Here is one way the world could be. By far the best opportunities for making the world better can be supported by philanthropic money. They last for years. They can be invested in a vast number of times. They can be justified using widely available information and widely verifiable judgments.
Here is another way the world could be. By far the best opportunities are one-off actions that must be done by small numbers of people in the right fleeting time and place. The information that would be needed to justify them is half a lifetime’s worth of observations, many of which would be impolite to publish. The judgments needed must be honed by the same.
These worlds illustrate opposite ends of a spectrum. The spectrum is something like, ‘how much doing good in the world is amenable to being a big, slow, public, official, respectable venture, versus a small, agile, private, informal, arespectable one’.
In either world you can do either. And maybe in the second world, you can’t actually get into those good spots, so the relevant intervention is something like trying to. (If the best intervention becomes something like slowly improving institutions so that better people end up in those places, then you are back in the first world).
An interesting question is what factor of effectiveness you lose by pursuing strategies appropriate to world 1 versus those appropriate to world 2, in the real world. That is, how much better or worse is it to pursue the usual Effective Altruism strategies (GiveWell, AMF, Giving What We Can) relative to looking at the world relatively independently, trying to get into a good position, and making altruistic decisions.
I don’t have a good idea of where our world is in this spectrum. I am curious about whether people can offer evidence.
Meteuphoric 
Couldn’t it depend on the person, too? Maybe some people have skills/personalities that are better suited to working in large ventures (and would have more impact that way), and others are better suited to working in smaller ones.
This is especially relevant if our world is close to the middle of the spectrum, so there exist good opportunities on both sides.
It seems clear that given reasonably common benevolence, reasonable forms of social organization etc, the second version of the world is a long run equilibrium, while the first version is something you start with until you have reasonable social organization and benevolence. To claim to be in the first type of world is to claim to be in a transition towards benevolence and organization, an extraordinary and common claim that needs evidence.
“given reasonably common benevolence”
Farcical to assume this in reality. Not to mention people have grossly incompatible ideas what “benevolence” would imply.
Your mileage may vary, but I would assert that its farcical to assume that benevolence is very rare but that you personally are special and benevolent and therefore, whatever you personally do to advance your career is ‘for the greater good, in the long term’ and thus it is morally imperative to help you rather than more normal, possibly needier but less special and benevolent people. Sounds to me like the perfect central tendency of a privileged in-group’s perspective.
I never claimed any of these things.
I think we’re in a general low-benevolence equilibrium. Whatever benevolence people do have is limited to a narrow margin, range of recipients, and conditions, not well defined, and often contradict and therefore counteract what other people consider benevolence.
This comment clarified your perspective for me.
At least together with imagining a world where 50 years ago, aid to the 3rd world was actually extremely effective. And in that world it feels easy to imagine that people would have been extremely excited about it. That would continue until it stopped being effective.
Curious what you think of this:
My impression is that GiveWell really did signal boost AMF. Malaria nets really were underfunded. So the “we’re smart people, so lets get together and look at the evidence and be reasonable” thing did work out. It didn’t work out nearly as well as GiveWell etc. initially hoped, but it really did work out.
Does that story seem wrong to you?
The implicit claim by GiveWell and EA is that they know how to organize better.
They implicitly claim to be better at guessing what problems are relatively important and what methods are likely to succeed than average. That means they can find useful things faster and give them wider attention faster (Malaria nets, lab grown meat, xrisk, land use reform, etc.).
There’s a good reason to think that the very most effective actions are in the second category: http://www.overcomingbias.com/2012/11/marginal-charity.html
Thanks Robin!. This is critically important for the 80K Hours research program to internalize… And very psychologically difficult for almost anyone to internalize actually!
First thought: decompose into path dependence and parameter sensitivity
Four cases:
high path dependence, low parameter sensitivity:
world is mostly reliant upon large, slow moving trends that you can’t change.
high path dependence, high parameter sensitivity:
world state is influenceable by small actors, which parameters it is sensitive to is changing over time. Impact relies on predicting in advance that the world will turn out to be sensitive to a particular parameter during a particular window (AI safety)
low path dependence, high parameter sensitivity
world state is hard to predict, changes won’t necessarily cause any useful lock-in. Chaos.
low path dependence, low parameter sensitivity
world is falling into stable low energy states continuously, median voter theorem, memeplexes, moloch. Predictable but not changeable, cassandra complexes.
We act as if we live in the 2nd case because actions are mostly useless in the other 3? Not sure this is even a good frame, or what other implications might be.
Are we still pretending to really care? Or was it pretending to pretend to really care? I forget.
Just to spell out something that I wasn’t ‘holding’ in my mind at first after reading this, but we should generally support the search for the types of interventions best in the first class of world (relative to the cost of searching) even if we’re fairly well convinced that we live in the second class of world.