Look for any podcast host, guest or anyone

Listen

Description

If bail decisions were made by an Artificial Intelligence instead of judges, repeat crime rates among applicants could be cut by 25%. That is because an AI is consistent in its judgements: human judges are not.  


This variation in in bail decisions, as well as in sentencing, and many medical diagnoses and underwriting decisions are all examples of what Cass Sunstein calls "Noise" - unwanted variation in professional judgement, which is the theme of his new book Noise: A Flaw in Human Judgement, co authored with Danny Kahneman and Olivier Sibony. Professional judgement and discretion sound great in theory - especially to the professionals themselves - but in practice they end up creating a lottery in some high-stakes situations.


He tells me why there should be statues of the legal reformer Marvin Frankel all across the land; how we can reduce the "creep factor" of AI decision-making; how early movers influence opinion especially through social media, and much more. 


Cass Sunstein


Cass Sunstein is a professor at Harvard Law School, as well as the founder and director of the Program on Behavioral Economics and Public Policy. He has written hundreds of articles and numerous books, ranging from constitutional law to Star Wars. He has also served in several government positions, formerly in the Office of Information and Regulatory Affairs in Obama’s first administration and currently in the Department of Homeland Security to shape immigration laws. Sunstein’s influence is wide-reaching, most notably from his work on advancing the field of behavioral economics, making him one of the most frequently cited scholars. He is also a recipient of the Holberg Prize and has several appointments in global organizations, including the World Health Organization. 


More from Cass Sunstein



Also mentioned



The Dialogues Team


Creator: Richard Reeves


Research: Ashleigh Maciolek


Artwork: George Vaughan Thomas


Tech Support: Cameron Hauver-Reeves


Music: "Remember" by Bencoolen (thanks for the permission, guys!)