[Question] When does adding more people reliably make a system better?

Pre­dic­tion mar­kets have a re­mark­able prop­erty. They re­ward cor­rect con­trar­i­anism. They in­cen­tivise peo­ple to dis­agree with the ma­jor­ity con­sen­sus, and be right. If you add more traders to a mar­ket, in ex­pec­ta­tion they price will be more ac­cu­rate.

More traders means both more fish and more sharks.

(The movie “The Big Short” might be a very sad por­trait of the global fi­nan­cial sys­tem. But it’s still the case that a sys­tem in a bad equil­ibrium with deeply im­moral con­se­quences re­warded the out­casts who pointed out those con­se­quences with billions of dol­lars. Even though so­cially, no one both­ered listen­ing to them, in­clud­ing the US Govern­ment who ig­nored re­quests by one of the fund man­agers to share his ex­per­tise about the events af­ter the crash.)

Lots of things we care about don’t have this prop­erty.

  • Many so­cial com­mu­ni­ties de­cline as more mem­bers join, and have to spend huge amounts of effort build­ing in­sti­tu­tions and rit­u­als to pre­vent this.

  • Many com­pa­nies have their cul­ture de­cline as they hire more, and have to spend an in­cred­ible amount of re­sources sim­ply to pre­vent this (which is far from get­ting bet­ter as more peo­ple join). (E.g. big tech com­pa­nies can prob­a­bly have >=5 can­di­dates spend >=10 hours in in­ter­views for a a sin­gle po­si­tion. And that’s not count­ing the prob­a­bly >=50 can­di­dates for that po­si­tion spend­ing >=1h.)

  • On­line fo­rums usu­ally de­cline with grow­ing user num­bers (this hap­pened to Red­dit, Hack­erNews, as well as LessWrong 1.0).

In pre­dic­tion mar­kets the vet­ting pro­cess is re­ally cheap. You might have to do some KYC, but mostly new peo­ple is great. This seems like a re­ally im­po­rant prop­erty for a sys­tem to have, and some­thing we could learn from to build other such sys­tems.

What other sys­tems have this prop­erty?