Wednesday, November 07, 2007

Testing widgets

Web 2.0 provides many interesting ways to study taste in music and patterns and habits of listening to music. Dividing music into genres is an activity that music makers, vendors and researchers are interested in doing, although their motivations might be very different. Early musicologists were obsessed about musical style and finding objective ways to categorise works and composers. The activity goes on, and now we also do that socially. Most online music shops use metadata, or names of artists, songs, years of release, and genres so that customers can search for music they might like to buy. The genre allocations in iTunes for instance are sometimes quite ridiculous. Rufus Wainwright is rock while Maija Vilkkumaa is pop. A genre of a song turns into soundtrack once it has been released on one.

Since having meaningful searches from the audio is so amazingly difficult, describing music, tagging it, is the best way to label the stuff so that it can be found later. One way to do that is to hire a bunch of music students to do that for you, as they did in Pandora. The service is based on the idea of musical genome, characterisations of the songs and pieces of music that can then be used to find similar songs from a database. Very interesting, but listening to these web radios gets boring very soon, as you quickly find out that there really is much more of the same, no matter where you start from. has another take on this. Similarity is based on the listeners. Just like Amazon's recommendations, uses the data that members provide of their listening to make recommendations for you. Those who listened to this song, also listened to that one. And again, you can listen to a web radio that compiles its playlists based on these algorithms.

From music research point of view these both are interesting services. And especially the social aspect of musical taste that has created. One thing are the genres themselves. You can tag the music you like and listen to in any way you want. And others see those tags. These tags are now being researched. See blog entry on the topic.

I'm now getting interested in the various ways in which we could obtain, process and analyse the eams of data people leave behind in when they use it. Just looking at my own iTunes listening behaviour (more than 11 000 songs played) after I had my iTunes play history scrobbled to was interesting. And perhaps mostly so because I think it is different from how I would reply to a questionnaire about my music listening. From methodological point of view, this is important. Collecting real world data about how people "use" music interests researchers, but observing has so far been tedious, required people to keep diaries of their listening habits. This is intrusive and makes people very aware of what they are doing, and that they are now being monitored, and this skews the result. With, we can obtain their real data, as the data is collected all the time, before they even know they are going to be asked to take part on a study.

And from social psychological point of view, the opportunities that a community based on what music people listen to gives, are immense. People are being linked based on their musical tastes. In the "real world", the causation often goes to the other direction. How does it influence your music listening that you know your choices are being scrobbled and made available to the world? That when you look at anyone else's profile you see you "musical fit" with that person? That your friends and "neighbours" are being constantly ranked based on how much music you share during a given week?

A lot to do there, I'd say. I'll test some of the widgets that provides here. If you are not satisfied with the audience gives your musical identity, you can use these kinds of widgets to broadcast your habits even wider, by adding these to your Facebook or MySpace profiles or blogs.

This should broadcast what I've been listening to lately:

No comments: