Duke Listens!: Visit my main blog at MusicMachinery.com

The social news site digg has been having some growing pains lately.  Since Digg is so popular, getting a story on the front page of Digg is a big deal - such a big deal in fact, that there is an incentive to game the system - to influence digg through organized voting to get a story dugg to the front page or buried where it will never be seen again.  Digg has been fighting back, tweaking their promotion algorithm, looking for digging patterns that will point to organized digging and banning the shills.  But everytime Digg updates their algorithm they end up pissing off some of their users - many of them long time contributors.  Recently, a group of hardcore diggers have put together an open letter to Kevin Rose - to object to the latest set of algorithm tweaks.   Digg isn't the first social site that has had these problems. Google has been fighting against search engine optimizers for years - whole industries have surfaced to help websites get good placement in Google.

 It won't be long before the same sort of thing starts to happen in the music discovery world as well.  Nearly all of the music discovery sites (for example last.fm, Qloud, iLike) rely on social data and user behavior to track and recommend music.  Just like Digg, these sites are easily gamed.  It won't be long before there's an economic incentive in gaming these music sites - using them to promote or bury songs.  And when that happens, people will get bad music recommendations.

 So what should you do if you are building a social recommender for music and want to protect your system from gaming?  One place to start is by reading the papers by Bamshad Mobasher on topic.  Bamshad spends a lot of time thinking about how social systems can be attacked and how to construct defenses against such attacks.  Of course, if your recommender doesn't rely on social systems but also uses other info (such as content) it is less vulernable too.