Bio: Ryan Shi is a final-year Ph.D. candidate of Societal Computing in the School of Computer Science at Carnegie Mellon University. He works with nonprofit organizations to address societal challenges in food security, sustainability, and public health using AI. His research has been deployed at these organizations worldwide. Shi studies game theory, online learning, and reinforcement learning problems motivated by nonprofit applications. He is the recipient of a Siebel Scholar award, an IEEE Computer Society Upsilon Pi Epsilon Scholarship, and a Carnegie Mellon Presidential Fellowship. Shi grew up in Henan, China before moving to the U.S., where he graduated from Swarthmore College with a B.A. in mathematics and computer science.
Talk Title: Towards the Science of AI for Nonprofits
Talk Abstract: I work on AI for nonprofits, with nonprofits, so that it could be used by nonprofits. This talk is about delineating AI for nonprofits as a research discipline of its own. I will start with a 4-year collaboration with a food rescue organization. We developed a series of machine learning-based algorithms to help the organization engage with food rescue volunteers. Our tools have been deployed since February 2020. Our randomized controlled trial showed that the algorithm not only improved the accuracy of engagement, but also made it significantly easier for rescues to get claimed. We are rolling it out to over 15 cities across the US. Motivated by the pain points we experienced in this and other works, we proposed bandit data-driven optimization, a new learning paradigm that integrates online bandit learning and offline predictive models to address the unique challenges that arise in data science projects for nonprofits. We prove theoretical guarantees for our algorithm and show that it achieves superior performance on simulated and real datasets. I will conclude the talk with an inquiry of the state of AI for nonprofits as its own research discipline. Using another project where we developed and deployed an NLP tool for with a conservation nonprofit as a primer, I argue that AI for nonprofits needs its own set of research questions that will systematically guide future projects. I will share a few of these directions that I am currently working on and plan to work on in the near future.