Monthly Archive for April, 2007

iPod sil­hou­ettes by roy­al­blue

Here is a fas­ci­nat­ing arti­cle from the New York Times. It refers to “cumu­la­tive advan­tage”, the phe­nom­e­non we see with a hit TV show like Lost, an iPod, or a overnight music star such as Norah Jones. As more and more peo­ple dis­cover that their friends enjoy a par­tic­u­lar prod­uct, the inter­est in this prod­uct snow­balls. The arti­cle is based on data from the Music Lab project, a study which I par­tic­i­pated in. It drew some inter­est­ing, although not unpre­dictable, con­clu­sions. Read it in full to find out more.

Con­ven­tional mar­ket­ing wis­dom holds that pre­dict­ing suc­cess in cul­tural mar­kets is mostly a mat­ter of antic­i­pat­ing the pref­er­ences of the mil­lions of indi­vid­ual peo­ple who par­tic­i­pate in them. From this common-sense obser­va­tion, it fol­lows that if the experts could only fig­ure out what it was about, say, the music, song­writ­ing and pack­ag­ing of Norah Jones that appealed to so many fans, they ought to be able to repli­cate it at will. And indeed that’s pretty much what they try to do. That they fail so fre­quently implies either that they aren’t study­ing their own suc­cesses care­fully enough or that they are not pay­ing suf­fi­ciently close atten­tion to the chang­ing pref­er­ences of their audience.

The common-sense view, how­ever, makes a big assump­tion: that when peo­ple make deci­sions about what they like, they do so inde­pen­dently of one another. But peo­ple almost never make deci­sions inde­pen­dently — in part because the world abounds with so many choices that we have lit­tle hope of ever find­ing what we want on our own; in part because we are never really sure what we want any­way; and in part because what we often want is not so much to expe­ri­ence the “best” of every­thing as it is to expe­ri­ence the same things as other peo­ple and thereby also expe­ri­ence the ben­e­fits of sharing.

The rea­son is that when peo­ple tend to like what other peo­ple like, dif­fer­ences in pop­u­lar­ity are sub­ject to what is called “cumu­la­tive advan­tage,” or the “rich get richer” effect. This means that if one object hap­pens to be slightly more pop­u­lar than another at just the right point, it will tend to become more pop­u­lar still. As a result, even tiny, ran­dom fluc­tu­a­tions can blow up, gen­er­at­ing poten­tially enor­mous long-run dif­fer­ences among even indis­tin­guish­able com­peti­tors — a phe­nom­e­non that is sim­i­lar in some ways to the famous “but­ter­fly effect” from chaos the­ory. Thus, if his­tory were to be some­how rerun many times, seem­ingly iden­ti­cal uni­verses with the same set of com­peti­tors and the same over­all mar­ket tastes would quickly gen­er­ate dif­fer­ent win­ners: Madonna would have been pop­u­lar in this world, but in some other ver­sion of his­tory, she would be a nobody, and some­one we have never heard of would be in her place.

Because it’s not pos­si­ble in the real world to test the­o­ries about events that never hap­pened, most of what we know about cumu­la­tive advan­tage has been worked out using math­e­mat­i­cal mod­els and com­puter sim­u­la­tions — an approach that is often crit­i­cized for gloss­ing over the rich­ness of real human behav­ior. For­tu­nately, the explo­sive growth of the Inter­net has made it pos­si­ble to study human activ­ity in a con­trolled man­ner for thou­sands or even mil­lions of peo­ple at the same time. Recently, my col­lab­o­ra­tors, Matthew Sal­ganik and Peter Dodds, and I con­ducted just such a Web-based exper­i­ment. In our study, pub­lished last year in Sci­ence, more than 14,000 par­tic­i­pants reg­is­tered at our Web site, Music Lab (www.musiclab.columbia.edu), and were asked to lis­ten to, rate and, if they chose, down­load songs by bands they had never heard of. Some of the par­tic­i­pants saw only the names of the songs and bands, while oth­ers also saw how many times the songs had been down­loaded by pre­vi­ous par­tic­i­pants. This sec­ond group — in what we called the “social influ­ence” con­di­tion — was fur­ther split into eight par­al­lel “worlds” such that par­tic­i­pants could see the prior down­loads of peo­ple only in their own world. We didn’t manip­u­late any of these rank­ings — all the artists in all the worlds started out iden­ti­cally, with zero down­loads — but because the dif­fer­ent worlds were kept sep­a­rate, they sub­se­quently evolved inde­pen­dently of one another…

The rest of the arti­cle, includ­ing its results and sub­se­quent con­clu­sions, can be found at: Justin Tim­ber­lake — Cul­ture — Hol­ly­wood — Idea Lab — New York Times.