AI in Adjudication and Administration: A Status Report on Governmental Use of Algorithmic Tools in the United States

Cary Coglianese, Lavi Ben Dor
2019 Social Science Research Network  
The use of artificial intelligence has expanded rapidly in recent years across many aspects of the economy. For federal, state, and local governments in the United States, interest in artificial intelligence has manifested in the use of a series of digital tools, including the occasional deployment of machine learning, to aid in the performance of a variety of governmental functions. In this Article, we canvass the current uses of such digital tools and machine-learning technologies by the
more » ... iary and administrative agencies in the United States. Although we have yet to see fully automated decision-making find its way into either adjudication or administration, governmental entities at all levels are taking steps that could lead to the implementation of automated, machine-learning decision tools in the relatively near future. Within the federal and state court systems, for example, machine-learning tools have yet to be deployed, but other efforts have put in place digital building blocks toward such use. These efforts include the increased digitization of court records that algorithms will need to draw upon for data, the growth of online dispute resolution inside and outside of the courts, and the incorporation of non-learning risk assessment tools as inputs into bail, sentencing, and parole decisions. Administrative agencies have proven much more willing than courts to use machine-learning algorithms, deploying such algorithmic tools to help in the delivery of public services, management of government programs, and targeting of enforcement resources. We discuss already emerging concerns about the deployment of artificial intelligence and related digital tools to support judicial and administrative decision-making. If artificial intelligence is managed responsibly to address such concerns, the use of algorithmic tools by governmental entities throughout the United States would appear to show much future promise. This Article's canvass of current uses of algorithmic tools can serve as a benchmark against which to gauge future growth in the use of artificial intelligence in the public sector. AI in Adjudication and Administration Artificial intelligence has begun to permeate many aspects of U.S. society.1 In settings as varied as medicine, transportation, financial services, and entertainment, digital technologies continue to emerge that rely on machinelearning algorithms to process vast quantities of data and make highly accurate predictions that can often outperform human ability to perform similar tasks.2 As a result, the potential utility of artificial intelligence in the legal field has not gone unnoticed, with scholars, attorneys, and judges beginning to examine the implications it could have for the United States legal system.3 This Article seeks to capture the state of the art of the current uses of digitization, algorithmic tools, and machine learning in domestic governance in the United States. It serves, in effect, as a status report on non-military governmental
doi:10.2139/ssrn.3501067 fatcat:7pe7gky3zngptktwbx3dk3ghpq