Holding Algorithms Accountable: Hidden Biases in Machine Learning

November 1, 2016 12:00PM E51-275

We live in an increasingly data rich world, and in recent years there have been profound breakthroughs in teaching computers to infer meaning from petabytes of information. These breakthroughs in machine learning and AI promise to make our lives more efficient, but unfortunately in many areas hidden biases in the underlying data can taint any classifier's decisions. For the past few years, Jeff Larson has been investigating the use of these algorithms, and he'll discuss implications and downsides of using them in areas like criminal justice, surveillance, and natural language processing.

Jeff Larson is a data reporter at ProPublica. From 2013 to 2015 he reported extensively from the Snowden leaks. He exposed a top secret NSA project to weaken web security, and detailed the close relationship between the intelligence community and AT&T. He is a winner of the Livingston Award for the 2011 series Redistricting: How Powerful Interests are Drawing You Out of a Vote.

ProPublica is an independent, non-profit newsroom that produces investigative journalism in the public interest.

GOV/LAB Events