Using The Math of Social Network Analysis in Association Mapping
Is it just that it’s called “Social Network Analysis” or is the “social” artificially maintained?
Social network analysis, graph theory, and the various forms of network analysis have developed in concurrence with computation. These methods — seemingly ignored during Gabriel Tarde’s time due to their complexity — are now commonplace and can be achieved with a click of a button through various SNA related software. As SNA developed in complexity, different attempts to use it have produced results that are still having an impact in the present. Some of the more famous applications include the impact of urbanization on individual well-being, how many human-relationships we are from anyone on the planet, the spread of innovations, the formation of markets, Communities, and citation networks in academia.
How SNA has been used and how AM will differ is the focus of this section. Specifically, this section will approach the philosophical underpinnings of the mathematical principles of SNA. AM being a straight application of SNA seems straight forward. After all, SNA is concerned with matricies and frequencies of associations between pairs in as much as AM is. As a result, this section approaches the question, “does the inclusion of non-human actors require new equations?” Or, does the math exist independently of the theory and the only reason the word “social” exists with “Social Network Analysis” is that when it was created, the place of non-human objects was still relegated to a non-consequential space?
There are additional concerns about generalizability and purpose of this method. While the purpose of SNA is not positivistic, the nature of SNA within HCI has often been mostly tangential and specifically human-centric to the point of its place as positivist or descriptive being unknown. The typical deployment of SNA is to evaluate human-to-human interactivity at scale through technology. What those results mean are often applied in larger and larger scales.
The reason that AM is being developed through this research is that I believe that SNA has applications that matter for the design of technologies. Much like anything, the inter-connected nature of the simple act of using a piece of software is enormously complex. We do not often treat using software with this in mind. In addition to decrypting users, AM has applications in evaluating the details of socio-technical systems at scale in more than just frequency and content of human-created messages in a network. After a brief discussion of a number of applications, I will return to discuss AM.
Within many SNA applications, human beings are simultaneously the cause and affected unit of any analysis. Two events sit at the center of both the setup and conclusion.
- Humans are the cause of action.
- Humans face the consequences of those actions.
Yet within all of this work, there are untold numbers of non-human actors that create the possibilities of this action. Additionally, many of these studies discuss the impact of a shallow understanding of an obsessively complex issue through the numerical frequency of some aspect of human-to-human contact. Rarely do these studies engage the way that non-human actors shift the focus of, and possibilities of, human action. Nor do they discuss the actions of those same non-human actors within those systems.
For example, the urbanization study linked at the top of this article discusses the results of a large-scale survey about the impacts of urban life on individual relationships. At its core, it sought to evaluate if community had been destroyed by modern society. It does this by evaluating the people individuals have direct contact with each day. It does not by map out the uniquely intimate relationship between humans through that of non-humans (subways, electricity, the highway, busy markets, pedestrian centered mass transit, and the like) in these urban areas. By its own definition, the author seeks to ferret out if the concept of community itself has been destroyed.
Additionally, the small world study from the infamous Stanley Milgram sought to evaluate the probability that two individuals from different places knew each other. The researchers of this study did this by sending packets of information with a letter stating that the intended target of this packet was somewhere in Boston, MA. Respondents were asked to mail this package to the target if they knew them personally. If they did not, they were asked to forward the packet to someone who might know them. Each time this occurred, the respondent also mailed a postcard to the researchers to know that it had been forwarded elsewhere. This study generated the concept that each person on earth is separated by 6 other people.
The spread of innovations study linked in the beginning of this piece concerned itself with medical doctors prescribing new medicines. By watching when doctors prescribed new medicines and for what reasons those medicines were prescribed, the researchers could evaluate how new ideas diffused across social networks. Generally, these researchers found that within the first few months of a drug’s release, strong social ties had a tremendous impact on prescribing new medicine. If the doctors had not prescribed the drug due to social network ties within 6 months, they often inevitably did so due to factors outside their social networks.
Each deployment of social network analysis presumes that humans only are those entities that communicate and spread ideas. However, within each of these studies are non-human actors that are hidden by other humans. For example, in the Urbanization study, the researcher ignores black boxes like mass transit, bus drivers, store clerks, and other entities that — while invisible when they work — can cause tremendous social interaction when they do not. In this way, the non-human actors of the urban network are constantly causing interactivity but we only notice it when it breaks down.
In the small world study, the researchers evaluate how many humans received the packets before they get to their destination. Through this, they conclude that 5.5–6 people must be contacted to get one stranger from anywhere to connect to another stranger — regardless of distance. This is a uniquely romantic notion and has been subject to a wide array of treatments in pop culture. Yet within that note of sociality are untold numbers of post office workers, information organizers in things like phone books, the socio-economic limitations of population drift, and other such factors that sit passively behind and within any of the respondents in this study.
Finally, the piece on diffusion of innovation sought to evaluate how quickly doctors prescribed medication based on their integration with their social network of other physicians. However, this work does not take into account drug representatives, advertising, patient questions, patient advertising, patient education, or even the research behind these medications.
Essentially within each of these studies is an absorption of non-human action that are masked by human-action. If we want a more detailed picture of these studies, all that need done is to add those non-human actors to the network to see where the centrality resides, to see where the influential actors are, to comprehend the struggle that resides within that network. To answer the question at the start of this post: “does the inclusion of non-human actors require new equations?”
The answer is no but the inclusion of non-human actors presents other issues. The most pressing of these issues is exactly what the results mean for designers or the design of new pieces of technology. For example, agency within AM is not tied to any particular object — human or otherwise. Instead, agency is something no single object can express. Only by recruiting objects to assemble can agency be expressed.
Through Catan then, the expression of agency could be a single player achieving their desire to win. However, this is not done without the dice producing resources to aid them. Agency is not achieved without understanding rules well enough to form a plan. Within that plan, the existence of other objects and the possibility of their recruitment toward the goal of Catan is meaningless. In this way, no single actor can be agentive.
There is of course the argument that without humans the activity would not form at all. Yet, without the pieces this is also true. Humans cannot play Catan if Catan is not present. The existence of Catan cannot be achieved without a designer yet that designer could not have created Catan without the existence of board games, without the existence of cardboard cutting factories, without the existence of wooden pieces, without the existence of colonialism.
And through this, we see the power of AM — it presumes none of these things. Association Mapping simply examines the interaction of objects within a given space for a given amount of time. While the game of Catan is used here as an example that very purposefully manifests these concepts, it is not a method that can only be used in games, game design, or play studies. This will be covered in the design fiction at the end of this dissertation. In the next section, we take our first steps into an overview of the findings.