Human and Artificial Intelligence in Decision Systems for Social Development
In this two-part thesis, I investigate and discuss the transformative potential of hybrid decision systems—which integrate human and artificial intelligence elements—in the promotion of development goals, with applications in poverty alleviation and health. Part I of this thesis focuses on a sequence of three studies that investigate advantages and disadvantages of a variety of multi-agent decision systems, starting from complex human-to-human networked systems, and progressing into human-AI systems. The sequence also progresses from controlled online experiments to field deployments with substantial societal impact and institutional complexity. Part II addresses two key challenges that arise as one applies cutting-edge science and technology in the context of real-world decision systems for development. In particular, this work provides academic and practical contributions on: 1) achieving algorithmic fairness and cost-efficiency via adaptive information collection; and 2) preserving privacy and mapping its tradeoff against utility for development goals.
Committee members:
Alex "Sandy" Pentland, Toshiba Professor of Media Arts and Sciences
Iyad Rahwan, Director, Center for Human and Machines, Max Planck Institute for Human Development
Esteban Moro, Associate Professor, Universidad Carlos III de Madrid
Rayid Ghani, Director, Center for Data Science & Public Policy, University of Chicago