Looking deeper at romantic networks
Of all the examples of networks I have come across in our Textbook, the most striking was the illustration of six months worth of romantic ties between high school students. I have copied the network below:
This network was taken from a study at UPenn, the link of which can be found at the bottom of this post. What makes this network stand out is not the large size of the large component, nor the large number (63) of perfectly monogamous relationships. What drew me back to the image was the extra piece of information it provided: Gender.
We can study this network using the tools we have learned in class, and determine the stability of each component, or to come up with theories for whom will join in a relationship with whom next, but the addition of gender adds an entirely new dimension to the network, as none of the tools at our disposal apply to the metadata of a Node. With respect to the tools at our disposal, the notation of gender is little more than an interesting fact. However, clearly the original study was doing something with this metadata field, so I searched it out, and was pretty amazed at what they were able to do with not only gender, but a whole series of other metadata fields. They compared the GPA of each student across relationship bonds, and were able to find striking correlations. Again, with popularity, and again, with family wealth, correlations could be found across relationship bonds.
The researchers took these correlations, relationships, and metadata fields for each Node, and used it as a way to uncover the “governing [social] norms” of the high school. In some cases, these social norms overruled rules we learned about in class like Triadic closure, or the correlations in metadata fields the researchers expected to see. The best example of this was the fact that in the large component, though triadic closure would suggest that if Node A and Node B were dating, and Node C and Node D were dating, if Node A and Node D each broke up with B and C respectively and began dating each other, Nodes B and C should in many cases begin dating. However, in the system shown in Fig 2, this almost never occurred. From this fact, the researchers concluded that there was a social stigma between dating an ex’s ex.
As I start to build my own networks, I will be sure to note the metadata. If this study has taught me anything, it is that sometimes that metadata can be almost as valuable as the network itself.
http://www.cis.upenn.edu/~mkearns/teaching/NetworkedLife/teensex.pdf