Contribution
It is the first study of how laypeople judge the badness of human extinction. It takes Parfit’s thought experiment from Reasons and Persons and asks people the question Parfit predicted they would answer wrongly. It then varies the vignette to find out why they answer as they do. It separates two reasons extinction could be bad: the deaths of everyone now alive, and the loss of humanity’s whole future.
Key insight
Ordinary concern about extinction is concern for the people who die, not for the people who will never live. Asked plainly, most people put the badness of extinction in its victims. Concern for the lost future appears when the future is pointed out, and most strongly when it is described as good. “An important reason why people do not find extinction uniquely bad is that they focus on the immediate death and suffering that the catastrophes cause for fellow humans, rather than on the long-term consequences” (p. 1).
Assumptions
- Parfit’s comparison measures unique badness. Extinction is “uniquely bad” if the difference between 80% and 100% dying is judged greater than the difference between no catastrophe and 80% (pp. 1–2). Parfit used peace, a nuclear war and 99%; the authors use “no catastrophe”, “a catastrophe” and 80%, so that recovery is plausible (p. 6).
- Only participants who rank no catastrophe best and extinction worst are asked the key question (p. 7). Others are dropped from it, so every percentage below is of that subset.
- The long-term view is the reflective one. The authors take Parfit’s and Bostrom’s view, that the future’s length and quality make extinction uniquely bad, as the position to be explained. Their Discussion suggests lay answers “should not necessarily be seen as reflecting people’s well-considered views” (p. 5).
- A forced choice between two differences, with no option for “equal” or “cannot compare” (p. 7).
Key results
-
Study 1 (US, n = 183 after 27 exclusions; pp. 2, 5).
- 78.14% (143/183) judged extinction bad on a binary item. On 1–7 scales: bad M = 5.61; needs to be prevented M = 6.01; a moral obligation to prevent it M = 5.69; funding it more important than education, health care and social security M = 5.43.
- Participants expected the future to be about as good as the present (M = 4.48). The better they thought it would be, the worse they judged extinction (r = 0.51). Optimists judged it worse (r = 0.32).
- Whether a happy person coming into existence makes the world better was near the midpoint (M = 4.45). Those who agreed found extinction worse (r = 0.22).
-
Study 2a (UK, Prolific, n = 1,251, pre-registered; pp. 2–4, 6). Five conditions; percentages are of those giving the expected ranking.
condition expected ranking extinction uniquely bad control 82.88% (213/257) 23.47% (50/213) animals (zebras) 89.84% (221/246) 44.34% (98/221) sterilization 82.54% (208/252) 46.63% (97/208) salience 77.82% (193/248) 50.25% (97/193) utopia 86.69% (215/248) 76.74% (165/215) - Animals: the same outcomes for the world’s zebra population.
- Sterilization: 80% or 100% of the world’s population “go sterile, meaning they cannot have children”, instead of dying.
- Salience: the control vignette plus a reminder that a humanity that survives “could go on to a long future”, and an extinct one has “no future” (p. 6).
- Utopia: survival leads to “a future which is better than today in every conceivable way”, with no wars, crimes or depression; the 80% catastrophe is followed by recovery; extinction means “all of human knowledge and culture will be lost forever” (p. 6).
- Every condition differed from control, and utopia from salience.
- Higher Cognitive Reflection Test scores predicted finding extinction uniquely bad.
-
Study 2b (US, n = 855, pre-registered; p. 4). The same five conditions.
- Control 18.75% (27/144); animals 34.59% (55/159); salience 39.45% (58/147); utopia 66.88% (105/157).
- Sterilization 28.75% (44/153), not significantly above control (χ²(1) = 3.55, P = 0.059).
- Utopia again exceeded salience; reflection again predicted the judgement.
-
Study 2c (Oxford students, N = 196, control and utopia only; pp. 4–5). Control 36.92% (24/65); utopia 76.83% (63/82). Only 65.7% ranked the outcomes as expected in control.
-
Study 3 (N = 71 recruited through the Effective Altruism Newsletter and existential-risk groups, who named existential risk their “most important cause”; p. 5). Control 85.71% (24/28); utopia 94.59% (35/37); no significant difference.
-
The authors’ reading (pp. 4–5). The animals and sterilization results show that sympathy for the victims drives the default answer. The salience result shows that the long term is neglected. The utopia result shows that beliefs about the future’s quality matter.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Laypeople judge human extinction bad and to be prevented | moderate: one US sample, Likert and binary items | Study 1 |
| C2 | Most laypeople do not judge extinction uniquely bad relative to an 80% catastrophe | strong for this task: three samples, two pre-registered, consistent | Studies 2a–2c |
| C3 | They answer so partly because they focus on immediate harm to the victims | moderate: two manipulations that weaken that focus raise the share, one of them (sterilization) not in the US sample | Studies 2a, 2b |
| C4 | They answer so partly because they neglect long-term consequences | moderate: a one-paragraph reminder doubles the share but leaves it at or below half | Studies 2a, 2b |
| C5 | Whether the future would be good matters to lay judgements of extinction | strong for this task: utopia beats salience in both general samples | Studies 1, 2a, 2b |
| C6 | Judging extinction uniquely bad may motivate devotion to preventing it | weak: a self-selected group differs from laypeople; direction of cause is open | Study 3 |
Concepts
- uniquely bad (extinction): the second difference (80% to 100%) judged greater than the first (none to 80%). It is a comparison of differences, not a rating of badness.
- existential risk: risks of human extinction, or of drastically curtailing humanity’s potential; the paper studies only the first (p. 5).
- immediate harm versus long-term consequences: the two grounds of extinction’s badness the design sets against each other.
Connections
- Parfit (LIT-469). The task is his, slightly changed. His prediction that most people would find the first difference greater is confirmed.
- Bostrom (LIT-1773). Cited for the claim that the future could hold “at least 10^16 human lives of normal duration” and be much better than the present (p. 2). The utopia condition puts something like his picture into the vignette.
- Caviola et al. (LIT-1749). By two of the same authors. Study 1 asks the same question their population studies ask, whether a new happy person makes the world better, and finds the midpoint. Their later studies find that people say yes when asked about one person.
- Scheffler’s afterlife conjecture (LIT-1754), unread. The sterilization condition is close to his infertility scenario: no one dies, and no one comes after. As the record states the conjecture, much of what we value depends on confidence that humanity will go on after we die. If that confidence weighed heavily in ordinary judgement, the step from 80% to 100% sterility, the step that ends humanity, should dominate. It did for fewer than half (47% UK, 29% US). That is an inference, and a weak one. The task asks how bad outcomes are, not whether one’s present projects would lose their point, which is what the conjecture is about. The vignette does not spell out that universal sterility means no one ever lives again.
- Not the legacy literature. Nothing here concerns the participant’s own descendants, reputation or continuity with the future (Wade-Benzoni, LIT-1716; Hershfield, LIT-1702).
Bearing on the record
- What the living owe the far future (QUESTION-299). The question cites this paper for “lay judgement does not treat extinction as uniquely bad unless the lost future is made salient”. That needs a correction. Salience alone brings it to about half (50% UK, 39% US). A majority judge it so only when told the future would be long and utopian.
- The owner’s thesis (CLAIM-491). These are
inferences; the paper measures no self, identification or felt
continuity.
- Nothing suggests concern for far-future people tracks an extended self. What raises it is information about the future’s length and quality. Concern that grows with how good a future is, for people no one will meet, looks like an impersonal valuation of outcomes.
- The default concern is for “fellow humans” now alive. The authors explain it by sympathy with victims. That fits the thesis’s larger self, the human group that exists now, better than it fits concern for a future.
- Species-membership works against the long view. More people judged extinction uniquely bad for zebras than for humans. The authors’ explanation is that people feel less for zebra victims. If so, a felt bond with the victims, which is what the thesis would count as extended self, crowds out concern for the lost future rather than carrying it.
- The triviality objection (CLAIM-492). Not addressed. The test the record wants, whether concern for the future follows a measured self–future continuity apart from caring, is not in this paper.
- A dated note should be added to CLAIM-491. Proposed text: “Schubert et al. (NOTE-1270) find that laypeople locate the badness of extinction in its victims, fellow humans now alive, and reach the lost future only when it is pointed out, and most when it is said to be good. Nothing measured there is a self, but what moves concern toward the far future is information about how good that future would be, which reads as an impersonal valuation rather than an extended self. The thesis’s hardest case in time is then not a hypothetical: where ordinary concern does reach the far future, the evidence so far shows it reaching there on the future’s merits.”
- No THEORY filed.
- No ML instruction.
Limitations
-
One task. Every main result is a forced choice between two differences, asked only of those ranking the outcomes as expected.
-
Study 3 is self-selected and its key extra questions are in the unread Supplementary Materials.
-
Statistics that do not match the counts. Recomputed from the reported counts (2×2 χ² with Yates’ correction), Study 2b’s values all match, but Study 2a’s do not (p. 4).
- Animals vs control is printed χ²(1) = 8.82, which is Study 2b’s value; the counts give about 20.1.
- Salience vs control is printed 29.90; the counts give 30.30.
- Utopia vs control is printed 30.30; the counts give about 119.
- Utopia vs salience, 29.90, matches.
Every difference is still significant.
-
Other discrepancies.
- The abstract’s total N = 2,507 (p. 1) does not equal the reported final samples (183 + 1,251 + 855 + 196 + 71 = 2,556).
- The Cognitive Reflection Test result is printed “Exp(B) = 0.15 … Odds ratio = 1.6” for Study 2a and “Exp(B) = 0.21 … Odds ratio = 1.2” for Study 2b (p. 4). Exp(B) is the odds ratio. The figures look like B, with 1.6 a slip for about 1.16.
- Study 3’s “14.00% female” of 71 is not a whole number of people (p. 5).
- “Published: xx xx xxxx” left on p. 1.
Open questions
- Does concern for the lost future follow the participant’s felt continuity with future people, or with humanity, when the future’s quality is held fixed? A continuity measure added to the salience condition would test the thesis’s reading against the impersonal one.
- Does the sterilization condition, spelled out as “no one will ever live again”, produce the weight on continuation that Scheffler’s conjecture predicts? Read Scheffler first (LIT-1754).