How Algorithms & Machine Learning Work

This session will begin with a tutorial on how algorithms and machine learning work in order to provide lawyers with a better understanding of how these technologies apply to solving real world problems. For example: how does machine learning help a review site spot fake reviews, a social media platform identify misinformation campaigns, or sites identify a banned user trying to rejoin the site under a new identity? Our tutorial will explore the limits of what algorithms and machine learning can and cannot do. The demonstration will be followed by a broader policy discussion, which will explore some of the practical, legal and ethical challenges of using algorithms:

• Since it’s almost impossible to run a large network with millions of users without algorithms, how do you strike the right balance between machine learning and human moderators for legal compliance and/or takedowns to comply with company policies, e.g., copyright, pornography, hate speech.

• Does more reliance on machines to make decisions create new problems like unfair takedowns and lack of transparency?

• Under what circumstances does legal liability for machine-made decisions attach?

• What happens when a government agency (such as under the new GDPR “right to an explanation”) requires platforms to disclose an explanation of algorithmic decision making and – not only is the algorithm proprietary – but the complexity of machine learning may make it impossible for even the platform to know precisely why a particular choice is made, e.g., why certain content was delivered.

Panelists:
Jim Dempsey, Executive Director, Berkeley Center for Law & Technology
Travis Brooks, Group Product Manager – Data Science and Data Product, Yelp (Tutorial)
Glynna Christian, Partner, Orrick
Cass Matthews, Senior Counsel, Jigsaw