Wednesday, September 05, 2007

Creating Online Indenty TrustRank

There has been much discussion recently about making portable/decentralizing the Social Graph/Social Grid, otherwise known as the collection of personal relationships, "friends", email addresses, photos, and other shared data which exist in your address books, (Gmail, Outlook, etc.), in your social networks (Facebook, LinkedIn, etc.), and in personalized content sites (YouTube, Flickr, etc).

There has also been discussion about social vs algorithmic search engines such as this one that focuses on a debate about the Jason Calcanis/Matt Coffin search engine Mahalo vs. Google.

However, as someone who has a strong financial services industry perspective, the
issue that interests me the most is the development an algorithmic way to assign trust and reputation to an online identity.

The central question is this: How do we create the ability to attach a trust factor to an online contributor (identity) so we can appropriatly assign weight to his or her perspective. We all know the problem that exists currently, without an adequate trust model:

1. Untrustworthy users create multiple or malicious accounts with which to "game" or "spam" the system. We can see this on Digg, Wikipedia, MySpace, Ebay almost any network which empowers an individual to control or influence published results.
2. Management struggles to identify and ban the bad actors. The remedy is only temporary, as the offenders are free to re-emerge under a new identity.

There are several models of trust already in existence, all of which are too simple to handle the complex requirements of identity and trust management, given the incentive which exists to abuse such. Some existing techniques include approaches such as:

1. Use of Captchas
2. Use of email address validation
3. Use of SSN Validation
4. Use of physical address validation (ie. Google Local)
5. Public display of account history (Ebay, Forums, Wikipedia)
6. Tracking and display of IP address (Wikipedia)
7. Friending schemes (MySpace, FaceBook, LinkedIn)
8. User empowered TOS monitoring (Wikipedia, YouTube)

Many social sites apply a combination of several of these techniques. Indeed a multi-layered approach is certainly necessary. The idea I'd like to explore is an portable algorithm for online identity reputation and trust.

Social Networks employing named-relationships (friends) have developed to a large enough mass to give us some real validation as to a user's singular identity. Using multiple social networks (Facebook, LinkedIn, Plaxo) it is possible to further validate identity and via the implied trust value of double-verified friends. Users who have to this point resisted social networking sites, still have an outlook or gmail address book full of relationships that could quickly be validated as a new entrant. More elements can be identified and the algorithm can be improved.

A social-algorithmic approach, combined with the techniques listed above (themselves alone not enough), could drastically improve the quality of the social internet, by forming a portable basis for reputation and trust.

No comments: