Fairness, Accountability and Transparency in Algorithmic Decision Making
Friday, September 22, 2017 — 3:30PM - 5:00PM
Decision making, once the exclusive purview of human experts, is now often offloaded to artificial intelligence (AI) systems. The range of such applications is steadily increasing – from benign recommendations by Amazon, Netflix and Facebook, to credit approval decisions and and law enforcement predictions. Algorithmic decision making systems promise data-driven predictions free from human bias. However, recent high profile cases such as modern face recognition systems which may be highly sensitive to skin color, and modern credit and loan decision systems that may be biased by race, remind us once again that artificial intelligence and algorithmic decision making systems are only as good as their design. How do AI decisions compare to the track records of the human experts they have replaced? What are the roles of systems designers, the law, and activists in the design, maintenance and oversight of these systems? The panel will these issues with a focus on both the successes and failures of automated decision making systems, their current and future impact on society, and what can be done to ensure that such systems are fair, accountable, transparent and free of bias.
Sanmi Koyejo, University of Illinois at Urbana-Champaign