Friday, October 19, 2007

Predictive accuracy is not enough

Sun Labs researcher Paul Lamere posts a list of properties of a good music recommender system:
A good recommendation has three aspects:

familiarity - to help us gain trust in the recommender
novelty - without new music, the recommendation is pointless
relevance - the recommended music has to match my taste
The emphasis on credibility is insightful. Recommendations that are too obscure may be perceived as low quality because the user cannot easily and quickly evaluate them.

This is just one of many examples of the yawning gap between predictive accuracy -- accurately predicting what people want to buy next -- and perceived quality -- the usefulness of the recommendations to users.

2 comments:

Anonymous said...

Thank you Greg, thank you Paul! This post just crystalized a thought that I've been chasing for about 8 years now, and it is not unrelated to the discussion Paul and I were having a few weeks ago right before ISMIR. Give me a week or two to write it up, but I would be interested in hearing y'alls reactions once I do.

Anonymous said...

This is why I'm surprised we don't see more services that let people push music to their friends - like Last.fm's "Tell a friend about this artist" - or open source players that let people push out recommendations using a standard playlist format. Paul and Lucas Gonze talk about this here:

http://blog.gonze.com/2007/05/16/portable-song-ids-and-music-influencer-networks/