Revisiting Guide Dogs and Blindness Prevention

Link post

In early effective altruism it was common for people to give the example of deciding between donating to fund the training of guide dogs in wealthy countries or to reduce blindness in very poor countries. For example, here’s what Giving What We Can used to say:

For example, suppose we want to help those who are blind. We can help blind people in a developed country like the United States by paying to train a guide dog. This is more expensive than most people realize and costs around $50,000 to train a dog and teach its recipient how to make best use of it. In contrast, there are millions of people in developing countries who remain blind for lack of a cheap and safe eye operation. For the same amount of money as training a single guide dog, we could completely cure enough people of Trachoma-induced blindness to prevent a total of 2,600 years of blindness.

That writing went up sometime in 2011 or earlier, and I think it’s where this comparison entered proto-EA, but if you know of earlier usage I’d be curious!

Over time, however, EAs have mostly moved away from this comparison. Instead global poverty EAs are more likely to give comparisons like:

  • A typical US income vs how much the world’s poorest people live on.

  • How much the US is willing to pay to save the life of one of its citizens vs how much it costs to save a life where that’s cheapest.

  • What the UK National Health Service is willing to spend per Quality-Adjusted Life Year (QALY) vs what it costs to provide a similar benefit to someone in a very poor region.

What these have in common is that there are charities working in these areas with strong evidence to support their cost-effectiveness. With GiveDirectly you can send money to people who are living on under $1/​day, or with the Against Malaria Foundation you can provide bednets that cut malaria deaths in ~half, at around $5k per life saved (~$100/​QALY).

On the other hand, while GiveWell and other EA groups have looked into vision surgery, there are no high quality evaluations. [1] What do you do if you say something like “while a guide dog costs $X you can restore someone’s sight for $Y << $X” and someone asks where they can donate to make that happen? Much better to give actionable comparisons.


[1] You might think of Hellen Keller International, another GiveWell recommendation, but looking at their cost-effectiveness model, GiveWell is primarily rating it highly for mortality reduction. Looking at their vision benefits writeup and BOTEC, linked from the cost-effectiveness model, most of the supplementation is going to people who wouldn’t otherwise become blind: yes, it’s two pills for ~$2.70, but if I’m interpreting GiveWell’s rough estimate correctly only 1:1,100 people who get the pill would otherwise become blind, and the supplementation isn’t 100% effective, so it ends up being about $3k to prevent a case of blindness this way. (more)

No comments.