[click to see full size] Andrew writes:
This graph represents a network of science fiction authors and philosophers, with the authors linked to philosophers just in case the philosopher listed that author as philosophically interesting. Authors are labeled, and label size corresponds to the number of philosophers mentioning them. Label colors and positions are rough indicators of similarity. Colors represent groups of authors; as an intuitive gloss, if authors A1-An are the same color that means the connections between the As is ≥ their connections to authors in other groups. Author positions are determined by a combination of three forces - gravity, attraction, and repulsion - applied to the network until it has settled into a stable position (a local peak in the space of possible positions). All nodes gravitate to the center and repulse one another, and nodes are attracted just in case they are connected. So, positions and colors can be seen as weak indicators of similarity, whatever kind of similarity is highlighted by philosophers' choices.
But, given the relatively small sample size and lack of strong modularity in the network, we should be cautious in inferring anything about these authors (or philosophers) based on their relative positions or colors.
I'm curious about what program was used to create this. I've been using R for the past year in law school to do predictive analytics and data visualization, but I have yet to do a network analysis despite its applications in law. Would he be willing to post his code to GitHub to recreate this?
ReplyDeleteAww, yeah: the lower-left amber ones are awesome!
ReplyDeleteOf course they are, Pete, cause the recommender is awesome! Ryan, Andrew probably wouldn't mind an email about this: Google his name and "philosophy" and you'll find him.
ReplyDeleteRyan, I used Gephi for this one, and it's my favorite program for producing slick visualizations, but I would recommend NodeXL for folks who haven't done network analysis before. NodeXL is an inferior program for almost every purpose, but I found that it was a good starting point because of the familiar Excel interface.
ReplyDeleteThis is a fun info graphic map and this is a link that may explain Higgins comment some more. Really these programs display the data well. Love ur numeric epiphenomemalist content, Eric.
ReplyDeletehttp://www.coppelia.io/2012/06/graphing-the-history-of-philosophy/