How to Build Ethics into Robust Artificial Intelligence (2017-2018)
The goal of this project was to combine computational methods, philosophy, game theory and psychology to develop robust and ethical moral artificial intelligence (“moral AI”) to direct autonomous agents.
The project team’s research address several of the main challenges in designing moral AI through three subprojects.
- The team used in-person experiments and online platforms such as Amazon Mechanical Turk (MTurk) to ask people what types of factors they believe kidney exchange algorithms should and should not take into account. Team members built a website for large-scale data collection for online participants, analyzed participants’ responses to find common ethical themes and then revised current kidney exchange algorithms so that their recommendations of how kidneys should be allocated will take these ethical themes into account. This “bottom-up” approach relied primarily on new data to learn the moral features that should be incorporated into AI algorithms.
- The team also combined principles from moral philosophy and economic game theory to design scenarios and games that asked participants to judge the actions described or displayed in specific moral situations. Team members began by focusing on the “trust game” and had MTurk participants play the trust game against one another and then report how morally wrong or acceptable they thought their actions and their partner’s actions were. Team members collected data, began analysis on performance and used participants’ morality ratings to refine game theoretic notions so that they account for behavior in the trust game more accurately. This “top-down” approach relied primarily on game theory and ethical theory to generate algorithms that can make ethical choices.
- Team members also asked participants whether they thought humans or computers should be making decisions in a wide range of scenarios and applications.
Fall 2017 – Spring 2018
Max Kramer, Jana Schaich Borg, Vincent Conitzer, Walter Sinnott-Armstrong. “When Should Computers Make Decisions?” 2018. Proceedings of the 2018 AAAI/ACM Conference on AI, Ethics, and Society 204-209.
Rachel Freedman, Jana Schaich Borg, Walter Sinnott-Armstrong, John P. Dickerson, Vincent Conitzer. “Adapting a Kidney Exchange Algorithm to Align with Human Values.” 2018. AAAI Conference on Artificial Intelligence.
Adapting a Kidney Exchange Algorithm to Align with Human Values (poster by Rachel Freedman, Jana Schaich Borg, Walter Sinnott-Armstrong, John P. Dickerson, Vincent Conitzer), presented at Bass Connections Showcase, April 18, 2018
Vincent Conitzer, Walter Sinnott-Armstrong, Jana Schaich Borg, Yuan Deng, Max Kramer. “Moral Decision-Making Frameworks for Artificial Intelligence.” 2017. AAAI Conference on Artificial Intelligence.
Who Should Get the Kidney (game aligned with kidney exchange algorithm)
Vincent Conitzer and Walter Sinnott-Armstrong. How to Build Ethics into Robust Artificial Intelligence ($200,00 grant awarded from the Future of Life Institution, 2015)
Walter Sinnott-Armstrong. Towards a Culture of Questioning: Accountability, Humility, and Public Discourse ($225,000 grant awarded from John Templeton Foundation, 2017)
Cultures of Collaboration: Managing the Moral AI Lab (Kenzie Doyle)
This Team in the News
- Vincent Conitzer, Arts & Sciences-Computer Science
- Jana Schaich Borg, Social Science Research Institute
- Walter Sinnott-Armstrong, Arts & Sciences-Philosophy
/graduate Team Members
Cassandra Carley, Computer Science-PHD
Lok Chan, Philosophy-PHD
Abbas Zaidi, Statistical Science - MS
/undergraduate Team Members
Anika Mukherji, Interdept Neuro/Comp Sci (BS)
Weiyao Wang, Mathematics (BS), Computer Science (BS2)
/yfaculty/staff Team Members
Emmanuel Chevallier, Arts & Sciences-Statistical Science
Kenzie Doyle, Duke Institute for Brain Sciences
Joshua Skorburg, Fuqua School of Business
Siyuan Yin, Arts & Sciences-Philosophy
/zcommunity Team Members
John Dickerson, University of Maryland