I’m working on a couple of different projects that use social networking technology to help create value. The idea behind social networking can be as simple as ‘It’s not what you know, it’s who you know. What has changed in the last couple of years is that technology and computing processing power is catching up to the point where we can model those types of relationships, even on very large scales.

One of the projects that I’m working on is the Birds of a Feather project at the Lehigh Enterprise Systems Center, the other is more commercially oriented, working with a network service provider to use social networking information to create more value for their customers.

One of the business values that social networking may be able to deliver is reduced fraud. If we have good modeling of communities within a network, we should be able to detect transactions that don’t fit the typical profile. We all see some fraud management activity from our credit card companies, for example, we may be called because a dollar amount is out of the ordinary, or the location or method seems wrong. Social networking could provide us with a new dimension for this.

Probing around this area, led to some interesting discussions regarding trust and social networks. Mike Catalano, who is an expert in wireless technology at Booz Allen Hamilton is working on a project for the Gates Foundation, and was starting to consider the elasticity of trust in a social network. How far can it be extended? Hank Korth, a faculty member at Lehigh had worked on real time fraud management back when he was at Bell Labs. He thought that trust should be another attribute that we look at and capture when we build social network maps.

If you think about examples, I’m sure that you may have family networks where you have trust in something even though it may have travelled through 4 or 5 people, while there are others, such as business deals that you may only trust as far as “a friend of a friend”.

My thoughts are that in addition to weak links and strong links, and direction, frequency, and duration of communications, we need to think about adding a trust attribute to the equation.

I am wondering whether trust can be extracted passively from a network by analyzing data, or you would need to establish and capture trust separately, but then model using the value.

Please let me know what your thoughts.