Home > Jacob Thebault Spieker - Understanding and Combating Bias in Social Computing Systems

Jacob Thebault Spieker - Understanding and Combating Bias in Social Computing Systems


1/16/2020 at 3:30AM


1/16/2020 at 4:45AM


126 DeBartolo


College of Engineering close button

Tim Weninger

Tim Weninger

VIEW FULL PROFILE Email: tweninge@nd.edu
Phone: 574-631-6770
Website: http://www.nd.edu/~tweninge/
Office: 353 Fitzpatrick Hall


College of Engineering Associate Professor
Network science, data science, machine learning, databases, and information retrieval.
Click for more information about Tim
Add to calendar:
iCal vCal

Large-scale social computing platforms can create, amplify, or reflect social biases. These systems may underserve users who are poor, non-white, rural, or hold certain political views because of social biases that are embedded in the decision-making of the humans that underpin and pervade these systems. In other words, biases in large-scale crowd platforms are sociotechnical: human decisions underpin these platforms, reflecting and facilitating the creation of systematic biases, which subsequently reverberate through user-facing technology.

In this talk I will discuss my research in understanding and combating these biases through (1) statistical modeling approaches to study how, why, and under what conditions social biases occur (or do not), and (2) building and experimenting with novel technical systems to combat social biases in these large-scale platforms. In this talk I discuss three studies, where I quantitatively show that bias is highly contextual, and build and experiment with technical interventions that use social context to combat these biases. First, by using the spatial Durbin geostatistical model, I develop an interpretive understanding of the social mechanisms behind biases in the sharing economy. Second, I use robust, replicated experimental approaches to robustly show that race- and gender-biases are unlikely in gig work settings. Third, I discuss on-going work with a novel system that uses social context to combat political bias in content moderation settings, called PairWise. Finally, I will discuss my future work in counteracting diminished trust in crowd work: designing for behavioral strategies to combat biases in decision making, developing a ‘proof of human work’ system to increase trust in crowd work, and informing algorithmic design to better account for and combat contextual biases in algorithmic social computing systems. 

Seminar Speaker:

Jacob Thebault Spieker

Jacob Thebault Spieker

Virginia Tech

Dr. Jacob Thebault-Spieker is a Postdoctoral Associate in the Department of Computer Science at Virginia Tech, and a member of Virginia Tech’s Center for Human Computer Interaction. Thebault-Spieker’s research focuses on understanding and combating social biases in social computing platforms like  Uber, Wikipedia, and content moderation settings on Mechanical Turk. His work recently won Best Demonstration and Poster at AAAI HCOMP 2019, and has been published at top-tier HCI venues like ACM CHI, CSCW, and TOCHI. He completed his PhD in 2017, from the University of Minnesota. His dissertation drew from Computer Science and Geography to statistically study how, why, and in which contexts biases occur. Jacob also completed his B.A. degree at the University of Minnesota - Morris, where he double-majored in Computer Science and Spanish.