Aaron Roth
Associate Professor of Computer and Information Science
University of Pennsylvania

with commentary by:

Richard Berk
Chair, Department of Criminology
Professor of Statistics and Criminology
University of Pennsylvania

 

View event recording and presentation materials

 

Machine learning undergirds many technological advances today, from email spam filters to self-driving cars. It is also increasingly being used to make consequential decisions in domains such as lending, policing, and criminal sentencing. When machine learning moves from the private to the public sectors, and government uses algorithms to make decisions, additional concerns emerge, especially if machine learning produces inequitable outcomes. In this seminar, Professor Roth will offer a basic tutorial on machine learning, while also pointing out some of the basic pitfalls that can lead to discrimination even when an algorithm’s objective function was not designed with discriminatory intent.  Professor Berk will offer additional commentary.

The Optimizing Government Project is a Fels Policy Research Initiative and is open to the public. 

Panel discussion featuring:

Samuel Freeman
Avalon Professor of the Humanities, Department of Philosophy
University of Pennsylvania

Nancy Hirschmann
Professor of Political Science
University of Pennsylvania

Seth Kreimer 
Kenneth W. Gemmill Professor of Law
University of Pennsylvania

Moderated by:

Cary Coglianese 
Edward B. Shils Professor of Law and Professor of Political Science 
University of Pennsylvania

 

View event recording and presentation materials

 

As machine learning is increasingly contemplated to support decision-making by governmental officials, questions have arisen about the fairness of decisions generated with the aid of artificial intelligence. Technologists seek to understand whether they can design algorithms to address fairness and equality concerns, while policymakers and citizens seek to evaluate available technological applications against standard moral and policy principles.  Both technical and policy deliberation demands clarity about several key questions.  What does “fairness” really mean? Does a commitment to equality demand merely that algorithms do not rely on characteristics such as race and gender to generate forecasts?  Or does equality demand more?  This workshop will bring together leading scholars from law, philosophy, and political theory to illuminate how fairness and equality are conceptualized in each of these fields. The workshop, the second in a series of four taking place throughout the fall, seeks to inform current deliberations about the design and use of machine learning in government.  

The Optimizing Government Project is a Fels Policy Research Initiative and is open to the public. 

This program has been approved for 1.5 total CLE credits (1.0 substantive, 0.5 ethics) for Pennsylvania lawyers. CLE credit may be available in other jurisdictions as well. Attendees seeking CLE credit should bring separate payment in the amount of $60.00 ($30.00 public interest/non-profit attorneys) cash or check made payable to The Trustees of the University of Pennsylvania.

Michael Kearns
Professor and National Center Chair
Department of Computer and Information Science
Founding Director, Warren Center for Network and Data Sciences
Founding Director, Penn Program in Networked and Social Systems Engineering
University of Pennsylvania

with commentary by:  

Sandra Mayson
Research Fellow
Quattrone Center for the Fair Administration of Justice
University of Pennsylvania

View event recording and presentation materials

Opaque machine learning models continue to be adopted by governments in a range of contexts, raising questions about whether such automated decisions are compatible with notions of fairness and equal protection. Answering these questions will require both careful policy study and a technical understanding of what makes algorithmic decision-making effective. This workshop will review technical solutions to these challenges and how they might impact government use of automated decision-making processes. The workshop is the third in a series taking place throughout the fall, dedicated to exploring policy and technical challenges to using artificial intelligence in government.  

The Optimizing Government Project is a Fels Policy Research Initiative and is open to the public.

Featuring:

Tom Baker
William Maul Measey Professor of Law and Health Sciences  
University of Pennsylvania

Benedict G.C. Dellaert 
Professor
Erasmus School of Economics 
Erasmus University Rotterdam

Moderated by:

Richard Berk
Chair, Department of Criminology
Professor of Statistics and Criminology
University of Pennsylvania

Consumers are increasingly turning to automated systems for suggestions about how and where to invest. The proliferation of “robo-advisors” – automated financial advice tools based on machine learning algorithms – points to the potential need for new types of financial services regulation and raises important questions about markets for complex goods and services. More broadly, automated investment products provide a robust case study for exploring the limits and possibilities of consumer protection as automation extends into throughout the marketplace.

This event is a Special Seminar co-sponsored by the Optimizing Government Project and the Penn Program on Regulation.

For the People, By the Robots? Democratic Governance in a Machine-Learning Era

A panel discussion featuring: 

John Mikhail
Agnes N. Williams Research Professor and Professor of Law, Georgetown University

Helen Nissenbaum
Professor of Media, Culture, and Communication & Computer Science, and Director, Information Law Institute, New York University,  

David Robinson
Principal, Upturn, and Adjunct Professor, Georgetown Law 

Machine learning algorithms are being used to automate decision-making across a wide array of government functions, from financial regulation to criminal justice. Yet the very features that make machine learning so useful – speed, scalability, and automation – may also come into tension with democratic principles like transparency and accountability, especially if human decision-making by accountable public officials is replaced with automated decision-making by machine. This workshop will convene scholars from law and computer science to elaborate on the potential tensions between machine learning and democratic governance and to highlight possible legal and technical responses. The workshop is the first in a series of three taking place throughout the spring, dedicated to exploring challenges related to transparency and accountability in the government’s use of machine learning.

The Optimizing Government Project is a Fels Policy Research Initiative and is open to the public. For questions about the event, please email optimizing@law.upenn.edu or visit the Optimizing Government homepage.

This event will be streamed live via this URL.

This program has been approved for 1.5 substantive CLE credits for Pennsylvania lawyers. CLE credit may be available in other jurisdictions as well. Attendees seeking CLE credit should bring separate payment in the amount of $60.00 ($30.00 public interest/non-profit attorneys) cash or check made payable toThe Trustees of the University of Pennsylvania.

Can Technology Be Democratic? Transparency and Accountability in Machine Learning

A panel discussion featuring: 

Sorelle Friedler 
Assistant Professor of Computer Science, Haverford College

Andrew Selbst
Visiting Fellow, Yale Information Society Project  

As governments turn to machine-learning algorithms to automate their decision-making, concerns arise about how those tools might affect transparency, accountability, and other core democratic values. Machine-learning algorithms pore through vast datasets to generate forecasts used to support government action.  Although they can increase accuracy in decision-making, such algorithms substitute an opaque analytic process for the democratic vision of an informed and engaged citizenry. Some commentators further worry that an embrace of machine learning by governments could signal a new era of surveillance and social control. Despite these concerns, a growing body of research suggests that machine learning processes can be built to prioritize transparency and can be adapted to enhance, not dilute, democratic government. This workshop will bring together scholars at the intersection of computer science and public policy to discuss possible ways to translate democratic values into algorithmic processes. The workshop is the second in a series of three taking place throughout the spring dedicated to exploring policy and technical challenges raised by the use of artificial intelligence by governmental institutions.

The Optimizing Government Project supported by the Fels Policy Research Initiative and is open to the public. For questions about the event, please email optimizing@law.upenn.edu or visit the Optimizing Government homepage.

This program has been approved for 1.5 substantive CLE credits for Pennsylvania lawyers. CLE credit may be available in other jurisdictions as well. Attendees seeking CLE credit should bring separate payment in the amount of $60.00 ($30.00 public interest/non-profit attorneys) cash or check made payable toThe Trustees of the University of Pennsylvania.

Big Data and Government: Meeting the Real-World Challenges

A panel discussion featuring: 

Cary Coglianese   
Edward B. Shils Professor of Law
Professor of Political Science
Director, Penn Program on Regulation
University of Pennsylvania Law School

Stephen Goldsmith 
Daniel Paul Professor of the Practice of Government
Ash Center for Democratic Governance and Innovation, Harvard Kennedy School
Former Mayor of Indianapolis and Deputy Mayor of New York 

Dennis P. Culhane
Professor, Dana and Andrew Stone Chair in Social Policy 
Co-Principal Investigator, Actionable Intelligence for Social Policy 
Director of Research, National Center for Homelessness Among Veterans
School of Social Policy & Practice, University of Pennsylvania

The automation of government decision-making by artificial intelligence is no longer a theoretical concern. As legal and technical experts continue to wrangle with fundamental questions about how machine learning impacts democratic governance, policymakers are actively implementing A.I. initiatives in the public and private sector. They must not only grapple with conceptual questions about fairness, transparency, and efficiency, but also with more practical concerns about navigating the political process. How can political leaders be convinced of the value of digital advances in government? What are the roadblocks to more widespread adoption of machine learning and other digital advances in government? This workshop will bring together scholars who will speak to the challenges of implementing machine learning and digital innovations in government. This is the final workshop in a series of three dedicated to exploring challenges raised by the use of artificial intelligence by governmental institutions.

The Optimizing Government Project, supported by the Fels Policy Research Initiative, is open to the public. For questions about the event, please email optimizing@law.upenn.edu or visit the Optimizing Government homepage.

 

This program has been approved for 1.5 substantive CLE credits for Pennsylvania lawyers. CLE credit may be available in other jurisdictions as well. Attendees seeking CLE credit should bring separate payment in the amount of $60.00 ($30.00 public interest/non-profit attorneys) cash or check made payable toThe Trustees of the University of Pennsylvania.