Machine learning models are becoming increasingly opaque to human examination, even to their designers. Yet these models are also increasingly used to make high-stakes decisions; who goes to jail, what neighborhoods police deploy to, and who should be hired for a job. But how can we practically achieve accountability and transparency in the face of increasingly complex models? And how do we know if the algorithmic decisions are fair or discriminatory - what does it mean for an algorithm to be fair? In this talk, we’ll discuss recent work from the new and growing field of Fairness, Accountability, and Transparency in machine learning. We will examine societal notions of fairness and non-discrimination and explain how these notions have been defined using a mathematical framework. We’ll also discuss recently developed strategies for auditing black-box models when given access to their inputs and outputs and for white-box interpretability in decision-making.
Please click here to access a reading list suggested by Professor Friedler from a related course at Cornell University.
This event is cosponsored by the University of Minnesota Libraries Digital Arts, Sciences + Humanities (DASH), the University of Minnesota Informatics Institute, the School of Social Work, the Institute for Research in Statistics and its Applications (IRSA), and the Departments of Computer Science and Electrical and Computer Engineering
Sorelle Friedler is an Assistant Professor of Computer Science at Haverford College and an Affiliate at the Data & Society Research Institute. Her research interests include the design and analysis of algorithms, computational geometry, data mining and machine learning, and the application of such algorithms to interdisciplinary data.
Sorelle is the Program Committee Co-chair for the new Conference on Fairness, Accountability, and Transparency (FAT*) and is one of the organizers of the Workshop on Fairness, Accountability, and Transparency in Machine Learning (FAT/ML). She has received a Fellowship and recent NSF grant for her work on preventing discrimination in machine learning. Her work on this topic has been featured in IEEE Spectrum, Gizmodo, and NBC News and she has been interviewed about algorithmic fairness by the Guardian, Bloomberg, and NPR.
Sorelle is the recipient, along with chemistry professors Josh Schrier and Alex Norquist, of two NSF Grants to apply data mining techniques to materials chemistry data to speed up materials discovery. Their paper on this work was featured on the cover of Nature and was covered by The Wall Street Journal and Scientific American.
Before Haverford, Sorelle was a software engineer at Alphabet (formerly Google), where she worked in the X lab and in search infrastructure. She received a Ph.D. in computer science in 2010 and an M.S. in computer science in 2007, both from the University of Maryland, College Park. She is a 2004 graduate of Swarthmore College.
Click here to download a .pdf transcript of Professor Friedler's talk.