The problems of visualising social networks…

From a pithy and somehow true post by Stewart Butterfield on the problems of creating visualisations of social networks:

Artist/curator friend Mark Soo did a piece for one of the Infest openings where he visualized the curators’ social network using balloons with people’s names printed on them as the nodes and ribbons tying them together as the edges (the data comes from “invites” he got the curators to send to one another). This was a great, inviting, tactile “graph manipulaton interface”. But the reason I liked it so much was that it really brought out the problems of social networks visualizations as a way of learning about the networks being visualized: too confusing!

He also cites a few examples of some of the attempts to visualise them – the problems should become self-evident:

Two things immediately occur to me – firstly how do we as humans make sense of this data in our everyday lives (because we’re incorporating at least some of it into our mental models, surely, and understanding that would make it easier for us to enhance those models rather than creating new ones that create nothing but cognitive overload), and secondly What would Tufte do?.

10 replies on “The problems of visualising social networks…”

Visualizing Social Networks
Tom Coates asks "what would Tufte do?" when it comes to the problems associated with visualizing social networks. Tom rightly points out that we’re constantly using our social networks in our daily lives, but when it comes to visualizing thei…

I think the way you display the data is linked to how you want to use the information. I assume browsing the data to understand relationships is important, and tagging the paths you like for retrieval would be a good thing.
The data -as it’s represented- is flat, looking at the structures in 3D might help. Imagine a real-time browser of the dataset, where you can zoom in and leave breadcrumbs of your journey, highlighting people and places. Paths leading from a person (or people) could be reduced to minimal or complex branches, like a phylogenetic relationship -in these cases you can mute (grey out) extraneous data and create a 2D representation of demonstrable paths.

The obstacle to creating visual representations of social or other complex networks is that nodes occur redundantly. You cannot visually represent Maryís relationship to her pals Sue, Betty and Louise in the same visual network as Maryís spot in the piece of the network comprised by her and her siblings because Mary appears in both sub-networks and none of the other individuals do. It is at best very messy to depict even two sets of relationships simultaneously. In the overall web of the people in her life Mary is redundant over and over, if you will, showing up in many configurations of different individuals. Tufte and the rest of us will, I think, never be able to reflect this visually. The wonderful thing is that we have the amazing new expressive powers of virtual representation in digital networks in which we can capture and manipulate the redundancies, calling up the relationships at will of Mary and her friends or Mary and her siblings. Because this new medium is dynamic and interactive, we can visualize what we need at the moment. Doing so is a big step toward mirroring thinking.

Off the top of my head I can think of reasons why these maps are wrong: they are the partial external link to people, without explaining what kind of relationships are there, and they are not a map or model even, but a byproduct. We do develop mental models of relationships, but these are based on concepts such as attraction, ideology, affinity, coincidence etc. These maps lack that subtlety and resolution, and simply jumble people together: You would need different resolutions for different kinds of relationships, redundancy assigning different weights, and definitely more than 2D to indicate the closeness and weight of appreciation for any person.
And just like a mental model of a city is not related to the geographic city itself, but to your interactions with it, our mental social models are based on how we interact with people and the stimuli we exchange, the messages that comprise our joint activity.

Just rambling here. If this is to be done in two dimensions, I propose any mapping be comprised of fading ink, that will indicate the connectedness of individuals. Less visible is a great metaphor for not well-connectedness. Someone not well-connected hardly deserves an equal spot in the overall network. As they become connected, they come up on the radar. Size of the individual symbol or text could symbolize how integral the individual is to the network.

Comments are closed.