Have you heard of Human Science yet?
It is not surprising, since in Ukraine you can not often here this topic mentioned in conversation. So we decided to clear things up by having a chat with Bogdan Yamkovenko who is a People Research Scientist at Facebook and an Assistant Professor in the Department of Psychology at Towson University.
Bogdan: Just to anchor this and make sure we all use the same terminology, we can refer to this area of work as either social network analysis (SNA) or network science. I’m assuming these questions all stem from the HBR article, and in it I mainly talk about methods used in SNA.
SNA is a method that stems from graph theory (a branch of mathematics), sociology, and psychology. Network Science is closely related, but also incorporates methods from physics (eg. spin glass algorithm for community detection).
The basic idea behind these methods is that we are all embedded in social networks (friends, coworkers, etc) and our position in these networks influences our behavior. Some argue that the effect of networks on behavior is stronger than factors typically studied in psychology (motives, personality, etc). Network structure can be used to understand how the information flows through the network (rumors or infection), who are the people who are likely to transmit this information faster, which individuals have power and prestige, and who serves as a bridge between otherwise disconnected communities.
In other applications, with dynamic networks or networks that are observed over time, we can study the effects of contagion. For example, if a person has friends who smoke, does that person also begin smoking?
Bogdan: We can probably talk about this from two angles — academic and business. In academic circles people always look for new/better theories to explain behavior. With that people look for new variables to study. In psychology in particular people talk about the replication crisis, indicating that many older theories do not hold much water or at least cannot be used with as much certainty any more. So naturally researchers look to other approaches, and SNA is promising. It’s not a new approach, I must say, and has been used over the last 50-60 years in many studies. But it’s experiencing something of a renewal. I think the realization is that the individual behavior is driven very much by what their networks look like. For example, research on strong and weak ties discusses benefits of having these types of ties in various contexts like job search and team cohesion. In leadership research we can investigate whether leaders who broker between disconnected groups or individuals are more successful. In team research we can examine how centralized vs decentralized teams differ in performance and socialization. We can also study network composition (the extent to which we have ties to similar others) and how this affects life outcomes.
In business, SNA is popular because people hope for easy answers. Many consulting organizations picked up the idea that you can collect some data and show a visual depiction of a network (sociogram). They always look cool so this stuff sells. But visual is one thing and the prediction of behavior is another. SNA methods are relatively complex and apply to a limited set of questions. They require careful design. In many cases SNA projects fail to deliver actionable insights because SNA is not used properly. So I would say that if this keeps going eventually business will lose interest. It usually has a very short attention span.
Bogdan: I worked with a colleague who studied this. He and I both worked on a study to examine job seeker behavior. That context was very appropriate for a network approach (because jobs leads often come from weak ties and individuals don’t often look to these ties as a source for job leads). Also over time I read more and more about this and branched into other applications.
Bogdan: there are many of them. I already mentioned sociology and psychology — to predict/understand behavior of individuals and groups, to understand social mobility, diffusion of innovation, adoption of new practices. In medicine — to study drug adoption and infection spread. In criminology — to understand terrorist and criminal networks. In neuroscience and biology — to understand brain networks and protein networks. In business — to understand relationships within and between teams, to identify key individuals (hubs) for whatever purpose.
Bogdan: There is no short answer. Network measures can be split into individual and whole-network. For example density is the number of ties that are present in the network as compared to the number of possible ties. This is a network measure and can be used to compare networks of the same size. Centrality is the individual level measure, which can be used to characterize an individual’s position in the network. There are various types of centrality that are used depending on the question we want to answer. There are many others, too many to list here. A good source for these is a book by Wasserman and Faust (1999).
Many of these come from graph theory and linear algebra. For example shortest path is a measure of how quickly information will travel through the network. It’s a number of steps that are needed for each node in the network to reach every other node. A diameter of the network is a measure of the shortest distance between two most distant nodes in the network.
Bogdan: Not sure I’m the right person to answer this question. It’s broad and cannot be answered without establishing some parameters. Social networks theory will tell you that we are social animals embedded in networks. But that’s still too broad and doesn’t provide any concrete insight.
Bogdan: There are many good sources, some of the best ones are listed below:
Wasserman and Faust (1999). Social Networks Analysis
Scott (2017). Social Network Analysis
Barabassi (2017) Network Science
Borgatti and Everett (2013) Analyzing Social Networks
Burt (1995). Structural Holes
Granovetter — anything