In this talk I will reflect on the varieties of expertise I think are needed to build useful AI. More specifically, I will focus on what I know best: the opportunities and challenges at the intersection of lived experience, social theory, statistics, and machine learning. I will explore this intersection in the context of work early in my career on identifying bad actors using machine learning (e.g. terrorist-supporting accounts on social media), and more recent work focusing more holistically on understanding how AI functions in a complex socio-technical system (e.g. in the context of the child welfare system in New York state). I will discuss what I think this work has gotten right, what it has gotten wrong, and where I hope that we, together as a community at UB and in Buffalo more widely, can go from here.

Kenneth (Kenny) Joseph is an associate professor in the Department of Computer Science and Engineering and the Associate Director for AI & Society in the Institute for AI & Data Science at UB. He also runs the computation and equity lab at UB (https://cse.buffalo.edu/cubelab/). Prior to UB, he spent time as a postdoc at the interdisciplinary Network Science Institute at Northeastern University, and received a doctoral degree from Carnegie Mellon University in the field of Societal Computing. Kenny identifies as a computational social scientist, and has published widely in the social sciences and in applied machine learning venues.