This is a draft of an upcoming blog post for my blog, http://www.greatplay.net. I post it here because it deals with my changing thoughts on philanthropy, and because I'd like comments and feedback.

~

I've been thinking a lot about philanthropy again, and specific styles of how to think about philanthropy. How would you answer the question "Where should I give money to?". Most people just give to whatever charities come their way, send them advertisements, etc. A lot of people are also concerned with giving back to their local community, supporting their church, or their university.

I don't condemn any of those motivations -- I think they're wonderful, local communities do deserve support, and there's something to be said for "paying it forward" and giving back to those who have set you up for success. But recently, I've been reflecting on how I think about philanthropy and where I think my money should go, and I've changed the way I think.

My (Previous) Way of Thinking
The way I have always thought about philanthropy, however, was to be cause blind, and think solely on giving wherever I could do the most. If one opportunity came up where I could help someone cope with blindness by spending $50000 to get them a seeing eye dog, and another opportunity came up where I could cure blindness altogether through $25 cataract surgery, I would jump at the opportunity to fund the $25 surgery. I wouldn't care if that person were all the way in Africa and the blind person needing the seeing eye dog were in my local community.

It's not that I don't care about the local blind person. It's that I also care about the non-local blind person, and that I could treat 2000 people like him or her for the same amount of money. I think it's ludicrous for the local blind person to insist that she or he is 2000x more valuable to me solely because he or she lives nearby.

Then when it turned out that this large variation wasn't some made up hypothetical example, but actually real life, I was motivated to get up and do something about it. I saved up some of my money and looked for the most cost-effective place to donate it. Given my calculations, I decided to donate it to Effective Animal Activism, with the aim that somewhere between $0.01 and $16.48 is enough to prevent one year of animal suffering on a factory farm.

This is the general philosophy of GiveWell, Giving What We Can, and Effective Animal Activism -- investigate lots of non-profits, find the one organization that can do the most good with an additional dollar, and give them a dollar. ...Or far more.

However, I've recently come to see that this way of thinking is incomplete; perhaps even mistaken.

The Streetlight of Quantification
An old parable goes as follows:

Late one evening a police officer comes across a man on the way home from a party. He is quite drunk and looking for something under a streetlight.

"What are you looking for?" asks the policeman.

"My keys," the man replies, pointing down the road a little way, "I dropped them somewhere way over there."

The policeman is baffled, "Then why are you looking for them here?".

"Because there’s no light over there. The light is over here"

Obviously, we would reject the behavior of the drunk man in the parable as irrational -- no matter how much light is over where he's looking, he's not going to find his keys, because they're not in the light.

The bottom goal of all of my donating has been to have the largest impact per dollar, where I particularly measure impact in vague utilitarian terms of making lives go better. Currently, I've been looking at the opportunities that can be easily quantified. However, I have been ignoring the opportunities that cannot be quantified.

Applying the metaphor of the streetlight, quantification is that very light -- only certain things can be quantified easily. I, put not so kindly, are the drunk person, searching in the light of easy quantification because... well... that's where the quantification is. Even if the best opportunity for maximal impact exists somewhere else.

Of course, there's something to be said for the streetlight. Givewell writes "In Defense of the Streetlight Effect", pointing out that there are plenty of good opportunities to be explored within the areas of easy quantification, and I agree. On an impact per dollar point-of-view, there is a lot to be gained by switching from the Seeing Eye Dog charity to the cataract surgery charity, based on our assessment of their impact. GiveWell's top charities do really great work for not that much money, and are definitely worth funding.

However, I just don't think they're our absolutely best bet. I think the best bet lays outside our streetlight.

The Far Future Blind Spot
What could be missing from our streetlight? I think we're missing a long-term view of the far future.

When I consider spending $2300 to save a life by donating to GiveWell's top charity, the Against Malaria Foundation, the first thing that I question is what it actually means to "save a life". The person I save may, instead of dying at age five from malaria, go on to live a pretty decent life, but still die of old age at 60. I haven't "saved" the life -- instead, I've happily extended it by 55 years.

But I'm also not considering the further impacts of extending this life. It turns out that by helping reduce the mortality burden of malaria, I'm helping to lead people to have less children overall, because parents in the developing world have less kids when they're less concerned their kids will die of malaria. Overall, this is going to prevent many years of life that otherwise would have been lived.

So where is my actual impact? More realistically, it's not in "saving a life", but rather in reducing overpopulation, helping to reduce concerns from malaria, and helping turn people's attention to better and happier things. In saving a life, I'm really helping to develop the developing world.

The same is true in considering Vegan Outreach. Sure my penny might, under more optimistic assumptions, be all that is needed to prevent a year of intense animal suffering. But my real impact is going to be from making the person I've helped indirectly convert to vegetarianism more compassionate, more likely to recruit others, and more likely to get the ball rolling on ending factory farming.

Right now, there are billions of people alive, hundreds of billions of factory-farmed animals, and trillions of nonhuman animals in the wild. There's a lot I can do to help them, and I do my best to care about all of them. But, as I outlined in discussing existential risk, this is going to pale in comparison to the trillions of people and the quadrillions of nonhuman animals who will be born over the next thousands of years.

If I can change the present a little bit, I may help the people alive now a little bit. But, through compounding effects, movement building, and momentum, I'll have a lot more impact on the people in the farther future. I just might not be able to quantify it.

To be fair, GiveWell has been looking away from the "cost of a life" metric and toward measures of economic development. But this way of thinking for the far future is pretty alien to most people, and was alien to me even back in Dec 2012. I don't think we should discard quantification altogether, of course. But I do think it needs to be retooled in our thoughts toward focusing on where the real impact may lie, not in quantifying our marginal, more minimal impact on the present.

Going Meta With My Money
So where am I giving my money now? With a look toward the far future, existential risk reduction looks like a much better bet. But, more realistically, I have to say that no one I know has a convincing enough answer yet on where we should donate. There's a lot of good work to be done in reducing existential risk, in reducing nonhuman animal suffering, and in raising the standard of living for the developing world. But I earnestly can't tell which opportunity is the smartest.

I had thought at first I should just start saving my money and wait until either I die, or a convincing enough answer comes along. But then I thought that there's a lot that can be done by just recruiting more and more people to start thinking about these problems. If I can recruit just one person to think about these problems as deeply and as intelligently as I currently am (or more!), I will have doubled my impact. Then, imagine what happens if they start recruiting people as well!

So, I want to invest my money in the movement. Effective Animal Activism was a pretty good bet, though now for completely different reasons than I once thought. I'm going to start giving my 20% - $2400 toward The Centre for Effective Altruism, which does a lot to collect a group of people together to think about where to give and to give more. I think there's a good chance that not only could I get more than $1 to go to a good cause for every $1 I give them, but that I will be able to get a lot more thought and energy going toward figuring out how to better improve our world.

After all, if they didn't exist, I very likely never would have even started thinking about these issues.

Interesting. I sort of disagree with your argument and support your conclusion

It seems to me that while in some theoretically optimal universe we should look outside of quantifiable areas for potential better bets, the problem is that without a proper search tool (and inasmuch as we have one, the problem doesn’t arise), we’re basically picking a point at random on a near infinite map. Even if we pick the right point, we might never know it.
To revisit the street lamp analogy, we could search in the dark for the keys for our entire lives, whereas if we search under the street lamp we might just find a fiver to at least fund our new set of keys – not as good as having them now, but a lot better and a lot quicker than the alternative. (obviously when you stretch the analogy far enough, you probably give up and stay at a friend’s, but I think we’re assuming only two basic possibilities here)

Because of the distribution of effectiveness in quantifiable causes, the gain we get from exploring them is potentially huge. Some people would only have been looking under the street lamp anyway, and we’ve made them far more efficient by signposting them to the best thing they’re going to find under it. Others would in principle search wherever seemed highest expectation, but signposting under the streetlamp has so much more powerful an effect than signposting in the dark that there’s a strong chance it will tip the balance for them, too.

By the nature of unquantifiability, if we look at all unquantifiable causes, we’re only ever going to have an expected return of the mean value of them all, so the fact that there might be some incredibly good ones available is largely irrelevant.

Obviously this division between streetlamp and darkness is hugely simplistic in practice, and we can often make semi-informed estimates about things on which we have no real data (eg the mean expectation of unquantified activities). But dealing with (non-absolute) uncertainty is still a form of statistical analysis, so it’s not that we shouldn’t be doing it, just that we shouldn’t restrict ourselves to an arbitrarily chosen level of confidence in our expectation before we recommend it.

But actually, Givewell and CEA (esp GWWC) have had that dialogue for at least a couple of years now. See eg Toby Ord’s responses in Why We Can’t Take Expected Value Estimates Literally (Even When They’re Unbiased). There was also a discussion on another Givewell post between Holden and Will on a similar subject, which I now can’t remember. To (over)simplify, Givewell are generally more interested in easily quantifiable estimates (eg cost to save a life), while GWWC seem to be happier with higher uncertainty at least if it ultimately allows you to quantify something closer to a measure utility (eg QALYs).

So on the whole, we’d probably both agree more with GWWC. That said, it is an oversimplification – Givewell obviously implicitly look at something resembling utility. Also Holden raises the point that focusing on quantification as a key has the benefit of providing a strong incentive for potential EA organisations to focus on providing quantitative data, which has a potentially huge long-term benefit of extending the range of the lampost. And I get the impression both organisations have been improved by the other.

So all that considerd, I’m a big fan of CEA, not because they look outside the light but because they look on the fringes of it – and because they combine it with persuading more people to look, and persuading those already looking to look more.

I ended up incorporating most of your feedback into the final draft. I'm still not sure if it's saying what I really wanted to say, but I think it's better.