Panel Paper: Reforming Administration: Replacing Human Decisions with Artificial Intelligence

Friday, July 20, 2018
Building 3, Room 213 (ITAM)

*Names in bold indicate Presenter

Justin Bullock, Texas A&M University

Administration scholars have devoted much attention to street-level bureaucrats and their motivations, discretion, and management capabilities, with particular focus on the relationship between management and performance. In the early 21st century administration scholars began discussing the role information and communication technology played in the implementation and administration of public policies. In particular, Bovens and Zouridis (2003), highlighted that in large, executive public agencies there was a significant change afoot in how these agencies implemented policy. The authors noted that there had been a shift from street-level bureaucrats implementing policy “on the street,” to screen-level bureaucracies where bureaucrats implemented policies, primarily or more commonly from behind a computer screen. Furthermore, they noted that this screen-level bureaucracy quickly shifted to a system-level bureaucracy where discretion landed in the hands of system analysts and software designers.

However, another dramatic change is on the horizon. Advances in machine learning and artificial intelligence suggest a not too distant future in which administrative decisions, street and system level, may best be implemented and administered by intelligent machines. In this paper, I discuss the strengths and weaknesses of human discretion as understood by cognitive and neural science, psychology, and administrative sciences. Following this I discuss recent advancements in machine learning and artificial intelligence and what this suggests for the future. Finally, I discuss the tradeoffs between allowing humans to make administrative decisions and allowing intelligent machines to make these decisions, and what set of tradeoffs can lead to a more effective, sustainable government.