[Question] What questions would you like to see forecasts on from the Metaculus community?

Me­tac­u­lus is an on­line plat­form where users make and com­ment on fore­casts, which has re­cently been par­tic­u­larly no­table for its fore­cast­ing of var­i­ous as­pects of the pan­demic, on a ded­i­cated sub­do­main. As well as dis­play­ing sum­mary statis­tics of the com­mu­nity pre­dic­tion, Me­tac­u­lus also uses a cus­tom al­gorithm to pro­duce an ag­gre­gated “Me­tac­u­lus pre­dic­tion”. More in­for­ma­tion on fore­cast­ing can be found in this in­ter­view with Philip Tet­lock on the 80,000 hours pod­cast.

Ques­tions on Me­tac­u­lus are sub­mit­ted by users, and a thread ex­ists on the plat­form where peo­ple can sug­gest ques­tions they’d like to see but do not have the time/​skill/​in­cli­na­tion to con­struct them­selves. Ques­tion con­struc­tion is non-triv­ial, not least be­cause for fore­cast­ing to work, clear crite­ria need to be set for what counts as pos­i­tive re­s­olu­tion. A use­ful in­tu­ition pump here is “if two peo­ple made a bet on the out­come of this ques­tion, would ev­ery­one agree who had won?”

Although there is already a sig­nifi­cant over­lap be­tween the EA com­mu­nity and the Me­tac­u­lus user­base, I think it is likely that there ex­ist many fore­cast­ing ques­tions which would be very use­ful from an EA per­spec­tive, but that have not yet been writ­ten. As such, I’ve writ­ten this ques­tion as both a re­quest and an offer.

The re­quest:

Have a think about whether there are any fore­casts you think have the po­ten­tial to have a large im­pact on de­ci­sion mak­ing within the EA com­mu­nity.

The offer:

If you do think of one, and post it be­low, and I’ll write it up for you and sub­mit it to the site. The closer it is to “fully formed”, the more quickly this is likely to hap­pen, but please don’t feel the need to spend ages choos­ing re­s­olu­tion crite­ria, I’m happy to help with this. I in­tend to choose ques­tions based on some com­bi­na­tion of num­ber of up­votes the sug­ges­tion has and how easy the ques­tion is to op­er­a­tional­ise.

Ex­am­ples of my ques­tion-writ­ing on Me­tac­u­lus are here, and I also re­cently be­come a mod­er­a­tor on the plat­form.

Some ex­am­ples of EA-ad­ja­cent ques­tions already on the plat­form:

How much will GiveWell guess it will cost to get an out­come as good as sav­ing a life, at the end of 2021?

On De­cem­ber 1st, 2023 how many com­pa­nies wor­ld­wide will pledge up­hold GAP stan­dards for broiler chick­ens raised for meat?

How many re­views will Toby Ord’s book The Precipice have on Ama­zon on Jan­uary 1st 2021?

How many in­fec­tions of SARS-CoV-2 (novel coro­n­avirus) will be es­ti­mated to have oc­curred wor­ld­wide, be­fore 2021?

If you’re in­ter­ested in hav­ing some­one make a fore­cast about a ques­tion that’s more per­sonal to you, and/​or some­thing that you wouldn’t ex­pect the Me­tac­u­lus com­mu­nity as a whole to have the right com­bi­na­tion of in­ter­est in and knowl­edge of, I’d recom­mend check­ing out this offer from aman­dango.

No comments.