Collaborative Micro-Filtering

Joshua Porter wrote a great post on circles of relationship a few days ago. It has generated many great responses and a new post that summarize all these contributions. Since it's a subject we have been working on with my partner, i would like to share a very concrete approach of it : U.[lik]

First let's have a look at the circles evolution from Ben Schneiderman (via Joshua) to People like us from Alex Mather:

The Schneiderman representation is clearly a pure social approach.

Alex Mather one is a pure Social Network approach mainly focus on the 18-30 years old group. According to him there are two main points that organize the circles:

1. Self, Friends, and “Those Like Us” are in a protected area.
2. Friends and even “Those Like Us” are more important than Family!!!

--> I agree at 100% !

The next step is from Sarah Cooper (Visual designer at yahoo) who introduces the idea of Collaborative Micro-Filtering. She has designed a quick demo of what it should look like. She also tries to build the specs for such a service.

The whole idea is to be able to get the right recommendation by using the right people for each different subject.
Her mom & danna for clothing & Joel, jim and rashid for movies. --> Again I agree @ 100%.

That's exactly these kinds of ideas that we have been working on over the last 2 years.

First of all, U.[lik] is a social cataloging website dedicated to entertainment i.e you can express you tastes with everything you like and dislike, the sum of it constituing your library.
The picture on the left (click to enlarge) is my lounge where my friends can get a taste of everything I like (they can also watch video based on my personal Hall of Fame).

In this lounge you can find a series of tags that describe my library. If you click on any of them (in the example below Rap & Writers) all the content of my library will be reorganized as well as my friends.

As you can see on the picture, I have a new set of friends for each subject.
Besides being an effective filter for the content you care about, using the same methodology will enable you to discover People Like You, either like the whole of yourself (the broadest part) or just a part of it depending on your mood (just someone Like Me who loves rap music). We are leevraging this micro-matching to leverage your access to content, in this case I can access the rock part of Dubus' library.

So now let's have a look at Sarah's specs :

1. Identify people I know
2. Allow me to judge my similarity to each person, for a variety of topics
3. Make recommendations based on first and second order relationships between me and the people I know
4. Continually refine this similarity index for each person I know, based on our rating patterns over time

I think the second point of sarah's process is too heavy.
--> You need the machine to do the bulk of the work and then have the human compute the last mile. In our case you can rate users (which we'll use to set the different circles).

We are still working on it to replicate real life but it's already delivering a very good sharing experience (users say so) and I hope that you will (as I do) make great discoveries thanks to U.[lik].

PS: Sarah, we will use the affinity to organize comments ;-)

Comments

[Rédigé par: Sarah Cooper]

Good discussion - I agree with your conclusion and I'm looking forward to playing around with U.[lik] !

[Rédigé par: fred]

when is this going to be available?