Tuesday, September 28, 2004

Social Networks in the Workplace

In my current project, I keep on coming back to social networks as a good place to do research. Of course, the power of social networks is completely part of today's zeitgeist, so in that way, I'm not being too original.

But my current set of questions revolves around social networks in the workplace. Are they useful? And if they are useful (it's not clear to me that social network analysis has power in the workplace beyond 'gee whiz' type applications), how should networks be measured and then used and/or maintained? I've been playing around with an interesting SN dataset here at work, but it's unclear that it's the 'right' data set. (I've also come up with about three possible SN data sets here, depending on how much time and effort I want to expend, and how deeply into people's lives I want to probe)

SN seems to have two remarkable weaknesses right now: multiple relationships and temporal data. All of the social network data I have identified has temporal aspects to it, and it's not clear how to handle time. (Of course, I don't think people have a good sense about how contacts decay over time which makes this analysis harder. I know there are some friends I can drift out of contact with for months at a time, and then re-build the connection later.)

The multiple relationships are also tricky; most analysis works with a link and a link-strength, not multiple links. Of course, one can create weightings of different relationships to come up with a real-valued weight, but I think that misses some of the richness and complexity of what real social networks are about. "Tools" like friendster and Orkut have only links and no weights (although I believe Orkut now has weights). There is also this related question that some of these resources are automatically gathered from data sources. Are the relationships constructed from these interactions "real"? Are they important? I am hedging here a bit, because I don't want to talk about the particular data sources that I am using here; some of them are considered proprietary data. But there is an argument that the relationships they embody aren't genuine social relationships. (They are genuine *business* relationships; that is perfectly clear from the data, but what of it?)

Not that anyone reads this (yet), but I'd be interested in thoughts.


Post a Comment

<< Home