Slashdot Mirror
The Social Web
person writes: "The Stanford Social Web calculates social interactions of members of the Stanford University network, using links and text from home pages, as well as information about mailing list subscriptions. Though the site's analysis of user relationships and similarities is limited to those with Stanford accounts, it is of interest to those studying the formation of social networks. The java applet is especially nifty." Don't even try the java applet if you don't have a fast machine, PIII or higher.
Sounds just like real life, right? One thing about measuring this via weblinks and such is that you are going to get different results than real life social connections. We all know that some of us are great socializers online but not so good at it in person, we all know of some people that might be great in real life but haven't a clue how they would manifest that online. I think the intersting paper is where these two 'sets' overlap and don't overlap. Considering the study is on individuals at a college campus with geographic proximity, the results would differ from people in general.
I only skimmed the paper and didn't see too much, however, an interesting chart was the difference between Stanford and MIT, MIT had twice the percentage of people that linked to someone else at school and nearly three times the percentage of both 'who are linked to by at least one other person' and 'with links in both directions'. So MIT has more online culture? Again, I'd like to see the overlap with real life social interaction.
One last thought, isn't this a social engineer's dream? Use this data to exploit connections between people? Impersonate a friend of a friend? It could backfire, what if a friend of a friend measured via links and mailing list participation are actually close friends in real life...
Chris Cothrun
Curator of Chaos
Bleh!
It would be much more informative to follow the social network by checking which users correspond by email, but the privacy concerns are obvious, even if nobody looks at the actual contents.
There must be a lot of research about social networks through email done at the NSA. I bet Carnivore, for example, is very interested in who is emailing who, possibly more than the actual contents for most of the time.
-
Stop worrying about the risks of nuclear power and start worrying about the risks of not using nuclear power.
> I'm not advocating being a luddite or anything,
> but more and more often these days we see
> online social interaction not adding to but
> replacing more traditional forms of human
> contact.
Funny, I know exactly what you mean. I remember when the written word used to be one of the traditional and main forms of social contact.
Heck, it used to be an art form. Then came along that newfangled, dehumanizing telephone. All asudden everybody could just ring someone up and start yakking away. It took all the thoughtfulness out of communicating, being able to just bug someone, none of the richness of a good thought out letter. It didn't even come close to the well-developed speeches one could work out on the walk to the town.
I wonder if written communication will ever come back into popularity:}
What about IM use? How did they or could they measure any use of instant messages between users on the network, would they measure the frequency of contact between different locations? it would be interesting to a see a map of all IM connections over a period of a week or longer, showing how they go to multiple connections.. a time lapse map would be perfect.... just a thought
Being a Stanford student, I was currious how well it did on me and my friends. The output is pretty much dead on. It picks out my roomate as my as the person closest to me, and most of my friends as well. Not particularly suprising, given that we are all on similar listserv lists (and all residence in each dorm are on a dorm list). Nevertheless, pretty striking. I'd be currious how it did in a less highly connected (in terms of listserv and homepage connectivity) community.
There have been studies of telephone calls within companies, and it turns out to be possible to derive the organization chart from call data. (A key criterion: how fast does someone get called back?)
Graphviz makes reasonably well sorted directed graphs.
http://www.graphviz.org for dev versions.
http://www.research.att.com/sw/tools/graphviz/ for release versions.
Optimal layouts with mimimum crossing edges for arbitrary directed graphs is a Very Hard Problem Indeed, so don't expect miracles.
I often wonder about the long-term effects of putting so much effort into creating - and now analysing - digital "culture". I'm not advocating being a luddite or anything, but more and more often these days we see online social interaction not adding to but replacing more traditional forms of human contact.
Usually you replace something that is broken or flawed. So what's so flawed about normal socializing that we feel this need to supplant it with something that is in reality more isolating?
Just some idle thought on my part. Feedback?
An applet? Not Flash? Am I on the right web?
-... ---
First, the authors are not studying friendship networks as the term is usually deployed in the network literature. They are studying citations on homepages, which is not the same thing. While they may be justified in arguing that these citations are in some way a proxy for the actual friendship network, the paper does not present any data to support this claim (or, more importantly, to assess the degree of measurement error inherent in the use of citations as proxy data). In showing, therefore, that there are associations between web page content or mutual citations and direct citations, the authors are merely using web pages to predict web pages. This would be expected to inflate the strength of the observed relationship (since there may be mechanisms encouraging (for instance) transitivity of web page citations which are not present between web page citations and friendship), and would lead me to question the validity of the research findings. At the very least, I find the title and abstract misleading and believe that the authors should be much more upfront about what it is that they are actually studying.
As a second concern, I was a bit irked by the fact that the authors continued to perpetuate the myth that degree (number of ties) for actors in most social networks is power law distributed. While it is true that some social networks exhibit this property, it is most emphatically not the case that all do. For instance, degree distributions resulting from the GSS network module are not even vaguely power-law distributed, and neither are most of the standard data sets distributed with UCINET (the bulk of them seem to be approximately normal based on my tests, which is exactly what you'd expect from a sum of random variables). The fact that web pages seem to exhibit power-law distributions in degree suggests that their networks are actually quite different from face to face networks, a finding which really shouldn't be a surprise to anyone. Alas, since citations are being used here as a proxy for friendship relations, this does not exactly inspire confidence in the data.
All in all, this is an interesting paper; unfortunately, I think it suffers greatly from the "siren song of the web." As the authors note, it is very easy to collect large quantities of data from web pages...alas, methinks it might be a bit too easy. I am aware of no extant data to support the proposition that hypertext citation networks taken from the W3 can be used as a reasonable proxy for more general interpersonal relations, but I fear that the lure of trivial data collection may seduce many researchers to look the other way when opportunities like this arise. Maybe it will turn out that web-based citation networks can be used in this way, but I'm not holding my breath. In the meantime, I'll stick to my fogeyesque position that real progress vis a vis the study of interpersonal relations depends on more traditional observation and elicitation schemes.
-Carter
While the main page loaded, I can't actually get through to any of the 1,200 individual webpages which are part of the Social Web.
Brackets contain world's first nanosig, highly magnified:[.]
--MarkusQ