Notes from the resident sceptic: what effective altruism gets right, and where I get off

I joined my university's effective altruism group in first year because a friend promised free pizza and an argument. Four years later I'm still there, and I've become what the others call, with varying degrees of affection, the resident sceptic. This is my attempt to write down where I've landed: what I think the movement gets right, and why I can't follow it all the way.

The part I can't argue with

Effective altruism starts from Peter Singer's 1972 essay "Famine, Affluence, and Morality", the one with the drowning child. If you can prevent something very bad without sacrificing anything of comparable moral importance, you ought to do it, and distance makes no difference to that. Singer's The Life You Can Save turns the argument into a book for general readers; his Practical Ethics sets it inside a wider theory.

The movement added something Singer's essay didn't stress: if you are going to help, it matters enormously how. Some charities do far more good per pound than others. GiveWell, founded in 2007, tries to compare them using evidence, and Giving What We Can, which Toby Ord founded with William MacAskill, asks members to pledge at least ten per cent of their income to the most effective causes they can find.

I've watched people in my group take that pledge as students living on noodles and keep it after they started earning. Whatever else I say, I want this to be clear: that is admirable, and the core claim is right. If you are going to give, you should care whether it works. The objections that call this cold or arrogant have never convinced me. Wasting a gift that could have saved a life is not warmer than counting.

Three doubts

My first doubt is about what the counting leaves out. The interventions that rank highest are those whose effects can be measured: bed nets, vitamin supplements, cash transfers. These are real goods. But many of the things that most change poor people's lives over the long run, such as working courts, honest officials, secure land rights and strong unions, are very hard to measure, and so they tend to fall off the bottom of the spreadsheet.

The philosopher Leif Wenar made a version of this point in an essay called "Poverty Is No Pond". The pond case is simple: one child, one rescuer, one obvious action, no side effects. Aid is never like that. It enters a political economy, and it can strengthen or weaken the governments, markets and relationships around it. Singer's analogy, Wenar argued, hides exactly the complexity that decides whether help actually helps.

Defenders reply, fairly, that they know this and try to account for it. But I notice that the pull of the method is still towards the countable.

My second doubt is older, and it comes from Bernard Williams. Utilitarianism, he argued, treats a person's own projects and commitments as just more inputs to the sum, to be overridden whenever the numbers say so. But a person is not a channel through which the world's welfare flows. Our commitments, to people, to work, to a place, are what make our lives ours. A morality that asks us to stand ready to drop them whenever a better use of our time appears attacks our integrity in a quite literal sense.

I see this in my friends. The ones who take the logic most seriously sometimes end up treating their own careers, relationships and even hobbies as instruments. Some of them burn out. The movement knows this and now talks a good deal about sustainability. But I think Williams saw that the problem isn't only practical. It is a sign that something in the theory has gone wrong.

My third doubt is about where much of the movement has gone. In What We Owe the Future MacAskill defends longtermism, the view that positively influencing the long-term future is a key moral priority of our time, because the future could contain vastly more people than the present. Toby Ord's work on existential risk makes a related case.

I don't think caring about the future is wrong. But I've argued here before that tiny probabilities of vast payoffs shouldn't dominate our decisions, and longtermism leans on reasoning of exactly that kind. When a speculative research project beats malaria nets on expected value only because the number of possible future people is so enormous, I start to suspect the method is producing conclusions nobody would accept if they arrived any other way.

Amia Srinivasan made a sharp version of a related worry in a 2015 review of MacAskill's earlier book: that this style of reasoning tends to leave the existing order unquestioned, asking how much good each of us can do within it rather than whether it should change.

Where I've landed

So I'm still in the group. I gave ten per cent last year, mostly to the unglamorous, measurable things the spreadsheets like, because I think the first argument holds even if the others don't. What I've given up is the hope that one method could tell me everything I ought to do.

The most honest summary I can offer is this. Effective altruism is very good at answering the question "given that I'm going to give, where will it do the most good?" It is much less good at answering "what kind of life should I lead?" or "what kind of world should we build?" The mistake, I think, is treating the first question as if it could answer the other two.