FDS Talk: Serena Wang (UC Berkeley), “Bridging gaps between metrics and goals in modern machine learning ecosystems”



NOTE: This talk will be virtual only by Zoom.
Click the button below or go to: https://yale.zoom.us/j/92754270974

Abstract: The increasing sophistication and proliferation of machine learning (ML) across public and private sectors has been met with both excitement and apprehension – how do we study societal impacts in this new frontier? Key to understanding the societal impacts of ML is understanding the development and deployment of such systems, which is driven by numerical metrics such as accuracies, click rates, revenue, etc. Unfortunately, these metrics often don’t capture all developer goals or eventual societal impacts, which makes auditing and improving these systems difficult for both engineers and policymakers. In this talk, I will discuss two main approaches to bridging gaps between metrics and goals. First, I will discuss implementation gaps between theory and practice in Fair ML, using robust optimization approaches to handle distributional uncertainty. Second, moving beyond standard “fairness” paradigms, I will discuss recent work on understanding how metrics fit into an ecosystem of stakeholders. Specifically, I will show how causal metrics can improve incentives induced by ranking and accountability systems.

NOTE: This talk will be virtual only by Zoom.
Click the button above or go to: https://yale.zoom.us/j/92754270974

Serena L. Wang (UC Berkeley)

Speaker Bio: I am a final-year PhD student in Computer Science at University of California, Berkeley, advised by Michael I. Jordan. I am generously supported by the NSF Graduate Research Fellowship and the Apple Scholars in AI/ML PhD fellowship. I have also concurrently worked at Google Research at 20% time for the last six years, where I am part of the Discrete Algorithms Group with Ravi Kumar and previously worked with Maya Gupta.

My research focuses on understanding and improving the long term societal impacts of machine learning by rethinking ML algorithms and their surrounding incentives and practices. I’m particularly interested in gaps between metrics and goals, and how those gaps may be bridged through algorithmic improvements, analysis of multi-agent incentives and interdependence, and better understanding of intervention context. I employ tools from machine learning, robust optimization, causal inference, and most recently, economics.

Website: https://serenalwang.com/