Why Are AI Systems Racist, Sexist, and Generally Unfair, and Can We Make Them Fair?: Q&A with Philip Thomas
Abstract: In this talk I will discuss examples of AI systems that produce unfair behavior, reinforcing existing social inequalities. I will then discuss what it really means for an AI system to be "unfair", before reviewing how the unfair behavior of AI systems can sometimes be a consequence of bias present in the data used to create and train the AI system, as well as how AI systems often produce unfair behavior even when there is no bias present in the training data. Finally, I will discuss some of our recent efforts to create new machine learning algorithms that can easily be applied fairly and responsibly to real-world applications.
About the Speaker:
Philip Thomas is an Assistant Professor of Computer Science in the College of Information and Computer Sciences at UMass Amherst, where he is the co-director of the Autonomous Learning Lab. His research interests are in reinforcement learning, decision making, and ways to ensure the safety of artificial intelligence systems, with emphases on ensuring the safety and fairness of machine learning algorithms and on creating safe and practical reinforcement learning algorithms. Prof. Thomas earned BS and MS degrees at Case Western Reserve University and a PhD at UMass Amherst. He joined the CICS faculty in 2017 after completing a postdoc at Carnegie Mellon University. He has published in top AI conferences and journals, including the prestigious journal Science, and testified to the U.S. House Committee on Financial Services, Task Force on Artificial Intelligence in 2020.
About the Series:
The Computing and Social Justice Series brings CICS researchers and the public together to critically assess how computing innovation intersects with vitally important issues like structural bias, civic participation, economic inequality, and citizen privacy.