Effective Altruism Prediction Registry

Pre­dic­tions mat­ter be­cause they sep­a­rate the sig­nal from the noise. Any­one can make a vague procla­ma­tion that some­time in the fu­ture some­thing good or bad will hap­pen; it’s the peo­ple who put defini­tive dates and mea­surable claims on these state­ments that make it pos­si­ble to eval­u­ate whether they were ac­tu­ally right or wrong, and figure out what we should do next.

Su­perFore­cast­ing makes a com­pel­ling case that track­ing pre­dic­tions is use­ful on both the in­di­vi­d­ual and col­lec­tive level. When we track a pre­dic­tion we’re able to get feed­back on whether we were right or wrong, and over­time we get bet­ter at mak­ing pre­dic­tions. And it helps groups make bet­ter de­ci­sions by forc­ing dis­ci­plined think­ing and avoid­ing un­clear and con­fus­ing opinions and HiPPO.

When­ever you make a de­ci­sion you’re mak­ing a pre­dic­tion. Th­ese can be small pre­dic­tions, like if I de­cide to leave at 7:30 I’ll make it to work on time, or big pre­dic­tions, like if I at­tend a pres­ti­gious uni­ver­sity I will be able to make a lot of money when I grad­u­ate. For Effec­tive Altru­ists these de­ci­sions are of­ten re­lated to char­i­ta­ble giv­ing, not just about what cause is the ‘best’ to give to, but also what is the marginal im­prove­ment you ex­pect your dona­tion to cause. I’m in­creas­ingly con­vinced that we need to do a bet­ter job as a com­mu­nity sys­tem­at­i­cally track­ing pre­dic­tions around im­por­tant top­ics.

If I give $3,000 to the Against Malaria Foun­da­tion, I’m pre­dict­ing I will be able to save one life. Or, if I think that EAs should be mak­ing ap­peals based on equal­ity or jus­tice, in­stead of in­di­vi­d­ual rights or out­comes, I’m pre­dict­ing that this strat­egy, if im­ple­mented, will drive more dona­tions than the other.

So we’re mak­ing pre­dic­tions all the time, but the pre­dic­tions sit in iso­la­tion from one an­other, in ran­dom posts in the var­i­ous EA face­book groups and fo­rums. And they lack the rigor of pro­vid­ing ac­tual dead­lines and num­bers, for when/​how we should as­sess the claims. This is a prob­lem, be­cause we’re miss­ing out on the wealth of knowl­edge that would come from learn­ing if/​how these ideas ac­tu­ally worked.

I think this is low hang­ing fruit for us to im­prove EA. If we tracked pre­dic­tions about the out­comes of cam­paigns, in­ter­ven­tions, etc. we’d see a num­ber of benefits.

  • Estab­lish a strong track record of suc­cess for top perform­ing charities

  • Provide guidance for de­ci­sion mak­ers. Pre­dic­tions, from orgs and in­di­vi­d­u­als, about EA cam­paigns, could help guide dona­tions.

  • Ele­vate good fore­cast­ers, and boost effec­tive ideas.

In par­tic­u­lar EA orgs could make more ex­plicit pre­dic­tions about what the out­come of giv­ing money would be. It was difficult to track down con­crete pre­dic­tions from no­table EA Orgs for the Win­ter 2015 giv­ing sea­son, but two I found were from 80000 Hours, pre­dict­ing 50 plan changes by 10/​31/​2016, and CFAR, with 1000 new alumni by 12/​31/​2016.

Some type of cen­tral repos­i­tory of pre­dic­tions be­ing made about EA, where peo­ple can com­ment, provide their own pre­dic­tions, and up­date their pre­dic­tions as new ev­i­dence comes in. This could be as sim­ple as a blog post or a more full fledged sys­tem/​mar­ket. Pre­dic­tion track­ers have a long his­tory, but with lit­tle suc­cess­ful adop­tion. I think that EA’s might be differ­ent be­cause of the unique ori­gins of the move­ment. Given the an­a­lyt­i­cal cul­ture of EAs, this could be uniquely well suited to our com­mu­nity.

I see three steps to do­ing this:

  • A sim­ple fo­rum where any­one can sub­mit a declar­a­tive pre­dic­tion about EA re­lated events.

    • By De­cem­ber 31st, 2016 there will be 5,000 peo­ple signed up for the Giv­ing What We Can pledge

  • Any­one can sub­mit their pre­dic­tions tied to this event

    • JohnSmith89 thinks there is a 20% prob­a­bil­ity this is true

  • After the dead­line has past judges/​peo­ple can vote on whether it ac­tu­ally hap­pened, and you can see who suc­cess­fully pre­dicted it and with what ac­cu­racy.

With 75% con­fi­dence I’d say that by Fe­bru­ary 10th at least 15 peo­ple will have ex­pressed in­ter­est in pre­dic­tions about effec­tive al­tru­ism.