Hybridity, Purity, and Space as Roadblocks for Social Network Analysis
For the past few years, I have been attempting to use Social Network Analysis to map out the network that is created during an activity like a board game. These scenarios are pretty straight forward and simplistic to view. We can record and map out the various associations things make during a period of time. When we replace that game with an iPad what buttons, what screens, what actions, and how well a user employs a piece of technology can be mapped out as well. This is a mainstay of user studies. The complex series of associations use creates is often reduced to a few metrics. This is what prompted me to try and unpack that.
I have been frustrated for these past few years. What I seek to do (map out a network of human and non-human objects) seems straight forward. In use, the association between a computational object and a human creates an association. The result of that association and the context it takes place in are of paramount importance. For example, we know that users will rarely progress past mediocrity with any product, regardless of the “goodness” of its design. However, we don’t really know why that is. We simply witnessed it and design has almost universally said, “Good enough is good enough.”
Yet, we still study design. We still try really hard to design things well.
So to understand this issue, to re-consider “Good Enough,” I have sought to use Social Network Analysis (SNA) to create a map reflecting the network that is created during use. The only issue with this is that SNA favors humans OR things humans use (roads, for example), never a hybrid. As I have prepared and have been doing a hybrid network map that I’ve termed “Association Mapping,” I have been reading a lot about Social Network Analysis from a number of perspectives.
I thought I’d collect a few of the concepts here as a way to focus my own deployment but also as a way to consider what this method is trying to do and how (perhaps) it could do it better.
1. Purity and hybridity.
“A network is simply a set of relations between objects which could be people, organizations, nations, items found in a Google Search, brain cells, or electrical transformers.” Understanding Social Networks — Kadushin
The first concept here is that of OR. What has frustrated me is that while SNA focuses on the hierarchy of networks, it tends to collect data points uniformly. The issue with this is that by maintaining consistency of a dataset, many issues are masked.
For example, my own work surrounding observing use consists of only 4 humans. Within this use scenario, SNA would traditionally be deployed to understand the human-network over time. Yet, within that social network, all that is able to be seen is that four humans sat in a room and did something together. The map looks something like the picture on the left below.
This is not something worthy of study. SNA cannot consider a pattern, nor can it consider a structure. It is simply a small group of data whose interest may sit within a given structure or within other data being gathered here.
Yet, if we were to include non-human objects, this particular dataset explodes in nodes, edges, and all manner of association. In the example on the right above is an example of a board game using those same users. Human actors and the tools they deploy to gain a degree of agency within a given network is symbolically and literally given form. The hybridity of object, I believe, is often avoided because SNA often must consider the hierarchical nature of the objects of study.
2. Social as human action only.
“An actor’s position in a network determines in part the constraints and opportunities that he or she will encounter.” — Analyzing Social Networks — Borgatti, et al.
The quote above is of interest only because this statement falls in line with a variety of concepts within the criticism of Sociology. The decline of Sociology in recent years has often been attributed toward its inability to account for or provide meaning for the complexity of human existence. To put this another way, within the given nature of Sociology at current, humans are afforded more status than the objects they design and deploy.
However, the nature of these relationships is literally flat in that the simple nature of a stop light will often have as rich a social life as that of a human being. That the associations and possible actions are limited is no different than the limitations of human-to-human interactions. Even in Kadushin’s text, he notes that “we are concerned with social networks, and what passes through these networks — friendship, love, money, power, ideas, and even disease.” These human-centric ideas are of course specific to ourselves as we are the thing we value most. Yet, if the object of inquiry is humanity, we must be able to eschew the idea that we are somehow special beyond reproach in favor of a more complete phenomena. Consider this quote:
Humans differ because they are often themselves equipped with many instruments to gather, compile, represent or even calculate the ‘whole’ in which theyare said to reside — Latour, et al. Whole is smaller than the Parts.
By focusing only on humans and only on one type of data, we are missing a radical opportunity to create a map of complexity within a bounded space.
3. Space of a network.
Social network analysis calculates and displays the relationships that exist among a collection of people, like those on a project, in an organization, or participating in a blog — Hansen and Smith — Ways of Knowing in HCI.
Like the above, SNA is focused on people. Even within the study of technology like Human-Computer Interaction, HCI tends to try and focus on the deployment of technology as a way to influence people. Yet, within this deployment of technology is a simple fact that I have already stated — users tend toward mediocrity. This is not because users are bad at technology or illiterate, it is just not something that matters. For users (and for designers) good enough is great. We humans are really good at using systems as they can be used within the context they’re created in.
Yet, when we deploy SNA to examine the impact of a particular network, that aspect of non-human-human hybridity is lost. We simply measure the way that relationships are or that relationships have changed. Going back to the first note on hybridity, I will point out again that the human-only relationships even with a board game are radically different.
For example, if we take the above data and display it in a different way, we can see an unintuitive representation of a network. In these examples, I have simply aggregated the associations of “different” objects between two different activities. The first is a tabletop game, the second is that same tabletop game mediated by an iPad.
In this case, associations are simply one node making an association (or edge) with another node. The space of the network, where the action is taking place, is important and within SNA, we have the opportunity to expand and re-consider that space yet the tradition of SNA tends to favor purity.
In general, I feel as though the use of hybrid, impure datasets will not only strengthen SNA, but provide ample opportunity to map out the complex social life of human/non-human agency. After all, humans have long been unable to accomplish goals as an actor without recruitment of other objects.
Of course, the above approach is tedious and it’s taken us a great number of years to even be able to deploy graph theory so easily through tools like NodeXL, R, Gephi, or UCINet. The amount of development it would take to truly create hybrid networks that use internal log data of software and hardware is probably a long way off. In addition, the manual coding of associations through video-recordings and transcription is tedious.
Yet, the value of such an approach would be of tremendous value to designers. If a design could view the social life of a product could the design of new software reflect that level of sophistication? To be able to witness the various types of centrality and network structures a particular piece of a piece of software creates would be valuable at understanding exactly what portions of software are important. That importance would then become something akin to whether that importance was as a gatekeeper, a central node on a network, or possibly even a bottleneck.
Would products be less about dragging users across a line of mediocrity and thus become more about designing software as competent social actors?