Trying to do good, and knowing whether it worked.

Faith traditions

Giving as duty

Several of the world's great faiths made giving an obligation rather than a mood. In Islam, zakat is an obligatory annual alms and one of the five pillars of the faith.1 In Jewish law, Maimonides ranked eight levels of charity, from giving grudgingly at the bottom to, at the very top, helping someone become self-sufficient through a gift, a loan, a partnership or a job.2

That top rung is worth pausing on. Long before anyone ran a trial, the highest form of giving was defined by its result: a person who no longer needs charity.

What it learned about knowingGiving was judged mainly by the giver's manner and intent. But the best of the old thinking already pointed at outcomes.

1601 onward

The parish and the almshouse

In England, the 1601 Act for the relief of the poor made each parish responsible for its own poor, paid for by a local rate. It set the able-bodied to work, apprenticed children, and allowed parishes to build dwellings for those who could not work.3

Charity was local and personal. The giver often knew the recipient by name, which made it easy to see what happened next, and easy to judge people rather than results.

What it learned about knowingProximity gave donors feedback, but the feedback was about people's character, not about whether help worked.

1869 onward

Victorian “scientific charity”

The Charity Organisation Society, founded in London in 1869 and copied in Buffalo, New York, in 1877, set out to coordinate relief, investigate every applicant's “worthiness” and register cases so that help was not duplicated.4

It was the first serious attempt to bring method to giving, and it pointed its method in the wrong direction. It spent its energy on whether the poor deserved help, not on whether the help made them better off.

What it learned about knowingMeasurement arrived, aimed at the recipient's worthiness. The lesson: investigate the programme, not the person.

1889–1930s

Carnegie, Rockefeller and the great foundations

In 1889 Andrew Carnegie argued that the rich were trustees of their wealth and wrote that "the man who dies thus rich dies disgraced".5 The Rockefeller Foundation was chartered in 1913 "to promote the well-being of mankind throughout the world".6

Before it, the Rockefeller Sanitary Commission surveyed hookworm across the South, found about 40 percent of school-aged children infected, and treated it; economists later showed schooling and earnings rose where the disease had been worst.7 Julius Rosenwald's programme built nearly 5,000 rural schools for Black children, with gains in literacy and schooling that researchers could still measure decades later.8

What it learned about knowingBig money began aiming at root causes and building institutions. The best of it surveyed before acting, which is what made the results knowable.

1938–1955

Dimes, and a proper test

The National Foundation for Infantile Paralysis, later the March of Dimes, turned millions of small gifts into patient care and research.9 It backed field studies of the Salk vaccine involving about 1.8 million children, and on 12 April 1955 the vaccine was announced as "safe, effective, and potent".10

What it learned about knowingOrdinary donors funded a large, public test before the public was asked to trust the result. That is still the gold standard.

1943–1980

Science at scale: wheat, smallpox, salt and sugar

A Rockefeller Foundation programme in Mexico, begun in 1943, bred the wheat varieties that helped launch the Green Revolution, and Norman Borlaug won the 1970 Nobel Peace Prize for it.1112 Researchers estimate that high-yielding varieties raised food-crop yields by 44 percent between 1965 and 2010.13

In 1967 WHO launched its intensified campaign against smallpox; in 1980 the disease was declared eradicated.1415 In 1978 The Lancet called the science behind oral rehydration "potentially the most important medical advance this century", and BRAC went on to teach it to mothers in about 12 million Bangladeshi households.1617

What it learned about knowingWhen the outcome can be counted directly, such as cases, harvests or children rehydrated, steer by it. The biggest wins in history were the most measurable.

1999–2003

The global health machine

A $750 million pledge from the Gates Foundation in 1999 helped launch Gavi, the vaccine alliance.18 The Global Fund followed in 2002 and the US government's PEPFAR in 2003.1920 Each now publishes large estimates of lives saved: 70 million for the Global Fund since 2002, by its own count, and 25 million for PEPFAR, by the State Department's.1920

What it learned about knowingPooled funds brought scale and annual results reports. Their headline numbers are modeled and the credit is shared, so read them as careful estimates, not counts.

2003–2019

The randomistas

In 2003 Abhijit Banerjee, Esther Duflo and Sendhil Mullainathan founded J-PAL at MIT to test anti-poverty programmes with randomised trials.21 In 2019 Banerjee, Duflo and Michael Kremer shared the Nobel prize in economics "for their experimental approach to alleviating global poverty".22

Trials revealed what good intentions could not. Cash transfers did measurable good.23 Free laptops did not raise test scores.24 And taking young people on prison visits to scare them straight made offending more likely, not less.25

What it learned about knowingThe question became counterfactual: compared with what would have happened anyway? That one question retired more bad ideas than any scandal.

1972–today

Evaluators and effective altruism

In 1972 Peter Singer argued that if we can prevent something bad without sacrificing anything of comparable moral importance, we ought to.26 In 2007 two people who had worked in finance founded GiveWell to research where donations do the most good.27 Animal Charity Evaluators followed for animal welfare, and the Happier Lives Institute, founded in 2019, began comparing charities by their effect on wellbeing.2829 In 2025 Open Philanthropy, a major funder in this tradition, renamed itself Coefficient Giving as it began working with more donors.30

The movement's great contribution was cost per outcome and room for more funding: not just whether a programme works, but how much good the next dollar does. Its great embarrassment was the FTX collapse in 2022, which showed that judging where money goes is not enough if you never ask where it comes from.31

What it learned about knowingCompare cost per outcome, ask what the next dollar does, and stay humble about your own models.

Recent decades

The modern giving machine

Donor-advised funds made giving easy to start and easy to postpone. Almost 2 million people now give through them.32

The plumbing of giving has never been better. The plumbing of knowing has barely changed. A donor can move money in seconds and still have no idea, a year later, whether it did anything.

What it learned about knowingConvenience is not impact. Money parked for good is not yet doing good.

Now Our view

AI agents at the table

When someone asks an AI assistant where to give, the assistant reads what the evaluators publish, and what we publish, and then answers with great confidence. It can make good evidence available to everyone, or launder a guess into a fact at enormous scale.

That is why everything on this site carries its source and its date, labels estimates as modeled, not guaranteed, and keeps a dated record we check again later. We built it for people first and machines second, and we hold both to the same rule.

What it learned about knowingThe next era will be judged on whether machines quote evidence honestly. We intend to make that easy.