we chosen otherwise, different people would exist. So the policy is worse for no one: no one who exists under it would have been better off had we chosen differently (they simply would not have been). This guts person-affecting intuitions — “an outcome can be worse only if it is worse for someone” — which is why person-affecting views struggle to say depletion is wrong, and why they fail completeness/transitivity in the matrix below.

6The axiologies, side by side

Each view, its value function, the intuition it honours, and the bullet it cannot dodge. n = population wᵢ = welfare of person i

7The impossibility matrix

This is the punchline. List a handful of individually compelling adequacy conditions along the top and the axiologies down the side. Hover a cell for the reason. Every row has at least one ✗ — no axiology satisfies them all. That is the shape (though not the letter) of Arrhenius's impossibility theorems: the intuitions we bring to population ethics are jointly unsatisfiable.

✓ satisfies✗ violates — hover any cell for the specific reason
Honest caveat about the framing. The five columns are a pedagogically simplified, self-contained selection, not a verbatim reproduction of one numbered theorem. Arrhenius's actual impossibility results (Arrhenius 2000; “The Impossibility of a Satisfactory Population Ethics”, 2011) turn on a more technical condition set — e.g. Egalitarian Dominance, General Non-Extreme Priority, Non-Elitism, Weak Non-Sadism, Weak Quality Addition, and a Dominance-Addition/Mere-Addition principle. The robust, well-established claim reproduced here is the structural one — no axiology satisfies every plausible adequacy condition at once. The particular ✓/✗ pattern below follows directly from each value function (all cells are derivable from §6's formulas), but readers checking against Arrhenius should map these columns onto his, not assume a one-to-one match. Two rows carry an asterisk where a view's verdict is genuinely contested — see their tooltips.
Built as a self-contained page — no network, no external assets (bar the SVG XML namespace). Charts follow a colour-blind-safe scheme (blue/orange welfare sign; violet/aqua series), reinforced by position, dashes and direct labels. Sources named inline: Parfit, Reasons and Persons (1984); Arrhenius, Population Ethics (2000/forthcoming); Blackorby–Bossert–Donaldson on critical-level utilitarianism; Ng (1989) on number-dampened / variable-value utilitarianism; Narveson (1967/73) on person-affecting views. Nothing here is a substitute for those texts.