A Star Is Biased

In another example of being late to the game, I started playing around with iTunes over the summer, and finally started rating some songs to improve my use of iTunes Party Shuffle. Just in time, Slashdot points to this neat statistical analysis of using rated Party Shuffle versus random Party Shuffle. Besides doing a nice experiment to measure the bias that rated shuffle has towards each star level, there’s also a nice mathematical justification for why you may hear fewer 5-stars than 3-stars and even the occasional repeated song.
I was surprised that unrated songs, on rated shuffle, only come up about 4% of the time versus 1-star songs, which come up about 12% of the time. That is, if you don’t like something, you’re better off not rating it at all than giving it a low rating. I’m not sure that’s entirely intuitive – I had been thinking of three stars as neutral towards a song, and it seems that iTunes assumes that if you rate a song, you at least like it a little. So, iTunes ratings stars do not translate to Amazon ratings stars.

Leave a Reply

Your email address will not be published. Required fields are marked *