What's the effect of installing PlayPumps compared to funding hand pumps?

PlayPumps International

Good intentions aren’t enough

Imagine you’re in charge of a hospital…

Causes don't deserve support. People do.

How much should the government pay to save 200,000 birds?

How to avoid scope neglect

Which strategy do you think brings the greatest reduction in CO2e emissions per $100 of US government spending?

How many more lives does the most effective global health charity save per $100 spent than additional UK public health spending?

How much more can be accomplished per $100 implementing the most effective solution for Problem B than Problem A?

What is the best way for you to help?

We are in triage every second of every day

Why the problem you work on is the biggest driver of your impact

How to achieve hundreds of times more

Have more people died from causes directly related to terrorism or COVID-19 over the last 50 years globally? How much more?

Factor 1: Importance

How much does the US government spend on terrorism vs. pandemics?

Factor 2: Neglectedness

Do preventative-care home visits to elderly patients by healthcare professionals improve life expectancy?

Do drug substitution programs reduce criminal offending among drug addicts?

Does formally processing crimes committed by juveniles reduce criminal re-offending?

Factor 3: Tractability

The ITN framework

Which of the following problems is higher impact to work on according to the ITN framework?

The ITN framework, cost-effectiveness, and cause prioritisation

Introduction to EA – Why & How to Prioritize Problems to Work On

Global priorities research & shaping the ideas of intellectuals

We can use science to end poverty faster. But how much do governments listen to it?

Choosing the most critical problem

What happened to GDP per capita in England after 1825?

Our wealth across the ages

The industrial revolution changed everything

Comparing wealth across countries and time

We’ve seen progress across most areas of our lives

Billions now live free from extreme poverty

We’ve become vastly better educated

In what period did global life expectancy grow the most?

Our growing life expectancy

We could make the future radically better

Takeaways

The world is much better; The world is awful; The world can be much better

We radically improved the world, and could again

Know whether you’re actually helping

What impact did your colleague have from pushing the doctor out of the way to do CPR?

The real impact of saving a life

What was your true counterfactual impact?

Real positive change wouldn't have happened without you

Does Scared Straight help difficult kids commit less crime?

Scared straight?

You can accomplish more in hard-to-replace jobs

'Replaceability' isn't as important as you might think (or we've suggested)

Shapley values: Better than counterfactuals

How to be irreplaceable

Who deserves our concern?

Should you report the kidnapping to the police?

Kidnapper conundrum

We should care for moral patients

A golden rule

The all-knowing dictator

The blind dictator

We’re making moral progress

Same-sex relationships are now legal in the majority of countries

Approval of marriage between Black people and White people flipped from 4% to 94%

Many more women now enter the workforce than a century ago

We’re not done yet

Being a moral entrepreneur

Our descendants will probably see us as moral monsters

Identifying and tackling prejudice

Every answer has an element of uncertainty

Expected value is a tool for handling uncertainty

Assuming you could afford to lose, should you take this bet?

Let’s make a bet

Averting a nuclear disaster

What is the expected value, in additional years of healthy life, of a dose of Paxlovid?

Paxlovid

Expected value is used by leaders across industries

Which of your options has the highest expected value?

Sink or swim

Probabilities can be subjective and made up

If we accept GiveWell’s analysis, what increase should we expect for receiving deworming treatment?

Deworm The World

Expected value doesn’t handle extremes gracefully

The Fox And The Hedgehog: A Story of Triumphs and Tragedy

Moral cluelessness, population ethics, and probability within a multiverse

How to manage uncertainty in impact

Which problem should you work on?

Your career is loooooong…

You could change the course of history

There’s never been a better time to change the world

Which option would have a bigger direct impact on CO2 emissions?

Which of these factors has the strongest relationship with happiness on average?

Which graph best describes the relationship between income and happiness?

You can be financially secure and impactful

Take action

How to Pick a Career (That Actually Fits You)

80,000 Hours: Find a fulfilling career that does good

Making the most of your career

At what age is the average American’s salary the highest?

Build career capital to set yourself up for success

How often do we change our minds about our values, relative to what we expect?

Conduct quick tests to find your best long-term option

To get the best score you can, how many hours should you research before you start writing?

How much should you research?

What percent of technology startups fail to return a positive investment?

Reach high, with a backup

Score your options

Career exploration: when should you settle?

What to do when you’re uncertain

How many citations do the top 20% of scientists account for?

1_How many citations do the top 20% of scientists account for

You won’t know if something’s a good fit for you until you’ve tried it

What percent of American undergraduate students change degrees?

Take small steps to become more confident in a degree

How many passionate students are there in sports, music, or art for every job?

Don't blindly follow your passion

How often, on average, does a member of Gen Z change jobs?

Try out more options than you think you need to

Which undergraduate degree has the highest median salary?

Choose a flexible degree to not lock yourself in

High school is 12 years long. What is the value, in increased earnings, of finishing that last year of high school compared to finishing only 11 years?

Prestigious degrees open doors

Infrequently asked questions

Many Undergrads Should Take Light Courseloads

Half-assing it with everything you've got

What to study at university

Compare your long-term paths

Create a plan that makes sense for you

Your most impactful, fulfilling career

Common misconceptions

How many lives a doctor saves

What's more important to check to find out if this charity would actually help people?

Save The World By Helping Things™

The overhead myth

Compared to a doctor, how many more lives per year does Elijah, the engineer, save?

What does it cost in the field of global health to save a life, to the nearest order of magnitude?

How does the median American income compare to the rest of the world?

How rich will you be?

You can save lives without a medical degree

Join or build a community changing the world

Confidently taking your next steps

How you can help now

How to actually change the world

Will you catch COVID if you go to the movies? Will you be more successful as an academic or a software engineer? Does taking a painkiller make you feel better?

It’s not just rolling the dice in Monopoly. Almost every decision you make involves probability, whether you explicitly calculate it or not.

Nerdy aside: Why you shouldn't assign probabilities of 0% or 100% to almost any decision

1_Every answer has an element of uncertainty.png

Your decision isn’t just dependent on the likelihood of an event. Getting a COVID vaccine isn’t valuable because you’re highly likely to end up in hospital if you caught COVID, but because it’d be so bad if you did. You should wear a seatbelt despite the fact that the odds of a crash are low for any given trip.

So we need to combine the probability of an outcome, and how good or bad it’d be if it happened. Doing this is known as expected value reasoning.

When considering just one case at a time, the formula for expected value is very simple:

We flip a fair coin. If it’s heads, I give you $15. If it’s tails, you give me $10.

The expected value, or “EV” of an action is the average value, if you were to take that action many times over.

If you don’t take the bet, you keep your $10.

If you do take the bet, you have a 50% chance of winning $15, and a 50% chance of losing $10.

Because there’s an equal chance of heads or tails, the 50% chance of winning $15 is more valuable than the 50% chance of losing $10 is disvaluable.

50% chance of gaining $15 + 50% chance of losing $10 = 0.5 × $15 + 0.5 × (-$10) = $2.50.

If you really did take this bet every day for a year, there’s less than a 1 in 100 million chance that you’d end up having lost more money than you won.

Nerdy aside: how to know when to fold ‘em

Taking this bet repeatedly would make you very rich.

Even though the chances of a nuclear power plant melting down are very low, safety measures to reduce these chances are worthwhile because of the consequences of such a catastrophic meltdown.

The value of nuclear safety measures depends on both:

How bad a disaster would be if one did happen.

Neither factor in isolation is enough to make a good decision. To increase the expected value of some safety measure, we could (1) increase the chances it prevents a disaster or (2) decrease how bad a disaster would be given that one does occur.

Imagine you're working at the Food and Drug Administration evaluating whether to approve a new drug for the treatment of COVID, Paxlovid.

Paxlovid works to prevent hospitalisation and death from COVID in 90% of high-risk patients.

 If it doesn't work, it has no impact, and doesn’t increase the chances of complications.

Let’s assume that when it works, it gives patients an average of 1 extra year of healthy life.

There's a 90% chance of success, which would lead to one more year of healthy living for a patient.

Conversely, there's a 10% chance of failure (0 extra years of healthy living).

90% chance of gaining 1 year + 10% chance of losing 0 years = (0.9 × 1) + (0.1 × 0) = 0.9 years

Taking Paxlovid would lead to 0.9 extra healthy years of life.

When something is important enough, you do it even if the odds are not in your favor. 

— Elon Musk, entrepreneur, founder of Tesla and SpaceX

Thinking about expected value also applies when the downside is terrible. For example, even if the probability of your having cancer is low, it might pay to get yourself tested when you have a symptom just to make sure.

— Ray Dalio, investor, founder of the world’s largest hedge fund, Bridgewater

Take the probability of loss times the amount of possible loss from the probability of gain times the amount of possible gain. That is what we’re trying to do. It’s imperfect but that’s what it’s all about.

— Warren Buffett, investor, CEO of Berkshire Hathaway

You're a lifeguard at a busy beach when you notice a rare tide has resulted in 1,000 people being drawn out to sea.

You only have time for one of two options:

Run in and save 1 person chosen randomly, OR

Call the Coast Guard, who have a 1% chance of saving everyone if they happen to already be in the area.

Run in myself: 100% chance to save 1 life

Call the Coast Guard: 1% chance to save 1000 lives

You could save 1 person with certainty, or have a 1% chance of saving everyone.

Imagine you’re one of the 1,000 people drowning. If the lifeguard chooses to call the lifeguard, you have a 10 times higher chance of surviving than if they try to run in.

You’re unlikely to survive in either option, so it’s not a great situation to be in. But if you had to choose, one option seems much better for you than the other.

Taking the risk of calling the Coast Guard would result in 

What’s the probability of rolling a 5 on a fair 6-sided dice? Easy, 1 in 6.

But what would you say the probability is, if you’ve already seen the result and know that it’s a 5? Nearly 100%.

Probability can be merely a number you assign, even if you’re not confident. Even rough guesses can be useful.

For example, let’s say you’re trying to decide whether to take a new treatment for depression, Superion.

 A study funded by a pharmaceutical company demonstrates that Superion cured 99% of depression cases. You’re sceptical of the study’s accuracy and integrity given its funding source, and your subjective estimate is probably closer to 20%. You’d be better off relying on your 20% estimate, rather than the study claiming a 99% success rate.

You’re working at the United States Agency for International Development (USAID), in charge of recommending what health programs get funded with their limited budget.

You meet the founder of a (real) charity, Deworm The World, who tells you:

Hundreds of millions of kids have parasitic worms in their guts.

 Treatment is safe, cheap, and effective at killing the worms.

 The long-term benefits are potentially major but hotly debated.

One experiment finds that deworming children increases their expected lifetime earnings by $1,258 compared to the placebo group.

GiveWell, a charity evaluator that has studied deworming in detail, is concerned these results may not apply to today’s environment.

GiveWell’s best guess is that the true effect of deworming on lifetime earnings is only 10% the size reported in the above experiment.

The expected value formula works for both objective probabilities and subjective assessments.

So in this case, if we use GiveWell’s subjective assessment of the probability that deworming has this expected impact, we get:

How to manage uncertainty in impact

What you'll be able to do