Strong longtermism, Pascal's mugging, and whether existing is good for us | Hilary Greaves (2020)

Definition of Strong Long-Termism

• Axiological Strong Long-Termism: The thesis that in a wide class of decision situations, the option that is best ex ante (has the best expected consequences) is contained within a small subset of options whose effects on the very long-run future are best.
• Deontic Strong Long-Termism: The thesis that the action one ought to do morally is among those options whose expected effects on the very long-run future are best.
• Distinction: Axiology concerns which actions are best (consequentialist focus), while deontology concerns which actions one is obliged to do (non-consequentialist focus); these do not always align outside of consequentialism.
• Scope: The definition specifically targets timescales of millennia or longer (e.g., human extinction or 100+ years), distinguishing it from standard political "long-termism" (e.g., thinking beyond the next election cycle).
• Function: It allows decision-makers to filter options by first identifying a "small subset" of long-term-best actions, then performing trade-off analysis within that subset rather than across the entire action space.

Robustness of Long-Termism

• Moral Theory Independence: The thesis is robust across various population axiologies and decision theories, not just Total Utilitarianism.
  • Prioritarianism: Long-termism remains plausible because long-term interventions often target "bad futures" (e.g., totalitarian AI, climate collapse) rather than improving an already wealthy future, making prioritarian weighting favorable to long-termist actions.
  • Non-Consequentialism: Even if one rejects consequentialism, the argument holds because non-consequentialist scruples (e.g., keeping promises) are unlikely to override the axiological stakes of long-term interventions, which are often orders of magnitude larger than short-term concerns.
• Mechanism of Argument: The expected value of influencing the long run is driven by the astronomical number of potential future people (estimated at roughly $10^{15}$ in expectation) and the possibility of "lock-in" events (e.g., extinction, value lock-in via AI) that persist for millennia.
• Alternatives to Extinction Risk: Long-termism is not synonymous with reducing extinction risk alone; it also encompasses improving the average well-being of future people or influencing the trajectory of value systems (e.g., in AI).

Moral Cluelessness

• Simple vs. Complex Cluelessness:
  • Simple Cluelessness: Arises from unstructured causal chains where consequences are so numerous and symmetric that they likely cancel out in expected value calculations (e.g., helping an old lady across the road).
  • Complex Cluelessness: Arises when there are structured reasons to believe specific long-term causal chains (e.g., funding bed nets leading to population growth vs. environmental degradation), but no way to assign robust probabilities to them; the outcome depends on arbitrary numerical choices.
• Objection Status: Cluelessness does not inherently favor short-termism over long-termism; it creates epistemic humility that reduces confidence in both long-term and short-term predictions.
• Practical Implication: While cluelessness makes precise cost-benefit analysis difficult, it does not justify ignoring long-term consequences; rather, it suggests avoiding "fanatical" interventions that rely on highly sensitive probability estimates without robust backing.

Moral Uncertainty

• Approaches to Decision-Making:
  • Maximize Expected Choice Worthiness: The dominant approach; applies expected utility theory to moral theories, weighting actions by the "moral value" according to the probability assigned to each theory.
  • My Favorite Theory: Choosing actions based solely on the theory with the highest individual credence.
  • Bargaining Theory: Modeling moral theories as disagreeing agents negotiating a compromise (e.g., Nash bargaining solution).
• Hilary Graves' Stance: She prefers the "Maximize Expected Choice Worthiness" approach.
  • Inter-theoretic Comparisons: She rejects the view that comparing values across different moral theories is impossible; she argues that while arbitrary, these comparisons are a necessary "extra component" the agent must settle on, similar to assigning probabilities in empirical uncertainty.
  • Bargaining Critique: When modeled using the standard Nash bargaining solution, the results converge closely with expected choice worthiness, failing to solve the "fanaticism" problem (where highly improbable but high-stakes theories dominate decisions).
• Fanaticism/Pascal's Mugging: The problem of allowing tiny probabilities of astronomical value to dominate decisions exists in both empirical and moral uncertainty; no well-defined decision theory has yet successfully avoided this without arbitrary constraints.

Population Ethics: Existence Comparativism

• Core Question: Can it be better or worse for a particular person to exist than not to exist?
  • Existence Comparativism: Yes; if a life is worth living, it is better for that person to exist than not. If a life is not worth living, it is worse for them not to exist.
  • Anti-Comparativism: No; it is incoherent to compare the well-being of a non-existent person to an existent one because there is no subject to bear the comparison in the non-existent world.
• Significance: Accepting existence comparativism supports "Totalism" (adding more good lives increases total value), making premature human extinction astronomically bad. Rejecting it leans toward "Averageism" or "Sufficientarianism," where extinction is less catastrophic.
• Critique of Anti-Comparativism: The standard "semantic" argument against comparativism (that relations require existing relata) proves too much; it would also prevent claiming that a world where a person exists is worse for them than a world where they exist but are less happy.
• The "Lives" Framework: A proposed solution that shifts focus from "persons" to "lives" (types of well-being). In this framework, one compares the "life" a person lives to a "non-life" (null life), allowing for comparison without requiring the non-existent person to exist as a relatum.

Global Priorities Institute (GPI) and Research Trends

• Current Status: The field is nascent; Oxford is currently the primary hub, but the pace of academic publication is slow (often >1 year per paper).
• Growth Areas:
  • Fanaticism in Decision Theory: Developing frameworks that avoid "Pascal's Mugging" scenarios in high-stakes decisions.
  • Infinite Ethics: Resolving paradoxes that arise when applying ethical theories to infinite populations or time horizons.
  • Long-Run Prediction: Improving forecasting methods for events 100+ years in the future to ground long-termist empirics.
• Economic Integration: GPI is actively hiring economists to integrate economic modeling with philosophical analysis, noting that economists have historically been underrepresented in this specific niche.
• Career Advice:
  • Grad School: Attend the best institution possible; while Oxford offers a unique community, the "best" school remains the primary factor for career capital.
  • Motivation: Proximity to like-minded peers can sustain interest in long-term, high-difficulty research topics over the duration of a PhD.
• Hiring: GPI plans to open postdoctoral positions in the fall (application deadlines around Christmas) and hosts early-career conferences annually in June.

Personal Context

• Work-Life Balance: Academic roles offer flexibility to manage family life (even with homeschooling during lockdowns, though the speaker noted the need for a caravan as a home office due to space constraints).
• Upcoming Events: EA Global X Asia Pacific is scheduled for November 20–21 as a virtual conference to connect the region with the broader effective altruism movement.