CS547 Human-Computer Interaction Seminar   (Seminar on People, Computers, and Design)

Fridays 11:30am-12:30pm PT · Gates B3 · Open to the public
Previous | Next
Saleema Amershi
Microsoft Research
Towards Usable Machine Learning
March 6, 2015

It is widely believed that machine learning based applications will broadly impact human productivity and decision making. However, experts caution that the complexity of machine learning can also lead to unintended consequences ranging from irrelevant content recommendations to damaging financial and political decisions made on the bases of inaccurate predictions. While advances continue in improving the accuracy and efficiency of machine learning algorithms, a complementary strategy is needed in improving the capabilities of the people who are fundamentally involved in using and building these machine learning based systems. In this talk, I will give examples of tools we have been developing at Microsoft Research to support the people involved in the machine learning process. Specifically, I will present ModelTracker, a tool we designed to support performance analysis and debugging, and Structured Labeling, a technique we designed to facilitate concept evolution during model building. I’ll then discuss some of the open challenges and opportunities in working towards more usable machine learning.


Saleema Amershi is a Researcher in the Computer Human Interactive Learning (CHIL) group at Microsoft Research. Her research lies at the intersection of human-computer interaction and machine learning. In particular, her work involves designing and developing new tools to support both end-user and practitioner interaction with interactive machine learning systems. Saleema received her Ph.D. in computer science from the University of Washington’s Computer Science & Engineering Department in 2012.