Title:

Causal inference and counterfactual learning from relational data

Abstract:

Predictive systems are used to make sense of rapidly increasing amounts of data and support human decision making, from what to read, to whom to date, whether to invite a job applicant for an interview, and what drug to develop next. However pervasive, these systems are limited in their ability to reason about cause and effect, and the consequences of their own behavior. In this talk, I will motivate the need for developing algorithms that can answer causal questions from real-world data which is inherently noisy, relational, and multimodal. I will focus on the problem of heterogeneous treatment effect estimation, understanding the individual variations in outcomes between individuals. I will show how learning algorithms can address this problem and take into account people's individual characteristics and predispositions. I will demonstrate the value of these algorithms in the context of understanding what drives empathy online and whether a given piece of information will go viral. 

Bio:

Elena Zheleva is an assistant professor of Computer Science at the University of Illinois at Chicago. Her research spans machine learning, causal inference, graph mining, and privacy. She has presented her research at top-tier conferences, and is the co-author of the book "Privacy in Social Networks." Prior to joining UIC, Elena spent a few years in industry as a data scientist, working on large-scale recommender systems and social network incentives. She built and led the data science team at LivingSocial, and later was a principal data scientist at Vox Media. She was also a policy fellow in the federal government for a year, contributing to U.S. initiatives at the intersection of data science and public policy. Elena earned a Ph.D. in Computer Science from the University of Maryland, College Park in 2011. More information about her research can be found at https://www.cs.uic.edu/~elena.