Last week I had the great opportunity to talk to Professor Sinan Aral (follow him on twitter @sinanaral). Sinan is one of the Rockstars in Science that is pushing the limits of social science. As I talked about in my Persuasive Technology Keynote social science is on the verge of a paradigm shift due to the accessibility of data about peoples decisions, possibilities to measure networks effects, and new machine learning and statistical methods that allow for meaningful analysis of such data. Sinan is one of the scientists that is quickly pushing the boundaries of social science outwards.
Why Sinan Puts the ‘S’ in Social Science
Sinan explained the core of his work: understanding network effects. This understanding is important not just in many applications that he himself focuses on such as the effects of advertisements and new product introductions in networked environments. The research namely also answers a very general question: how are we influenced by the behaviors of others around us? This will in due time answer questions such as: How are my kids influenced by the behaviors of others in the class? How does the performance of one team member in a company influences the others in the team and the performance of the company? How does me drinking in a bar inspire you drinking in that same bar (given that you are there too)?
Sinan Aral is one of the few scientists—next to people like our previous meet up guest Dean Eckles—that is truly getting his head around network effects. Is the fact that you buy a product that one of your friends owns caused by your friend buying it? Or rather would you have bought it anyway since you share preferences with your friends? This distinction, between so-called peer-influence and homophily, is one of the important focuses of Sinan his research.
Some of Sinan’s latest work focuses on identifying influential people within networks. Some people introduce new ideas, while others seem to follow. How do influencers cluster in a network, and can we identify them a-priori so we are able to target them directly? Discussing this work with Sinan gave me a number of new insights as to influencers and followers: How come influencers cluster together, while even for influencers it should be easier to influence followers than their influential friends? Are clusters of influencers actually clusters of people with a lot of influence on different topics? Sinan published some great work in this area in Science.
Sinan on Persuasion and Networks
Finally we discussed heterogeneity in responses to persuasion—the thing we focus on with PersuasionAPI. If you are more susceptible to authority arguments than other customers, would it make sense that the same is true for your friends (due to homophily)? If so, could we leverage the known network structure to be able to learn individual profiles faster? Sinan suggested a number of ways in which we could introduce aggregated measures of properties of the network at an individual level to borrow strength in building a persuasion profile. I would like to extend this to including the full network information—not only a vector of (e.g.) tie strength or network influence. However, these are next steps. In any case I believe persuasion profiles and networks will find each other in the near future.
It was great talking to Sinan, who is besides a science rockstar also an amazing guy. We will have him over for drinks one day—and you will be invited.
For more on Sinan’s work, collaborations, and projects, see some of his recent publications:
The Diversity-Bandwidth Tradeoff (American Journal of Sociology)
Information Technology and Information Worker Productivity (Information Systems Research)
Distinguishing Influence Based Contagion from Homophily Driven Diffusion in Dynamic Networks (Proceedings of the National Academy of Sciences)
Computational Social Science (Science)