If you don't have good evidence one thing is better than another, don't pretend you do

byRobert_Wiblin4y21st Dec 20157 comments


One of the most common criticisms of people involved in EA from people who are not, is that we come across as arrogant and overconfident about our ideas even when we know very few concrete facts about the alternatives.

I think that is a fair criticism.

One of the most common ways this manifests is that we act as if we are confident that the projects we support are better than alternatives that are radically different.

How does the Against Malaria Foundation compare to lobbying for more effective US aid, developing better curricula for US students, basic cancer research, let alone publishing papers about global catastrophic risks?

I don't know with much confidence, and I've been hearing people try to answer these questions for years. A lot depends on specific details I can't currently keep up with.

As an example, if someone says they know about and are giving to one of the others in that list, some humility is called for. Here are some reasonable things that you can say:

  • I suspect lobbying for legislation in Congress isn't likely enough to work at the moment, but I'm happy to be convinced otherwise.
  • I suspect research into catastrophic risks is about as likely to have a positive as negative effect, so have not been convinced to support it yet.
  • I don't know anything about US education policy, so wouldn't feel comfortable giving to that currently. I also suspect focussing on very poor countries offers better leverage, as there are a lot of players in the US education space.
  • I'd rather give to an intervention I understand well, and GiveWell's recommendations allow me to do that. Maybe you should read GiveWell's research and see how persuasive you find it compared to what you already know?
In addition to being more polite, this is also likely to lead you to learn more, and is more intellectually honest than blustering through pretending to have answers we don't yet have.