I saw this NYTimes article and thought it might be of interest. Netflix (a DVD rental company) offered a $1 million prize to any person or team that would create a program to help subscribers pick movies to watch that they might like and do it at least "10% better" than their own Cinematch software. They ended up with two teams who succeeded. But they were given a database of 100 million movie ratings to calibrate with (!). This has led to a second contest where more data, including gender and demographic information, will be provided.
I don't know how Cinematch works but if it takes that much data to build models to do just 10% better, I think we're pretty much out of luck getting something similar for helping us pick books unless Amazon does it. No one else will have that much data available.
http://bits.blogs.nytimes.com/2009/0...ew-contest/?hp