AI for Social Good (AI4SG) is a research theme that uses and advances AI to improve the well-being of society. We introduce three lines of work that use machine learning and game theory to address real-world challenges in cybersecurity, food waste and security, and wildlife conservation. For cybersecurity, we provide a learning and planning pipeline for generic cyber deception and an algorithm to counter watering-hole attacks. In the food rescue setting, we develop a predictive model for the rescue claim status and a recommender system for volunteer engagement through push notifications. For wildlife conservation, we propose a media content monitoring system to provide early warning of infrastructure projects that might pose harm to conservation efforts.
Distilling lessons from these projects, we propose several "meta" research problems of AI4SG that, for the first time, attempt to address the common pain points in AI4SG projects. We develop the first iterative prediction-prescription framework to rigorously characterize the practical machine learning workflow in lots of social good projects across application domains. We propose to analyze the scoping phase of AI4SG projects to explore how AI researchers and stakeholders can work together efficiently. We also propose to study the deployment and adoption of AI4SG projects through an organizational learning perspective.
Fei Fang (Chair, ISR)
Rayid Ghani (MLD/Heinz)
Jason Hong (HCII)
Milind Tambe (Harvard University)
Zoom Participation. See announcement.