Longtermism and Effective Altruism: What Each Position Is Protecting

In November 2022, Sam Bankman-Fried's cryptocurrency exchange FTX collapsed, leaving an $8 billion hole where customer funds had been. Before the collapse, Bankman-Fried was the second-largest individual donor to the Democratic Party in the 2022 election cycle, a co-founder of the Centre for Effective Altruism, and one of the most visible figures in a philanthropic movement called effective altruism, which had become, in the years since philosopher Peter Singer's 1972 essay "Famine, Affluence, and Morality," a genuine intellectual and cultural force in how a certain segment of the wealthy and the highly educated thought about their obligations to other people.

Bankman-Fried had articulated, repeatedly and publicly, a philosophy called "earning to give": the idea that taking the highest-paying job available and donating a large fraction of the proceeds was more impactful than direct service work, because the money could be directed to interventions with proven, measurable impact at scale. He also articulated a version of expected-value reasoning under which normally prohibited actions could become permissible if the expected philanthropic impact were large enough. His lawyers argued at trial that his risk tolerance was calibrated to the long-run good of humanity. He was convicted on all counts of wire fraud and conspiracy in November 2023 and sentenced to 25 years in prison.

The collapse was not what caused the debate over effective altruism and longtermism. That debate had been building for years, inside the movement and outside it. But FTX made it legible to a wider audience: here was a case where the philosophical apparatus of a movement — expected value, earning to give, the tractability of present constraints by future impact — had been used to rationalize behavior that ordinary moral intuition would have stopped immediately. The question the collapse raised was not whether Bankman-Fried was an anomaly or an inevitable product. The answer to that question depends entirely on what you think effective altruism and longtermism are protecting.

What the strong longtermist position is protecting

The argument that the number of people who could exist in the future so vastly exceeds the number of people alive today that reducing even small probabilities of civilizational collapse is among the most important moral tasks available to us — and that most of what contemporary philanthropy and politics treats as urgent is, in relative terms, a rounding error. This position is associated most closely with philosophers William MacAskill (author of What We Owe the Future, 2022), Toby Ord (author of The Precipice: Existential Risk and the Future of Humanity, 2020), and Nick Bostrom (author of Superintelligence, 2014), as well as organizations including 80,000 Hours, the Future of Humanity Institute at Oxford, and the Machine Intelligence Research Institute.

The core argument is a math problem. If humanity survives and spreads through the galaxy over the next billion years, the number of people who will ever live could plausibly be in the range of 10²³ or more. If you assign even minimal moral weight to each of those potential people — if their lives would be worth living, if they matter morally at all — then a 0.001% reduction in the probability of extinction today has an expected value that dwarfs the entire present-day global health burden by many orders of magnitude. Ord estimates the probability of existential catastrophe in the next century at roughly one in six — "Russian roulette odds," he writes. MacAskill argues that this framing, if taken seriously, should reshape how we think about careers, institutions, and political priorities.

The practical implications are specific: AI safety (preventing artificial general intelligence from pursuing goals incompatible with human flourishing), pandemic preparedness and biosecurity (preventing engineered pathogens that could kill civilization), and nuclear risk reduction (lowering the probability of a civilization-ending exchange). These are the areas where longtermist funding has concentrated most heavily — through Open Philanthropy, through the FTX Foundation before its collapse, and through the direct career advice of 80,000 Hours, which for years ranked AI safety as the highest-leverage career path for people capable of doing it.

What this position is protecting: the moral seriousness of the very long run — the claim that our intuitive discounting of distant future people is not just psychologically explicable but philosophically indefensible. It is protecting the obligation to apply reason rigorously to the question of where moral concern should go, even when the answer is counterintuitive — even when it says that the child dying of malaria today matters less, in aggregate expected value terms, than reducing the probability that civilization collapses in 2150. The longtermist claim is not that present suffering doesn't matter. It is that our intuitive sense of scale is catastrophically miscalibrated, and that correcting that miscalibration is an act of moral seriousness, not callousness.

What the present-focused effective altruist position is protecting

The argument that effective giving requires rigorous measurement of actual impact — cost per life saved, dollars per quality-adjusted life year — and that the interventions with the strongest evidentiary base are overwhelmingly in global health and poverty reduction, not in speculative future-oriented programs whose impact cannot be measured. This position, closest to Peter Singer's original utilitarian case for effective giving in "Famine, Affluence, and Morality" and developed institutionally by GiveWell (founded in 2006), holds that the power of the EA framework lies precisely in its epistemic discipline: comparing interventions rigorously, demanding evidence, and directing resources to where they will do the most measurable good.

GiveWell's top charities — the Against Malaria Foundation, Helen Keller International's vitamin A supplementation program, the Malaria Consortium's seasonal malaria chemoprevention program — are recommended on the basis of randomized controlled trial evidence, independent replication, and cost-effectiveness modeling that has itself been subjected to repeated scrutiny and revision. The cost to prevent a death from malaria is in the range of $3,000–$5,000. That estimate is contestable in its details but grounded in a methodology that is transparent and replicable.

The present-focused position's critique of longtermism is epistemological before it is political. Expected value calculations for existential risk multiply very small probabilities by very large numbers. Those small probabilities — 1-in-6 odds of existential catastrophe, 10% chance of transformative AI by 2030, 0.001% annual probability of civilization-ending pandemic — are not derived from frequentist data. They are expert judgments about unprecedented scenarios. The mathematics of expected value is exact; the inputs are guesses. A calculation that looks like a rigorous optimization is, at its base, a product of intuitions dressed in quantitative clothing. If your probability estimate for AI risk is half as high as someone else's, the entire framework produces a radically different conclusion — and there is no agreed method for resolving that disagreement empirically.

What this position is protecting: the epistemic discipline that makes effective altruism effective — the demand that impact be accountable, measurable, and subject to revision based on evidence. It is protecting the claim that present-day suffering is real and legible in ways that hypothetical future suffering is not, and that directing resources away from preventable deaths toward speculative future scenarios requires a burden of proof that longtermism has not met. It is also protecting the reputational proposition that a movement calling itself effective should be able to demonstrate that its interventions work.

What the structural and political critique is protecting

The argument that both effective altruism and longtermism share a political silence — a refusal to ask why the world is structured so unequally in the first place — and that this silence is not a neutral omission but a substantive political position that forecloses more transformative alternatives. This position, articulated by philosophers including Amia Srinivasan (in "Stop the Robot Apocalypse," her 2015 essay in the London Review of Books), development economists skeptical of aid effectiveness, and advocates for systemic political change, argues that the EA/longtermist framework treats poverty and catastrophic risk as natural phenomena to be managed rather than as products of ongoing political arrangements that the movement's participants help create and maintain.

Malaria kills roughly 600,000 people per year. It is a preventable, treatable disease. The question EA asks is: given that it kills 600,000 people, how can we most cost-effectively reduce that number? The structural critique asks a prior question: why does it still kill 600,000 people when the interventions that could prevent it are well understood and affordable? The answer to that question involves trade policy that undercuts African agricultural development, intellectual property regimes that price essential medicines out of reach, debt structures imposed by the IMF and World Bank that constrain health spending in low-income countries, and colonial extraction that transferred wealth from the same places where malaria is endemic. EA's framework is not designed to address any of those causes. It is designed to work within them.

The structural critique extends to longtermism: the people most concerned about existential risk are overwhelmingly located in wealthy Western universities and tech-adjacent institutions. Their estimates of what constitutes existential risk — misaligned AI, engineered pandemics — reflect specific anxieties shaped by their location in the global economy. Climate change, which disproportionately threatens the global poor, receives a fraction of the longtermist attention that AI safety does, despite having stronger near-term probability estimates and significantly clearer causal mechanisms. The claim that longtermism is politically neutral — that it is just following the math — obscures the extent to which the choice of probability inputs, the selection of scenarios worth modeling, and the definition of what counts as civilizational catastrophe are all political acts.

What this position is protecting: the claim that charity and structural change are not the same thing and cannot substitute for each other; the proposition that the political context in which philanthropic money is generated matters morally, not just the impact at the margin; and the idea that movements that optimize for impact within the existing political economy are, by that choice, endorsing the existing political economy. The structural critique is not primarily an attack on malaria nets. It is an argument that malaria nets and structural change are different projects, and that directing philanthropic energy to the former while leaving the latter unaddressed is a choice that deserves to be owned as a choice, not presented as a neutral optimization.

What the cultural and epistemic critique is protecting

The argument that the EA and longtermist movements have developed institutional pathologies — a community culture that valorizes consequentialist overrides of ordinary moral intuition, concentrates philanthropic power in a small unaccountable network, and produces reasoning patterns that have historically enabled serious harm — that are structural features of the movement rather than incidental failures of specific individuals. This position is associated with critics including philosopher Émile Torres (who coined the acronym TESCREAL — Transhumanism, Extropianism, Singularitarianism, Cosmism, Rationalism, Effective Altruism, Longtermism — as a taxonomy of overlapping movements), journalists including Gideon Lewis-Kraus (whose 2021 New Yorker profile of the rationalist community traced the cultural genealogy), and former EA community members who have documented the movement's internal dynamics.

The cultural critique centers on what has been called "galaxy-brained" reasoning: the phenomenon where chains of seemingly valid logical steps lead to conclusions that should trigger moral alarm but instead get accepted because each individual step feels defensible. Bankman-Fried's reasoning about fraud was not irrational by the internal logic of expected value maximization — if you genuinely believed that the expected philanthropic impact of FTX's growth justified risk-taking with customer funds, the calculation was, in a narrow sense, coherent. The critique is not that consequentialism is always wrong; it is that any framework that allows sufficiently large expected future goods to override present deontological constraints is structurally vulnerable to producing conclusions that catastrophically fail the humans they claim to benefit. The evidence that the movement has this vulnerability is not only the FTX collapse; it includes documented cases of sexual harassment and abuse within EA organizations that were handled through in-group negotiation rather than external accountability, and a broader pattern of treating ordinary ethical norms as obstacles that sufficiently important work could override.

The concentration of power critique runs parallel. Open Philanthropy, the primary philanthropic vehicle associated with EA, directed roughly $750 million in grants in 2023, primarily toward AI safety, biosecurity, and global health. That money has reshaped entire academic fields — AI safety as a research area barely existed before EA funding made it possible to hire full-time researchers. The people making grant decisions are a small network connected by shared intellectual background, shared institutional affiliation, and often personal relationships. Longtermism worries explicitly about power concentration as an existential risk; it is, critics argue, less reflexive about the power concentration it is itself creating. MacAskill's What We Owe the Future explicitly warns against any group, including longtermist groups, seizing permanent control over the future. Whether the movement's current grant-making architecture is consistent with that warning is a question its critics take more seriously than many of its members.

What this position is protecting: the idea that the internal culture and power structure of a philanthropic movement is itself a moral object, not just a delivery mechanism for impact; the claim that universalist philosophical frameworks can mask deeply parochial assumptions about whose future matters and who gets to model it; and the concern that a community whose epistemics are optimized for internal consistency rather than external accountability will tend, under pressure, to rationalize its own interests as the good of humanity. The cultural critique is not an argument that the questions EA and longtermism ask are unimportant. It is an argument that the way a movement pursues good things is itself a moral question, and that getting the answer wrong at scale produces damage that the movement's own frameworks will be poorly designed to notice.