Effective Altruism — The Mathematics of Doing Good

A $5,000 donation can be treated as a moral fork: one path buys a used car, the other may avert a child's death from malaria. Effective altruism is the attempt to make that fork explicit, then ask why ethics should get less numerical discipline than investing, medicine, or engineering. Its power is the same thing that makes it dangerous: once lives become expected values, very small probabilities attached to very large futures can dominate the present.

The case

Effective altruism starts with Peter Singer's 1972 essay "Famine, Affluence, and Morality." Singer's claim was not gentle: if you can prevent serious suffering without sacrificing something comparably serious, you ought to do it. Distance does not matter much. A drowning child in front of you and a child dying from malaria 8,000 kilometers away both count.

The movement turned that moral pressure into a method. Causes are compared by scale, neglectedness, and tractability. A problem affecting 200 million people, receiving little funding, and solvable with cheap interventions beats a local problem with more attention and weaker evidence. GiveWell, founded in 2007 by Holden Karnofsky and Elie Hassenfeld, made this style concrete by ranking charities using public cost-effectiveness models.

The near-term case is strongest in global health. Anti-malaria nets, seasonal malaria chemoprevention, vitamin A supplementation, and direct cash transfers can be compared using QALYs and DALYs: quality-adjusted or disability-adjusted life years. The numbers are uncomfortable because they force tradeoffs. If one intervention buys 100 healthy life-years per dollar and another buys 2, sentiment has to explain why it deserves the money.

Where the calculation changes

Longtermism changed effective altruism from charity evaluation into civilizational triage. Toby Ord's The Precipice (2020) estimates roughly a 1-in-6 chance of existential catastrophe this century, with unaligned artificial intelligence and engineered pandemics doing much of the work. William MacAskill's What We Owe the Future (2022) argues that future people matter morally, even if they do not yet exist.

That move produces the movement's sharpest result. If humanity might last millions of years, then reducing extinction risk by 0.001% can appear to beat saving thousands of lives today. This is why EA money moved into concept ai alignment, pandemic preparedness, forecasting, biosecurity, and institutional decision-making.

The FTX collapse made the cost visible. Sam Bankman-Fried was the most famous "earning to give" case: get rich, then donate to high-impact causes. FTX failed in November 2022; Bankman-Fried was convicted of fraud in 2023. The scandal did not refute bed nets or QALYs. It did expose a failure mode: a person convinced of enormous future value can rationalize present harm with a spreadsheet.

What's contested

The central dispute is not whether evidence matters. It is whether expected value can carry ethics under deep uncertainty. A 0.000001% chance of saving a trillion future lives looks huge on paper, but tiny probabilities about unprecedented events are not like casino odds.

There is also a politics critique. EA often treats suffering as a funding allocation problem, while critics argue that poverty, factory farming, and pandemic risk are also power problems. A donor can buy malaria nets; changing trade rules, land use, or state capacity is harder to model.

The movement's own split reflects this. Near-term EA leans on measurable interventions. Longtermist EA leans on speculative stakes. Both use the same mathematical grammar, but the evidence is not equally solid.

Why this has to do with other realms

Effective altruism is philosophy behaving like operations research. That makes it a bridge between concept utilitarianism, concept expected value, and concept ai alignment. The same habit that ranks charities also ranks futures.

It also touches biology in a blunt way. If factory farming affects more than 80 billion land animals per year, then moral concern for nonhuman experience can outrank many human-centered causes by scale alone. That question runs straight into concept animal consciousness and concept hard problem consciousness: the spreadsheet needs a number for suffering, but consciousness has not supplied one.

An open question

If the highest expected-value action depends on probabilities nobody can verify, when does moral mathematics become a machine for laundering guesswork?

Key Sources

Peter Singer, "Famine, Affluence, and Morality" (1972) — the argument that made distance morally suspect.
William MacAskill, Doing Good Better (2015) — the clearest early public case for EA methods.
Toby Ord, The Precipice (2020) — the canonical longtermist risk argument.
William MacAskill, What We Owe the Future (2022) — the longtermist case in book form.
GiveWell cost-effectiveness analyses — public models for malaria, supplementation, and cash transfer interventions.
to verify: current Open Philanthropy grant totals for AI safety, biosecurity, and global health.

Abhishek