NOTE-1293: Responsibility to Future Generations: A Moral Reframing Strategy for Combatting Climate Change

Contribution

It proposes responsibility to future generations (RFG) as a “moral reframing” of climate change. The claim is that RFG is widely held, little tied to demographics or politics, and normally not salient, so appealing to it should move people whom appeals to climate responsibility leave cold. Thirteen studies test the three parts, and a large preregistered experiment compares five ways of raising RFG.

Key insight

Responsibility toward future people is a less partisan version of responsibility for climate. The two correlate strongly (r = .58 to .76) but behave differently: Republicans and Independents endorse RFG more than climate responsibility (RCC), Democrats about equally, and RFG’s link to conservatism is half as strong. One manipulation moved RFG without moving RCC, which the authors take to show “two separate mechanisms, despite their high correlation” (p. 32).

Assumptions

• RFG is a value people already hold. The rationale is that “caring for future generations is a widely endorsed principle” (p. 4). The paper uses caring, duty and responsibility interchangeably.
• Kinship may carry it. “Future generations typically include a person’s descendants”, and family receives the most moral consideration (p. 6, citing Crimston et al. 2016). Not tested.
• Self-report stands for behaviour. Behaviour frequency, intentions and policy support are self-reported; the one behavioural measure is a lottery-bonus donation.
• Single items carry RFG in Studies 1, 2A and 2B.

Key results

• Study 1 (pp. 8–12). Four Eurobarometers, 2009–2011, 34 countries, N = 109,513. Item: “We need to reform to benefit future generations even if that means making some sacrifices now” (1–4).
  • Majority agreement in 33 of 34 countries; Latvia 48%, Lithuania below 60%, all others above 70% (p. 10).
  • Meta-correlations with RFG: ideology r = .03, subjective status .13, age .07, gender (female) −.05. Only status passed the preregistered r = .10 threshold (p. 11).
  • Supplementary: binary RFG measures in 1988 and 2008 surveys ranged from 25% to 79%; 74.1% of surveyed Ecological Society members called RFG a relevant value for a scientist (p. 12).
• Studies 2A–2B (pp. 12–19; CloudResearch, N = 457 and 906). Matched items: “When deciding how to live, I have a duty to consider the impact of my actions on future generations” against ”… on climate change” (p. 13).
  • RFG above RCC: 2A, M = 5.48 against 5.17, d = 0.26, r = .67; 2B, 66.51 against 60.08 (0–100), d = 0.26, r = .58 (p. 14).
  • The gap was driven by Republicans (2B: d = 0.66) and Independents (d = 0.34). Democrats scored highest on both and slightly higher on RCC (d = −0.12) (pp. 15–16).
  • People thought other Americans held RFG more than RCC, and thought others found RFG more morally right (2B: d = 0.54), in every party (p. 17).
  • Across Studies 2A–3B (N = 3,558), RFG correlated with conservatism r = −.21 and RCC r = −.45 (pp. 18–19, Table 3). RFG barely correlated with age, income, education, gender, race or religiosity (all |r| ≤ .07 meta-analytically).
• Study 3A (pp. 20–22; Prolific, N = 395). Five-item scales, RFG α = .93, RCC α = .96, correlated r = .76.
  • RFG predicted self-reported proenvironmental behaviour (β = .48; .41 with covariates), concern (.52; .37), perceived harm (.50; .36), and expecting harm sooner (−.27; −.14). Donations to charity only with covariates (β = .12) (Table 4, p. 22).
  • The two responsibilities were never entered in one model, “due to the very high correlation” (p. 21).
  • Supplementary Detroit Area Study 2002 (N = 328): RFG related to 15 of 16 environmental outcomes, βs ≥ 0.20 (p. 23).
• Study 3B (pp. 23–28; Prolific, N = 1,800, 600 per party). RFG exceeded RCC however climate was framed (global warming, greenhouse gas, carbon emissions, climate change), r = .67 to .75. RFG predicted five Environmental Attitudes Inventory subscales, β = .47 to .54, adjusted R² .22 to .29 (Table 6, p. 27).
  • Supplementary Study S5: RFG predicted less endorsement of protecting the environment for anthropocentric reasons, β = −.35 (p. 28).
• Study 4 (pp. 28–38; preregistered, Prolific, N = 3,175, six conditions). Effects against a control (writing about one’s daily routine), Table 7 (p. 34):
  • On RFG (control M = 5.17 on 1–7): letter b = .37 (d = .29); sacrifice .29 (.22); legacy .27 (.21); video .19 (.15); longtermism .18 (.14). All significant.
  • On RCC: all but sacrifice (b = .14, p = .118). The sacrifice task did not mention climate (p. 32).
  • Intentions: all five; letter b = .54 (d = .54), others d = .18 to .24.
  • Policy support: letter (d = .18) and longtermism (d = .14) only; legacy p = .073.
  • Donations (of a $10lotterybonus;controlM=$3.31): letter only, b = $0.49 (d = .15).
  • The conditions: a letter to a person born in 2022 who will be 28 in 2050 (after Shrum 2021); reflection on one’s legacy (Zaval et al. 2015); reflection on sacrifices of the past generation (Watkins and Goodwin 2020); a 1-minute Grand Canyon video; a 385-word excerpt from the introduction of What We Owe the Future, section “Future People Count” (pp. 30–31).
  • Mediation (exploratory, preregistered; pp. 36–37). Every condition had a significant indirect effect through RFG on intentions, policy support and donations. RCC was not entered as a parallel mediator because of collinearity (r = .75); entered after RFG, RFG still predicted two of three outcomes (n. 5, p. 36).

Claims

id	claim	strength	support
C1	Most people in Europe and the US endorse a responsibility to future generations	strong for endorsement of the items; single items in Study 1, unweighted pooled surveys	Study 1, S1–S2, Studies 2A–3B
C2	RFG is nearly unrelated to demographics and is half as tied to conservatism as RCC	strong for the correlations, in WEIRD samples	Studies 1–3B, Table 3
C3	RFG and RCC are distinct despite correlating .58 to .76	moderate: one manipulation moved RFG and not RCC; RCC’s own effect was positive and non-significant	Study 4 sacrifice condition
C4	RFG predicts self-reported proenvironmentalism, with demographics controlled	moderate: consistent across samples, but RFG and RCC never compete in one regression	Studies 3A, 3B, S4, S5
C5	Brief interventions raise RFG and proenvironmental intentions	moderate: preregistered, N = 3,175, small effects, outcomes measured minutes later	Study 4
C6	Raising RFG is how the interventions work	weak: exploratory mediation with RFG measured after the manipulation, alongside the outcomes	Study 4 path model

Concepts

• responsibility to future generations (RFG): agreeing that one has a duty to consider, or to make sacrifices for, the effects of one’s actions on future generations.
• responsibility to reduce climate change (RCC): the same duty toward one’s personal contribution to climate change; in norm activation theory a core antecedent of proenvironmental action.
• moral reframing: recasting an issue in terms of values an audience already holds, to reduce polarisation (Feinberg and Willer).
• legacy motivation (as the paper cites it): concern “for being remembered by future generations as a good person” (p. 5).

Connections

• Syropoulos and Markowitz 2021 (LIT-1741), unread. Cited only as “preliminary work” finding concern for future generations “weakly or not at all related to political identity or partisanship” (p. 3).
• Legacy and the climate game (Hurlstone et al., NOTE-1279). Hurlstone’s messages combined death awareness, future generations’ powerlessness and paying forward a past generation’s sacrifice. Study 4 separates legacy from sacrifice and finds both raise RFG about equally (b = .27 and .29).
• Zaval et al. (LIT-1724), unread. Study 4 replicates its legacy effect on intentions but not on policy support or donations (p. 33).
• What We Owe the Future (LIT-1747) and the SPC framework (filed with this reading). The book’s excerpt was one intervention. The book’s notes (1.5) name the reciprocity ground the sacrifice condition tests.
• Lay valuation of the far future (Schubert et al., NOTE-1270; Caviola et al., NOTE-1267). Those studies found far-future concern appears when prompted. Here a prompt as short as a 1-minute video raised RFG, but “future generations” is never dated beyond 2050.

Bearing on the record

• The triviality objection (CLAIM-492). The paper does not supply a criterion of the self’s extent that comes apart from caring. These are inferences from what it measures.
  • No measure of self-extension appears: no overlap with future people, no identification with descendants or with humanity, no future-self continuity.
  • Responsibility is not separated from caring. The authors justify RFG as “caring for future generations” (p. 4) and ask whether it is “morally right to care about” the future (p. 13). The 3A harm scale averages harm to oneself and one’s family with harm to future generations and species (pp. 20–21), so self-regarding and other-regarding concern were pooled.
  • What it does separate is two responsibilities, RFG and RCC. That is not the separation the objection asks for.
  • Its kinship hypothesis, that “future generations” means one’s descendants (p. 6), is the thesis’s own reading, and the paper’s limitations grant it cannot say whether people think of descendants or “future people more broadly” (p. 43).
• The owner’s thesis (CLAIM-491). Study 4 compares routes to the same felt responsibility. Read as inferences:
  • The weakest route was the impersonal one. The longtermist excerpt raised RFG least (d = .14) and did not raise donations. The strongest was a letter to one particular person born today (d = .29; intentions d = .54).
  • A self-regarding route worked as well as most. Legacy reflection, concern for how one will be remembered, raised RFG (d = .21).
  • A route that is neither self nor impersonal good worked too. Reflecting on what past generations gave up raised RFG (d = .22) and not RCC. That is reciprocity, an obligation from benefit received, and it is not concern for an extended self.
  • So the experiment fits the thesis’s prediction that concern is stronger when the future is made particular and personal, but it also gives a non-self route of similar size. The effects are small, measured minutes later, and on self-report.
• QUESTION-299. A dated note is proposed (section 4). It updates the question’s “What predicts it” bullet, which cites the 2021 paper.
• No THEORY filed.
• No ML instruction.

Limitations

• Self-report. The only behaviour is a share of a lottery bonus, and only the letter moved it (p. 42).
• WEIRD samples only, by the authors’ account (p. 42).
• “Future generations” is undefined. No study says how far ahead or whose; Study 4’s letter targets 2050.
• No test of RFG against RCC in one regression; mediation measured RFG after the manipulation and with the outcomes.
• Study 1 merged four waves and so dropped the survey weights (p. 9). Its item asks about reform and sacrifice, not duty.
• Overstatement.
  • The discussion says “letter, legacy and longtermism interventions successfully increased support for policy” (p. 38); legacy’s effect was p = .073, which the results call “marginal” (p. 33).
  • The general discussion says the results “replicated work on the three existing interventions” (p. 41); the results report failures to replicate Zaval et al. on policy support and donations (p. 33).
• Inconsistencies and typos.
  • Study S2’s N is 1,215 in Table 1 (p. 8) and 1,285 in the text (p. 12).
  • The trimmed path model reports χ²(7) = 0.69, p = .686 (p. 36); with 7 degrees of freedom, 0.69 would give p near 1.
  • The text refers to “Table 11” for the indirect effects, which are Table 8 (pp. 36–37). There, legacy → donation has a 95% interval printed as 0.29 to 0.14, which excludes its own b = 0.18 (p. 37).
  • Policy-support items are described as capturing “proenvironmental intentions for the next three months” (p. 31).
  • “Watkins and Goodwin (2021)” for 2020 (p. 32); “Xiao & McCright, 2013” in the text, 2015 in the references; “MacAskill, 2022” in the text, 2023 in the references; “Stern, 200” (p. 7).
  • “RFT and covariates” (Table 4, p. 22); “batterυ” (p. 23); “t(1543 =” (Table 5, p. 25); “Averagte” (Figures 6a–6c, p. 35); “Participants were randomly to one of six conditions” (p. 29); “responsibility to future generations someone” (p. 29); “Fischer’s Z” for Fisher’s (pp. 18–19).
  • Table 4 prints standardized βs beside confidence intervals that are not on the β scale (p. 22).

Open questions

• Does felt responsibility to future generations vary with a measured extension of the self, apart from caring? A study adding overlap with one’s descendants, or identification with all humanity, and a separate caring measure, would test CLAIM-492 across generations. A later preprint by Syropoulos, Law and Young, “Caring for Present and Future Generations Alike” (PsyArXiv, 2023, 10.31234/osf.io/hzwrt), not held, reports by its abstract that identification with all of humanity may mediate longtermists’ moral regard.
• Does reciprocity toward the past move costly action toward the future, and does it work when the future people are strangers?