If there is a true moral theory, and if careful reflection has any tendency to move people toward moral truth, then people who think hard about population ethics should partially converge on the correct view. There are many ways to be wrong, but only one way to be right.

After I released a population ethics quiz in September, I looked at how people’s answers clustered together. What were the views where people had the strongest internal agreement?

Cross-posted to the Effective Altruism Forum.

Contents

Convergence on the total view

By far the strongest cluster was the total view, with 25% of respondents exactly agreeing with it—although a follow-up analysis suggests that 40% of those 25% probably hold negative-leaning views, which would bring the absolute percentage of totalists down to 15%.1 On the more conservative estimate, totalism still dwarfs the second-most popular exact answer profile, which is averagism at 3% of respondents. No other answer profile captured more than 2% of the people surveyed.2

Survey respondents were asked if they had pre-existing familiarity with the repugnant conclusion. Among those who said no, the most popular answer profiles represented single-digit percentages of people: 6% for the average view and 4% for the total view. Relatedly, Caviola et al. (2022)3 surveyed Mechanical Turk users and found that only 5.6% of respondents gave answers consistent with the total view.4 There isn’t a large cohort of “intuitive totalists”; it’s a view that people only arrive at when they’re familiar with population ethics.

The total view was also the modal answer set among familiar respondents: if you take the most popular answer to each individual question and put all those answers together, you get the total view.

This is convergence: among people who have thought about population ethics, a large share of them agree on the total view.5 There is no other view that people clearly cluster around. If some other view were correct, we would expect to see some visible convergence toward it.

Of course, this doesn’t prove that the total view is correct. We run into the is–ought problem: survey results are facts about people, not about the true nature of morality. But it’s evidence, as long as reflection tends to move people toward moral truth. If it doesn’t, then we have no reason to treat our beliefs as associated with truth in any way.

Alternative explanations of the survey results

I can think of three reasons why we would see these results even if totalism is false. The first two are incomplete explanations on their own, but the third explanation patches the holes in the first two.

  1. Sampling bias
  2. People tend toward simple views, even when they’re wrong
  3. The true theory of population ethics has yet to be discovered

Sampling bias

The survey sample was not representative—respondents disproportionately came from the EA Forum. Perhaps totalists are disproportionately attracted to effective altruism for reasons other than its truth,6 or effective altruists “convert” to totalism for social reasons.7 EA Forum readers are more thoughtful than most about ethics, but I don’t believe that’s the sole explanation for the prominence of the total view. For example, I expect totalism would be less popular among academic moral philosophers.

Supposing the survey sample had an unjustified bias toward the total view, if the true view isn’t totalism, then we would expect to see a second large cluster on the true view—perhaps smaller than the totalism cluster, but still noticeable. Instead, there were five times as many totalists as any other precise view, and no other position stands out as a clear second. (Averagism was the second-largest precise view, but it only beat #3 by a small margin—see Claude’s data analysis under “Whole answer profiles”.)

People tend toward simple views, even when they’re wrong

Many people find simple explanations more attractive—and there are good reasons to assume that simpler explanations are more likely to be correct. (Occam’s razor is the simple version.) But simple does not always mean correct. The total view is the simplest flavor of population ethics: you just add up everyone’s welfare.8 Perhaps people converge on totalism not because it’s true, but because it’s simple.

This explanation has the same problem as the previous one. It explains why the total view might be popular even if it’s wrong, but it still leaves us expecting to see a second cluster around the true theory of population ethics, and no such cluster emerges from the data.

The true theory of population ethics has yet to be discovered

For most of the 20th century, scientists could not agree on where the Moon came from. Did it form alongside Earth in the early days of the Solar System? Did it form elsewhere and then get captured by Earth’s gravity? Did early Earth spin so fast that it ejected enough matter to form the Moon? There was no scientific consensus. Then in 1975, Hartmann and Davis proposed that the Moon formed when a Mars-sized object collided with Earth. This was the first proposal that fit the evidence, and within a decade, the “giant impact hypothesis” became the dominant theory.9

Maybe population ethics has yet to have its “giant impact hypothesis” moment. Maybe in the future, some brilliant mind will discover the correct theory of population ethics, and then people will come to see the truth of this view. This would explain why we don’t see a second cluster around the true view: it’s because the true view has never been articulated.

Some evidence for this hypothesis is that population ethics is a young field, which received little attention until the last half-century or so. Some theories of population ethics were not described until the 21st century.10 It stands to reason that there are more views yet to be described, and the true view may lie among them.

Counterpoint: This population ethics survey does not ask people to describe theories, only to give answers to hypothetical scenarios. In science, researchers may agree on the observations without agreeing on a theory—the problem with the Moon’s formation was that (until Hartmann and Davis) nobody could come up with a theory that fit the known facts. By analogy, you would expect thoughtful people to converge toward a particular set of survey answers, even if they do not yet have a unified theory that explains them. If the survey results show any such convergence, then it’s toward the set of answers that match the total view, and the underlying theory is easy to see—it’s the total view.

How persuasive is the convergence argument?

If I didn’t have a strong position, this evidence would tilt me toward the total view, and give me a reason to extend more serious consideration toward the view.

I doubt this evidence is enough to convince someone who is committed to an alternative view—you could reasonably prefer a non-totalist view on its merits. From my perspective as a totalist, if the evidence had instead shown convergence around some other view, I would not have been swayed because I would still be able to articulate specific reasons why I believe the total view is correct. But it would’ve given me something to think about. (I would find myself faced with the task of explaining why people apparently converge toward a view that I disagree with.)

For a committed non-totalist, it seems difficult, but not impossible, to explain why their view doesn’t see greater convergence. Perhaps they could strengthen one of the alternative explanations described above, or they could explain why people have a strong unjustified bias against the correct view.

The evidence could be strengthened by surveying a different sample of people, such as academic philosophers. If the new sample also clusters around the total view, that strengthens the argument. If a second cluster emerges, then the convergence argument would support two candidate views.

Image credit: Lance Grandahl

Notes

  1. Quiz v2 had a smaller sample size, but it produced a similar figure: 14% of v2 answers exactly matched the total view. ↩

  2. I’m speaking of exact answer profiles: that is, sets in which every person gave the same answer to every question. You could also look at classes of views; for example, 5% of respondents aligned with a person-affecting view with a procreation asymmetry. But this is a collection of views, not a single view—these respondents did not agree on every question. ↩

  3. Caviola, L., Althaus, D., Mogensen, A. L., & Goodwin, G. P. (2022). Population ethical intuitions. doi: 10.1016/j.cognition.2021.104941 ↩

  4. I inferred this from the data on question 3b, which asked three questions, each comparing two populations with equal total utility but different average utility (e.g., 4,000 people at 60 welfare vs. 6,000 people at 40 welfare). A totalist would answer that the two populations are equally good; 5.6% of respondents gave that answer on every question, and 7.1% gave it on at least one of the three questions.

    Claude Opus 5.5 (who did the data analysis) claims that the original study included a coding error in which responses to a comprehension check question were assigned to the wrong code (respondents were reported as having passed the comprehension check when they failed, or vice versa). The 5.6% figure comes from the corrected version. I did not examine the data myself, but Claude’s reasoning passed my spot check. On the uncorrected version, the figure is 0.6%; for all respondents, it’s 3.5%. ↩

  5. I believe the important observation is that support for the total view is high among the familiar-with-population-ethics group, not necessarily that support is higher among “familiar” respondents than “unfamiliar” ones. The fact that there are few intuitive totalists is an interesting observation, but I don’t take it as much evidence either way. ↩

  6. I don’t have a strong hypothesis for why that might be. Effective altruist principles are obvious, yet accepted by few; those principles ought to be appealing to any ethical theory. But EA principles and the EA Forum are not the same thing. Perhaps the community discussion on the EA Forum disproportionately focuses on totalism-adjacent topics, and the vast majority of EA-principle-accepters (who donate to GiveWell charities, etc.) are non-totalists, are not interested in that kind of discussion, and therefore didn’t see my survey? ↩

  7. Beliefs are often socially driven, but I do not see much explicit discussion of population ethics. Perhaps people infer the total view from discussions of longtermism? ↩

  8. More accurately, totalism is the simplest plausible view. An example of a simpler view: “every outcome is exactly as good as every other outcome.” ↩

  9. My first draft used an anecdote about special relativity. Claude said I got some of my facts wrong, and suggested that the Moon example would work better. ↩

  10. Examples:

    • Broome (2004), Weighing Lives.
    • Asheim & Zuber (2014), “Escaping the repugnant conclusion: rank-discounted utilitarianism with variable population.”
    • Parfit (2016), “Can We Avoid the Repugnant Conclusion?” – The field’s founding figure was still publishing new work as recently as 2016.

    ↩