Lydia Liu `17, an operations research and financial engineering major who also earned the Center’s certificate in statistics and machine learning, was awarded a best paper award, and gave an oral presentation of her work to a packed crowd, at this year’s International Conference on Machine Learning in Stockholm, Sweden.
Liu presented the paper, “Delayed Impact of Fair Machine Learning,” at the six-day July conference, sponsored by the International Machine Learning Society. Liu is the first author with co-authors Sarah Dean, Esther Rolf `16, Max Simchowitz `15, and Moritz Hardt, an assistant professor at the University of California, Berkeley’s Department of Electrical Engineering and Computer Sciences. Liu, Dean, Rolf, and Simchowitz are all doctoral students at Berkeley.
The paper studies how certain machine learning systems that are built to minimize prediction errors may discriminate against underrepresented groups in terms of race or gender due to biased historical data. As an example, Liu and her co-authors examined bank lending data. To address this issue, the authors propose constructing models that predict how decisions or legislation may impact vulnerable populations. You can access the paper here and read about their findings in this blog post.