Yan Leng, Collective Behavior over Social Networks
Social networks are ubiquitous and shape individual behaviors; yet, behavioral data over networks are complex and stretch the limit of conventional analysis and models. In this talk, Dr. Leng presented two complementary projects that extend existing prediction and inference methods over social networks. First, she investigated how social network influences behavior. Using two large-scale mobile phone data, she found out that social influence on non-routine behaviors spread up to more than three degrees of separation with a decay pattern. To explain such a phenomenon, she built a Bayesian model in which individuals locally aggregate information and dynamically make adoption decisions. Second, Dr. Leng introduced a machine learning framework to infer the unobserved network structures from decisions. Specifically, sge modeled individual decisions using the linear-quadratic game and inverted the sparse network structures from the equilibrium actions. Together, this line of work contributed to the understanding of human behavior over networks via computational methods, which are important for managing large-scale behavioral change.
Yan Leng is a Ph.D. candidate at the MIT Media Lab. She will join the McCombs School of Business, the University of Texas at Austin as an Assistant Professor in Information Management in Fall 2020. She holds master degrees in Computer Science and Transportation Engineering, both from MIT. Yan is a network scientist working on social science problems. Her research lies in the intersection of machine learning, network theory, and causal inference. She uses large-scale behavioral data to understand collective human behavior over social networks and builds computational techniques for solving societal and organizational issues.
