My story

How I became a computational social scientist


I was using traditional quantitative tools of social science (e.g., surveys and experiments) to study human online behavior until 2009, when a huge turning point in my career and in my life came. I become a member of the Swarm Agents Club in Beijing, which is a scientific club of 500+ members (most of them are scientists and engineers) run by a small group of 11 physicists and computer scientists.

Since then I defined myself as a “computational social scientist” who embraces the methods from other fields of science such as physics and computer science in the research of human behavior.  The master thesis was my first trial, in which I applied the Page Rank Algorithm to evaluate the influence of social network members.
During 2009-2013 I was completing my PhD. program in Web Mining Lab at the City University of Hong Kong. My exploration of various methods for computational social science, including Webpage crawling, agent-based simulations, and complex network analysis, was greatly supported by Jonathan J. H. Zhu, the director of the lab.

My research

Since 2009 I have been working together with Dr. Jiang Zhang, a physicist at Beijing Normal University, to study the dynamics of collective attention. Our collaboration has lead to 4 articles in PloS ONE, Physical Review E, and European Physical Journal B, as well as several conference papers. In these papers we suggest that the Kleiber’s law, a universal scaling relationship between body size and energy consumption in living systems, can also be found online between users and clicks. Thus, websites can be viewed as “virtual living systems” that grow at the expense of the attention flow of users. The scaling exponent of clicks vs. users reveals the efficiency of websites in terms of converting human resource into computational resources. We also connect this efficiency to the flow structure of collective browsing and suggest that chain-like structure is more efficient than star-like structure. Our findings generated interest among mass media and was reported by New Scientist under the title “Why social networks are sucking up more of your time” and Science Daily under the title “Online activity grows in a similar pattern to those of real-Life networks”.
In 2011, during my visit in Australia National University I began to collaborate with Dr. Robert Ackland, an economist/Web scientist, to evaluate the efficiency of the Web and to study the preference of users. In our articles in Social Network Analysis and Mining we show the mismatch between clickstreams and hyperlinks, which implies that hyperlink is a failed design to serve for the navigation of users.  In another conference paper we used the index number theory in economics to characterize the consumption of  webpages as “information goods”. We find that users can be separated into different groups according to their deviation from the ideal and rational consumption behavior.

My industrial background

I benefit greatly from my one-year industrial experience. When I was working at Baidu, the largest Chinese search engine company, my work involves applying machine learning techniques (such as Bayesian models and collaborative filtering models) to PB-level data using parallel computing systems. The industrial experience enhanced my interest in generating software/algorithms as research outputs. I am now working on a Python module designed for flow network analysis. I also provide consultation to my research collaborator’s startup company, which is developing Apps for minute-level weather prediction using machine learning techniques. Recently, this startup team won the championship among more than 1,000 competitors in a contest and obtained an investment of $ 530,000 USD.

My social activities and teaching experience

I am now one of the 11 core members of the Swarm Agents Club and I am the manager of two mail groups, “Artificial Intelligence” (475 members) and “the Web and Complex networks” (30 members). These groups aims to connect engineers, scientists, and investor who are interested in topics such as Big Data, Machine Learning, Network Analysis, etc. Besides these two online groups, I also host online study groups with a particular emphasis on computational social science.

During these group I have developed skills to provide hands-on Python/R programming guidance to zero background social science students. Recently, I tied multiple documents together into an e-book Data Mining in Social Science, which aims at providing a comprehensive introduction on data mining skills including data collection, analysis, and visualization. The goal is to equip social scientists with a minimum set of Python programming skills to handle computational social science studies.

2014 Fall Social science data mining, Center for the Study of Institutional Diversity, Arizona State University, 5 members (graduate students). Reference Book: Data Mining in Social Science by myself.

2014 Spring Programing collective intelligence, Center for the Study of Institutional Diversity, Arizona State University, 11 members (graduate students). Reference Book: Data Mining in Social Science by Toby Segaran.

2013 Winter Python and Machine Learning Methods, Center for the Study of Media and Market, Peking University, 5 members (senior data analysts in the industry). Reference Book: Building Machine Learning Systems with Python by Willi Richert & Luis Pedro Coelho.

Leave a Reply