Businessweek Archives

Apple iTune's Genius: Is it racist race-based?


Update: This post angered quite a few people. I’ve used their input and my own thinking to write another version of the post. Please check it out. Thanks.

ITunes users probably have been asked in recent weeks to sign up for the Genius program—an arrangement where Apple gets permission to study our song lists and then creates play lists for us. I’ve been trying it out. It comes up with good stuff, dredging up forgotten or unknown tracks and mixing them with favorites.

This goes along with the the digital butler trend, in which we sacrifice data about our tastes and preferences, and receive (automated) personalized services and recommendations in return. Amazon, Netflix, Last FM, Pandora, they’re all doing it more and more. (And with Google’s Latitude and other geo-tracking services, this trend is extending into the physical world.)

Back to iTunes. I think its decision-making algorithms focus on race to the exclusion of nearly everything else—at least at this stage in their development. The way Genius works, you highlight a song in your library, and with the push of a button, the system creates a playlist of 25 or 50 songs (or even more) that fit the same musical theme. (If your starting song is too obscure, Genius tells you to pick another.)

For a test, I try Aretha Franklin. There’s no denying, of course, that she’s African American. She embodies Motown’s soul tradition. So I’d expect most of the songs grouped with hers to be by black musicians. It turns out that every single one of them is. I would point to other, non-racial, variables to consider. She’s a woman, and could be placed with other women influenced by blues, like Bonnie Raitt. The Aretha song I picked, “You Make Me Feel (like a Natural Woman”), was written by Carole King. One of her songs could go on an Aretha list.

But its easier for a new system like this one to start with clear lines. Motown goes with Motown. As we use Genius more, presumably the system will deepen its analysis. Perhaps it will study the playlists we create, or incorporate the other songs that Aretha buyers download.

I test drove the recommendation engines at to Last FM and Pandora. Both stick to race. I was a bit surprised, because Pandora claims to represent the "Music Genome Project." It sounds like it would dig for similarities beyond race. But no. Type in Aretha, and Pandora dishes up Al Green, Marvin Gaye, The Staple Singers--all the regulars.

Pandora offers other deciding factors, such as "major key tonality," "acoustic rhythm piano," even "heavy melodic ornamentation." But despite all these nods to complexity, where race figures prominently, as it does in Motown, it trumps all else.

Elsewhere, there are plenty of exceptions. Tracy Chapman and Joan Armatrading fit into folksy or jazzy groups that are multi-ethnic. Jimi Hendrix is with the rockers. But those are cases of black musicians who don't fit within one recognizable group, like Motown.

I think Genius reflects the normal human way of dealing with complexity. It reaches for a simple rule in the beginning. These people are like this, those ones like that. And then, hopefully, it starts to learn.

So there you have it: Just as society is showing signs of looking beyond race, our futuristic services come programed with yesterday's simpler formulas.

UPDATE: In comments, HJ asks if I experimented with other ethnic groups. I did. I have lots of Brazilian and Spanish-language music. Genius blended them, which is about what I'd expect. But when I started a list with Italian singer Paolo Conte, who bangs out caberet-style songs on his piano--a bit like Randy Newman--I get the Irish Chieftains, Africans Mamadou Diabake and Ali Farka Toure, Algerian Rai singer Cheb Mami, and some Brazilian Bossa Nova. I think the only thing that group has in common is that they're all foreign to the United States. Again, a single and simple variable.


American Apparel's Future
LIMITED-TIME OFFER SUBSCRIBE NOW

(enter your email)
(enter up to 5 email addresses, separated by commas)

Max 250 characters

 
blog comments powered by Disqus