Tuesday, December 28, 2010

Quantifiable Data, Specificity and the rest of the stuff I think and feel

I really enjoy Pandora, and if you haven't checked them out yet, I highly recommend you do. The people behind the Music Genome Project did an incredible thing, that iTunes then imitated with their Genius mixes, etc. and it all boils down to this: it's like having a friend who knows more music than you ever thought it possible for one human to know, and he kinda knows you, and when you tell him what you like, he makes recommendations for you... only he's slightly more accurate in predicting your tastes than the guy who set you up on a blind date with his buddy back in '92. Most of the time, I'm able to create great stations from a single song or artist with minimal futzing. I get excited when somewhat obscure songs that I've liked for years get played along with pop gems and then I hear something I've never heard before and go off on a discovery tour of new music.

Sometimes, you get bizarre songs that you not only don't like, but can't figure out how you got from A to B. That's part of the problem with not being able to accurately identify things that are intangible. See, last week, I was in a mood. Not a bad mood, not a good mood, just a moody mood. I kept flipping through the stations that I had created* trying to find something that would fit. I didn't want my 80s Mix, Perky Dance, Glee-ful, Mr. Mister, Mellow, Grrrl Music, Tarlowski or Classic Christmas Carols. I couldn't even piece together what I wanted by using the QuickMix feature, which is awesome in that it lets you sort of shuffle lots of stations together.

So I started a new station. That wasn't really working very well either. I finally figured out that the problem wasn't that Pandora wasn't playing songs that I liked, but that I hadn't correctly defined what I wanted to listen to. Once I did that, I identified other problems.
What I wanted to listen to was moody songs, but Pandora doesn't really know what my moody songs are, and the list I came up with off the top of my head was musically diverse enough to be confusing. Fiona Apple, Terrence Trent D'Arby and Alkaline Trio have little in common musically for the Music Genome Project to draw on. The common element that caused them to be included in the seed songs was the way they make me feel. Luckily, the Music Genome Project looks at more than just the easily quantifiable data. The list of attributes identified includes objective observations like "Electric Guitar Solo" and somewhat subjective labels like "funky raps." Unfortunately, even with their comprehensive view of music and sounds, they can't accurately pinpoint all the things that make us like music. When creating a station, I want to be able to look at the list of attributes and put the ones I want into a list, in addition to picking the songs that are inspiring me.

However, the things that I'm thinking about aren't often on the list. There isn't an option for "bass lines that make me want to drive faster" or "guitar solos that sound awesome when you try to sing them" or "mopey girl songs" or "songs that remind me of that guy that I should probably never talk to again because I seem to lose my mind, my cool, and any semblance of independent thought whenever he's around." There's not even something as simple and quantifiable as "songs written or released in this decade". Best I can figure, I just have to put enough of the right songs in, and not be afraid to use the thumbs-down button to tell Pandora that what I'm looking for is different, even though that song has "mystical qualities" and "passionate atmosphere" and so did the last song I thumbed-up. I don't share the same vocabulary as the Music Genome Project to know that what I consider "joyful lyrics" is the same as their definition of "joyful."

I suppose you could take this to the nth degree and say that in the extreme we can't say that we ever know that our "joy" is the same as another person's definition and get all esoteric about no two people defining anything the same or seeing color the same way but that neglects the reality that the function of language is to communicate, even abstract concepts, if imperfectly due to the unique nature of each human experience but still with a degree of success because of the breadth of language available and the vast number of shared experiences.** No other person has the exact same mix of experiences that makes me think and behave the way I do, but for each experience, there's someone somewhere who can relate because of a similar, if not exact experience... granting that there will be exceptions to this, as with anything... but that's sort of peripheral to what I started writing about.

The Music Genome Project has no way of knowing that when I'm in a semi-nostalgic, somewhat bitter, contemplative and cautiously optimistic mood I want to hear old Dashboard Confessional, new Muse, some Justin Timerblake, some Whitesnake and only half of Introducing the Hardline According to Terrance Trent D'Arby but all of O.K. Computer. So I will tell Pandora just that, and use the thumb buttons to keep things in line.

Gee, that was simple.



*Interesting side note--my husband & I share an account. Of the 23 stations we've created, only two of those are my husband's brain child, plus a joint station.

**Someday I might learn how to make my sentences feel like my speech while employing proper grammatical form, but I doubt it. I think and speak in run-on sentences. If I cared more, I might fix it. I don't.

1 comment:

Bud said...

I wish Pandora would get their act together and enable service outside the US...
And somehow using proxies has never worked out the way I want it to.


I was unaware that I was set up on a blind date with a guy in '92. Then again I was unaware of most things back then.

Related Posts Plugin for WordPress, Blogger...