April 6, 2021

12:00 pm / 1:00 pm

Venue

ZOOM

Recorded Seminar:

https://wse.zoom.us/rec/share/-0p1JhMCNx91Z4zHR5gqucib5YGIMXpNXAlczzfH82CYC_FSMnQ0FK8CpOZPRV61.8zv-KgpAyIOymP3O?startTime=1617724020000


Maria De-Arteaga, PhD 

Assistant Professor
Information, Risk, and Operations Management Department
McCombs School of Business
TheUniversity of Texas at Austin


“Mindthe gap: From predictions to ML-informed decisions”

Abstract Machine learning (ML) is increasingly being used tosupport decision-making in critical settings, where predictions havepotentially grave implications over human lives. In this talk, I will discussthe gap that exists between ML predictions and ML-informed decisions. The firstpart of the talk will highlight the role of humans-in-the-loop through a studyof the adoption of an algorithmic tool used to assist child maltreatment hotlinescreening decisions. We focus on the question: Are humans capable ofidentifying cases in which the machine is wrong, and of overriding thoserecommendations? The second part of the talk will focus on the gap between theobserved outcome that the algorithm optimizes for and the construct of interestto experts. We propose influence functions based methodology to reduce this gapby extracting knowledge from experts’ historical decisions. In the context ofchild maltreatment hotline screenings, we find that (1) there are high-riskcases whose riskis considered by the experts but not wholly captured in thetarget labels used to train a deployed model, and (2) the proposed approachimproves recall for these cases.

 

Bio Maria De-Arteaga isanAssistant Professor at the Information, Risk and Operation Management (IROM)Department at the University of Texas at Austin, where she is also a corefaculty member in the Machine Learning Laboratory. She received a jointPhD inMachine Learning and Public Policy from Carnegie Mellon University. Herresearch focuses on the risks and opportunities of using machine learning tosupport experts’ decisions in high-stakes settings. Her work has been awardedthe Best Thematic Paper Award at NAACL’19, the Innovation Award on Data Scienceat Data for Policy’16, and has been featured by UN Women and Global Pulse intheir report Gender Equality and Big Data: Making Gender Data Visible. She is arecipient of a 2020 Google Award for Inclusion Research, a 2018 MicrosoftResearch Dissertation Grant, and was named an EECS 2019 Rising Star.