Rob Wiblin

So the idea, yeah, even though you could have more impact in one particular village doing all three, it makes more sense with a given budget to do the two that are best and then try to reach more areas, because I guess that's easier to scale and

Yeah, so I didn't realize that charity entrepreneurship was quite so central to getting this started as it was.

Because for those who don't know, charity entrepreneurship is this project, I guess, yeah, an incubator run by people who are very sympathetic to effective altruist ideas.

And I guess they do a bunch of research where they try to kind of shortlist interventions that they would love to see people going and actually trying to implement.

in the real world.

And then I guess they try to match those with people like you who are thinking, I want to start something, but I'm not sure exactly what and I'm not sure who to do it with.

And I guess, yeah, pair up founders and then pair them up with an idea that's a good fit for them and then help them actually get moving.

I guess, yeah, maybe we'll come back and talk about that later because it could be an excellent resource or an excellent thing to go through for some people in the audience.

I saw on your website that you have this kind of table summarizing the evidence on SMS reminders.

And I

I took a quick look over and it seemed like there was five trials that had kind of positive effects and then five where it didn't really seem like the SMS reminders had moved the needle very much.

Did you go away and kind of either yourself or get someone else to do this kind of literature review to learn like other lessons about under what circumstances these things work and like what things you can vary about the intervention to make them have a bigger impact?

Yeah, I was really happy to see the table that both had studies with positive results and studies with no results where it couldn't find an effect.

Because an intervention, even one that's effective, you should find some studies that by chance have found that it didn't work in this particular case or perhaps were underpowered.

There weren't enough people in the sample to detect even a real effect.

And if people aren't aware of that when they're setting out, it suggests that they haven't really been quite comprehensive enough in looking at all of the evidence out there to get a proper average kind of effect size across the full range of results rather than just cherry picking a few of them.

Yeah.

So were there any studies that didn't find an effect where it was possible to kind of identify that some particular thing had been important?

You were suggesting, for example, literacy might mediate the effect of the messages.

If people can't read them, then that's going to reduce the effect size pretty substantially.