Complexity and Social Networks Blog

« March 12, 2007 | Main | March 14, 2007 »

13 March 2007

Small worlds-- the degrees of separation between Cambridge, MA, and Fargo, ND

I recently conducted a small world experiment in my networks class, selecting an individual in Fargo, North Dakota as the target recipient. I selected Fargo for much the same reason that Milgram selected a small town in Nebraska: psychologically and sociologically it is about as far away as from Cambridge, Mass, as is conceivable and still be in the US. My students were instructed to send an e-mail (with explanation) to someone they knew, who was supposed to send the e-mail to someone they knew, and so on, until it reached the target.

The results: for 25 students (and about 40 attempts) 4 chains were completed, plus a fifth chain completed from a split from one of the completed chains (i.e., one of the students hit the target twice with a single e-mail because one of the recipients down chain sent the e-mail to multiple others). The number of hops varied from 3 (a student in the class who happens to be from North Dakota), to 8), with an average of 5.25.

A number of interesting observations (some consistent with existing lit on small worlds—e.g., see Stanley Milgram, and Duncan Watts on small worlds, and John Kleinberg on navigation in small worlds):

1) These results confirm the basic intuition that we live in a small world (i.e., where a small number of degrees of separation is typical). Even for the majority of students who did not have completed chains, presumably there are a maximum of only 4 hops away from this individual in Fargo, since they have one classmate who is just three jumps away.

2) The navigation of the completed e-mails through the network was strikingly efficient, reflecting the crude but effective cognitive representations of participants of the macro- societal network. People used a variety of heuristics for choosing who to send the e-mail to: do I know someone from North Dakota? Do I know someone who is well connected and likely to know someone from the Midwest? Etc. The resulting paths were likely not the optimum paths, but couldn't have been far off. Even assuming that the world is “small” it is a remarkable (if understandable) thing that these e-mails could find a reasonably short path through the network. If only Boston roads were this navigable….

3) Information about the target improves the navigation through the network. To illustrate this, I varied the information that students were given—some were just told name and city of the target, and others were told name, city, and profession. Three of the four completed chains were for the second condition. Further, in the fourth case, someone along the way looked the guy up and incorporated information about where he worked into the e-mail. Nothing statistically significant, but notable.

4) The credibility of the message was essential in pushing it through the network. I suspect that if I had done this experiment 5 years ago, chain completions would have been higher. The vast majority of e-mails that people get now are junk. Further, everyone has received hoax e-mails forwarded on by acquaintances. One guesses that there was a concern not just that it was a hoax, but that it would be embarrassing to forward a hoax e-mail to someone else.

5) People relied on strong ties in sending the message. Interestingly, given the literature on the role of weak ties in disseminating information, when I polled students on whether they sent e-mails to close friends as compared to acquaintances, 90% of the students indicated that they sent the e-mail to close friends. This follows directly from point 4: their concern was to send it to someone who, in turn would forward it on.

6) Friends are helpful, but friends of friends far less so. Participants in chains were instructed to cc the originator, thus we have data on incomplete chains. Interestingly, almost everyone reported that the first person they sent the e-mail to forwarded it on (following from point 4); but there was a big (~50%) drop off at the next jump. My intuition is that this reflects more broadly on the epidemiology of information.

7) The act of “using” the network affects the structure of the network. In reading the chains of e-mails, it was striking to me how people used this e-mail to reconnect with someone—e.g., “I wouldn’t normally send this on, but it seemed like a good excuse to see how you were doing.”

Posted by David Lazer at 8:04 PM

Enterprise Social Networking Software

IBM has announced to launch its Lotus Connections software in the first half of 2007 and Cisco buys the technology assets of tribe.net. It seems as if social networking software has become an important business line within large software vendors.

From a researcher's perspective it makes sense for firms to connect their employees through social networking software. Finding information, locating experts and spotting project relevant knowledge effectively are promises social software seems to able to hold. If not, why would people be interested in paying annual membership fees on platforms such as xing.com or linkedin.com.

At the same time, software vendors haven't got much to offer than whitepapers, prototypes, or other studies. A persistent question software vendors might be struggling with thus is: What is the USP of online social networking software why is it worth a client's effort to go through a massive data migration effort from several expert or knowledge management databases to a consistent social networking platform?

Here are some arguments/talking points that might help:
- Validation through Existing Models: The success of Xing.com and LinkedIn.com as two prominent examples of popular professional social networking platforms shows that managers and practitioners are willing to spend time and money in locating contacts, knowledge, and information within social networks
- Tie Characteristics and Performance: Studies in the management literature have shown that the characteristics of ties among managers and employees can have strong effects on the firm's or a managers performance (Hansen 1999, Moran 2005, Obstfeld 2005, references see below)
- Privacy Concerns: People are willing to publish their profiles online (as it can be observed on prominent Web 2.0//online social networking sites and as described by Ines Mergel in her prior blog). Hence, people are used to publishing their profiles online, have experience with it and might ranke the expected benefits higher than potential data privacy concerns.

Hence, why should making ties among people within firms visible NOT help these people to become more effective or productive over time?

Related Literature:
Hansen, M. T. 1999. The Search-Transfer Problem: The Role of Weak Ties in Sharing Knowledge across Organization Subunits. Administrative Science Quarterly, 44(1): pp. 82-111.
Moran, P. 2005. Structural vs. Relational Embeddedness: Social Capital and Managerial Performance. Strategic Management Journal, 26(12): 1129-1151.
Obstfeld, D. 2005. Social Networks, the Tertius Iungens Orientation, and Involvement in Innovation. Administrative Science Quarterly, 50: 100–130.

Posted by Thomas Langenberg at 8:09 AM