shikora.org

So… what does it really take to create a trust network? I’m thinking not a lot, at least not a lot to get useful information. When we say we trust someone, what information are we really conveying? Really only 4 things. 1) I 2) trust this person 3) about this 4) this much.

An example of this would be <@manimal, @donpdonp, #rubyonrails, 9> or <@manimal, @strangelovelive, #podcast, 10>. Here I’m using twitter handles for unique identifiers and the hashtag format for subject taging. Is this perfect? Of course not. Will this data need updating? Absolutely. Does having even some of this information become useful to us? I think so.

As a simple use case, what if we wanted to be able to sort our email, not by some algorithmic means like a Bayesian metric, but by out own trust relationship with people. We could give all the people we receive email from a spam rating from 1-10 and our email would be sorted from most to least probable spamishness.

One more piece of information would make this a much more powerful tool. If we added how much we trust that person to trust others about the same subject, we can make inferences about people we might know nothing about personally. For example <@manimal, @donpdonp, #rubyonrails, 9, 7>. This says that I trust Don quite a bit to know Ruby on Rails, but less to know about trusting others about what they know about Rails. In this way we can share trust networks and gain information about others.

I talked a little bit about this at both the Bar Camp PDX 09 conference and on the Strange Love Live show at the WebVisions 09 conference. The video of the last can be found here.

Like twitter, I think the power here is in keeping the information simple, and allowing the clients to add the complex usage patterns. It so happens there is a format for sharing social network information in tuples called FOAF data. One way of accomplishing this would be to just add a tag and value tuple to standard FOAF data. Thankfully the RDF allows redefinitions pretty easily. One link that descriptes RDF and FOAF data is here.

I’ll be posting more about this as I have time, but I have lots of ideas for clients applications and sites to allow adding of data. Ideally this would be part of the DISO project, but again, we’ll see how much time I have.