How Trends Actually Spread; or, Six Degrees but No Connectors

The small sample size of Stan­ley Mil­gram’s small world exper­i­ment means that the the­ory of ‘six degrees of sep­ar­a­tion’ and the con­clu­sion drawn from it–primarily, the Influ­en­tial’s the­ory pop­ular­ised by Mal­colm Glad­well in The Tip­ping Point–could be deeply flawed. That was the start­ing point for Duncan Watts’ research that led him to say “the Tip­ping Point is toast”.

So to research how ideas and trends spread vir­ally, Watts (who is author of Everything is Obvi­ous, prin­cip­al research sci­ent­ist at Yahoo! Research (he dir­ects their Human Social Dynam­ics group), and found­ing dir­ect­or of Columbia Uni­versity’s Col­lect­ive Dynam­ics Group) ran large-scale repro­duc­tions of the small world exper­i­ment and hun­dreds of com­puter sim­u­la­tions that brought for­ward two con­clu­sions: the six degrees of sep­ar­a­tion the­ory is cor­rect, but there is no evid­ence for super-con­nec­ted ‘trend gate­keep­ers’ (such as Glad­well­’s ‘Con­nect­ors’):

But Watts, for one, did­n’t think the gate­keep­er mod­el was true. It cer­tainly did­n’t match what he’d found study­ing net­works. So he decided to test it in the real world by remount­ing the Mil­gram exper­i­ment on a massive scale. In 2001, Watts used a Web site to recruit about 61,000 people, then asked them to ferry mes­sages to 18 tar­gets world­wide. Sure enough, he found that Mil­gram was right: The aver­age length of the chain was roughly six links. But when he examined these path­ways, he found that “hubs”–highly con­nec­ted people–weren’t cru­cial. Sure, they exis­ted. But only 5% of the email mes­sages passed through one of these super­con­nect­ors. The rest of the mes­sages moved through soci­ety in much more demo­crat­ic paths, zip­ping from one weakly con­nec­ted indi­vidu­al to anoth­er, until they arrived at the tar­get. […]

[His com­puter sim­u­la­tion] res­ults were deeply coun­ter­in­tu­it­ive. The exper­i­ment did pro­duce sev­er­al hun­dred soci­ety­wide infec­tions. But in the large major­ity of cases, the cas­cade began with an aver­age Joe (although in cases where an Influ­en­tial touched off the trend, it spread much fur­ther). To stack the deck in favor of Influ­en­tials, Watts changed the sim­u­la­tion, mak­ing them 10 times more con­nec­ted. Now they could infect 40 times more people than the aver­age cit­izen (and again, when they kicked off a cas­cade, it was sub­stan­tially lar­ger). But the rank-and-file cit­izen was still far more likely to start a con­ta­gion.

I can­’t help but find it some­what iron­ic that, writ­ten almost four years ago, this argu­ment has­n’t really gained much trac­tion and Glad­well­’s ideas are still dis­cussed ad nauseam.