Machine Learning for Human Decision Making - Jure Leskovec

Abstract:

In many real-life settings human judges are making decisions and choosing among many alternatives: Medical doctor deciding a treatment for a patient, criminal court judge making a decision about a defendant, a crowd-worker labeling an image, and a student answering a multiple-choice question. Gaining insights into human decision making is important for determining the quality of individual decisions as well as identifying human mistakes and biases.
In this talk we discuss the question of developing machine learning methodology for estimating the quality of individual judges and obtaining diagnostic insights into how various judges decide on different kinds of items. We develop a series of increasingly powerful hierarchical Bayesian models, which infer latent groups of judges and items with the goal of obtaining insights into the underlying decision process. We apply our framework to a wide range of real-world domains, and demonstrate that our approach can accurately predict judge’s decisions, diagnose types of mistakes judges tend to make, and infer true labels of items.

Bio:

Jure Leskovec is assistant professor of Computer Science at Stanford University and chief scientist at Pinterest. Computation over massive data is at the heart of his research and has applications in computer science, social sciences, economics, marketing, and healthcare. This research has won several awards including a Lagrange Prize, Microsoft Research Faculty Fellowship, the Alfred P. Sloan Fellowship, and numerous best paper awards. Leskovec received his bachelor's degree in computer science from University of Ljubljana, Slovenia, and his PhD in in machine learning from the Carnegie Mellon University and postdoctoral training at Cornell University. You can follow him on Twitter @jure.