Posts Tagged ‘last.fm’

Recommendations: Content vs Friends

Friday, August 1st, 2008

Digg Recommendation Engine from Kevin Rose on Vimeo.

The geek world (aka bloggers) have been abuzz about the Digg recommendation engine and with good reason, recommendations are a crucial part of any service asking you to digest ridiculously insane amounts of information in a little time period. This same concept holds true for services like Last.fm or Google Reader… the services know they are more useful if you are presented with content you want/enjoy and can cut through the mountains of crap.

This being said, it seems that technologies are opting soley for content based recommendations over friends/network based recommendations and this makes me sad.

Yes… I understand that for content based recommendations… I like X, Y and Z site, you like W, X and Y… let me check out W … but what about my friends? I want to know what common things people who call me a friend are interested in… even if it may not be the most interesting thing to me.

Friend based recommendations work on two levels… first it gives you something to talk about at parties… all my friends are interested in a new band, maybe not my sort of music, but as a friend I want to stay connected with what they are talking about. Infact the fact that my friend likes a band, even if I don’t, may be of interest to me… I am constantly on the lookout for links for my school teacher sister even though I have NOTHING to do with the education system. Secondly as an author/blogger/marketter you want to know where your community is headed, especially in the days of Twitter and Facebook where many have amassed hundreads, if not thousands of followers, as a author you need to know what your public wants. This is marketing 101.

You built your network of friends based on something you have in common… some magical unquantifiable quality that as students of marketing know, word of mouth is worth more then gold, and yet software is being built with a complete disregard to these connections, but instead trying to “figure me out” based on simple yes and no answers to items… do you like this? Did you digg this? Did you listen to this? All unweighted answers (Did you like this? Did you REALLY like this? DId you LOVE this?) to try and provide for you content that you will find interesting. Seems like a real waste of a insanely valueable resource.

I am not saying that content based recommendations are bad, on the contrary, I actually use them to discover new friends. Content based recommendations are totally valueable in the case where you are looking for a topic that maybe your friends have no interest in or perhaps you have no network from the software’s perspective. I definately don’t want them to disappear… I just think software developers need to take a step back from the algorithms, the 1s and the 0s and take a look at how humans interact and ADD this functionality in (and even label them so…. “Recommendations Based on Habbits” and “Recommendations Based on Word of Mouth”) finally helping the machine to become more human.

About Lynn Wallenstein

My name is Lynn Wallenstein and I am one of the co-founders of Powered By Geek, a contracting and consulting firm formed by some friends and I who were sick of working 100 hour weeks while our bosses get richer but that is a whole other story. This is where I ramble about all things design, code, project or whatever based both for PBG and for my collection of personal projects.

Contact Me
My Portfolio
Buy Me a Coffee