“A sig­nif­i­cant per­cent­age of online con­tent is now pub­lished and con­sumed via the mech­a­nism of crowd­sourc­ing. While any user can con­tribute to these forums, a dis­pro­por­tion­ately large per­cent­age of the con­tent is sub­mit­ted by very active and devoted users, whose con­tin­u­ing par­tic­i­pa­tion is key to the sites’ suc­cess. As we show, people’s propen­sity to keep par­tic­i­pat­ing increases the more they con­tribute, sug­gest­ing moti­vat­ing fac­tors which increase over time. This paper demon­strates that sub­mit­ters who stop receiv­ing atten­tion tend to stop con­tribut­ing, while pro­lific con­trib­u­tors attract an ever increas­ing num­ber of fol­low­ers and their atten­tion in a feed­back loop. We demon­strate that this mech­a­nism leads to the observed power law in the num­ber of con­tri­bu­tions per user and sup­port our asser­tions by an analy­sis of hun­dreds of mil­lions of con­tri­bu­tions to top con­tent shar­ing web­sites Digg​.com and Youtube​.com.”

No plug-​​ins required – it can only use fea­tures natively sup­ported by the client
Com­pat­i­bil­ity – it has to work on every major browser on the mar­ket
Ease of use – no or near-​​zero con­fig­u­ra­tion needed for stan­dard use cases
Speed – it has to be fast, even for suf­fi­ciently large amounts of text
And now, after nearly a year of plan­ning and research we believe that these require­ments have been met.”

The algo­rithm com­bines pat­tern search and par­ti­cle swarm. Basi­cally, it applies a direc­tional direct search in the poll step (coor­di­nate search in the pure sim­ple bounds case) and par­ti­cle swarm in the search step.”