Algorithms That Rule the Web
The Love Algorithm
Chemistry may be the decisive factor in the human phenomenon of falling in love, but algorithms provide the matchmaking spark for many who use online dating services such as eHarmony and Match.com. Finding the perfect je ne sais quoi for potential lovebirds requires sites such as Match.com to number-crunch users' personal-attraction tests to look beyond the bare-bones facts that someone is a "Jewish nonsmoker who likes swing dancing."
At eHarmony, the service plugs your answers to the site's 258-question personality test into company's ultimate trade secret: its love algorithm. In a 2008 article in the New York Times, eHarmony said that 19 million people had taken its personality test, and a study it commissioned concluded that it was responsible for 2 percent of all U.S. marriages in 2007.
The Perfect Online Ad Algorithm
Online advertising sits at the crossroads of commerce and algorithm deployment. Its objective is to display the right ad to the right person at the right time. An advertising algorithm that succeeds in this mission can mean the difference between a sale and no sale. To better their odds, advertisers use algorithms to slice and dice a complex mix of data.
The algorithms are so byzantine that they can be very difficult to grasp. I have examined them in several articles including "Good-Bye to Privacy?" and in a point-counterpoint editorial, "Privacy Backlash Over Ad Tracking Debated." In a nutshell, sophisticated online advertisers pair offline demographic data about you with your Web surfing habits in order to entice you with targeted online ads.
Some observers argue ads that profiling you and presenting you with relevant ads based on your surfing habits help Web content owners stay in business and deliver high-quality content. Others say that trusting private companies with massive databases of user profiles is like putting foxes in charge of henhouse security.
Shopping and Recommendation Algorithms
Does Amazon's recommendation engine have you figured out? Probably.
Amazon's algorithm objectively analyzes the buying patterns of millions of customers. Then, if you buy the book State of Wonder by Ann Patchett, Amazon recommends other books based on the titles that other buyers of State of Wonder have purchased. As a result, Amazon may be able to sell you something else that you hadn't intended to buy.
Recommendation engines enable e-merchants such as Amazon to sell billions of dollars worth of merchandise by helping consumers find what they're looking for and by fostering impulse buys. In an interview with CNet, an Amazon spokesperson said, "Algorithms are what make our site run, [and] such a unique place to shop."
In 2009 Netflix doled out $1 million in prize money to a group of statisticians known collectively as BellKor's Pragmatic Chaos for their success in boosting the movie rental company's accuracy at predicting the movies that customers would like rent. To earn the prize, they had to consider demographic and behavioral data along with zip codes, genre ratings, and 100 million movie ratings.
Pandora Decodes Music
The music service Pandora has demonstrated uncanny accuracy in matching a music listener's tastes based on a single song--and once again, an algorithm is responsible. Pandora's Music Genome Project has as its goal to "capture the essence of music at the fundamental level," according to Tim Westergren, founder of the Music Genome Project and cofounder of Pandora.
Pandora says that it uses 400 attributes to describe a song. Next, according to the description on Pandora's Facebook page, the service's algorithm parses that data from one song and can "play a range of music that is 'musicologically similar' to your starting points in some way--but not always necessarily music that 'sounds similar.'"
While competing online music services have faltered in recent years, Pandora claims a growing user base of 100 million registered users, of whom 36 million are active users.
The Death of Serendipity
Do algorithms signal the death of serendipity--if not free will and privacy--online? The debate will rage for years. But as we quickly morph into the supercomputer age of gargantuan databases, algorithms need to be protected from exploitation by Orwellian governments, sociopathic hackers, and intrusive companies, privacy experts warn. Unfortunately, few laws have caught up with technology in this area.
Current do-not-track legislation is working its way through Congress, and other initiatives put forth on the state level, such as in California, tangentially address how companies gather and use data. But so far the lords of the algorithms have the upper hand.
Enjoyed reading this story? PCWorld's algorithm thinks you might also like to read these: