Creating lists

Sup­pose you are try­ing to cre­ate a list. It may be of the “best” pop­u­lar sci­ence books, or most con­tro­ver­sial movies of the last twenty years, tips for get­ting over a breakup or the most in­ter­est­ing cat gifs posted in the last few days.

There are many rea­sons for want­ing to cre­ate one of these lists, but only a few main sim­ple meth­ods:

  1. Vot­ing model—This is the sim­plest model, but pop­u­lar­ity doesn’t always equal qual­ity. It is also par­tic­u­larly prob­le­matic for reg­u­larly up­dated lists (like Red­dit), where a con­stantly chang­ing au­di­ence can re­sult in large amounts of du­pli­cate con­tent and where eas­ily con­sum­able con­tent has an ad­van­tage.

  2. Cu­ra­tor model—A sin­gle ex­pect can of­ten do an ad­mirable job of col­lect­ing high-qual­ity con­tent, but this is sub­ject to their own per­sonal bi­ases. It is also effort in­ten­sive to eval­u­ate differ­ent cu­ra­tors to see if they have done a good job.

  3. Vot­ing model with (con­tent) rules—This can cut out the ir­rele­vant or sug­ary con­tent that is of­ten up­voted, but cre­at­ing good rules is hard. Often there is no ob­jec­tive line be­tween high and low-qual­ity con­tent. Th­ese rules can of­ten re­sult in con­flict.

  4. Vot­ing model with sec­tions—This is a solu­tion to some of the limi­ta­tions of 1 and 3. In­stead of declar­ing some things off-topic out­right, they can be thrown into their own sec­tion. This is the op­ti­mal solu­tion, but is usu­ally ne­glected.

  5. Vot­ing model with se­lec­tion—This cov­ers any model where only cer­tain peo­ple are al­lowed to vote. Some­times se­lec­tion is ex­traor­di­nar­ily rigor­ous, how­ever, it can still be very effec­tive when it isn’t. As an ex­am­ple, Metafilter charges a $5 one-time only fee and that is suffi­cient to keep the qual­ity high.

The main point is that model 1 shouldn’t au­to­mat­i­cally be se­lected. The other mod­els have ad­van­tages too.