Overview of the Application of Machine-Learning in Administrative Law

This article was authored by Lauren Beadle, a student at American University Washington College of Law. The views expressed below are those of the author and do not represent the views of ACUS or the Federal Government.

Technology and administrative law are intertwined. The accessibility of computers increased public participation in agency rulemaking and improved access to government for citizens. E-rulemaking facilitates communication between the public and agency decision makers. The Social Security Administration serves millions of people through online forms and expects to expand its platform to serve more. Administrative law grappled with regulating new technology, particularly with the advent of computers and the internet, and will continue to regulate new advancements, like drones and driverless cars. However, administrative law faces a new challenge, how to regulate using technology to increase fairness, transparency, and efficiency.  

ACUS previously adopted recommendations about integrating technology and administrative law including Recommendation 2018-3, Electronic Case Management in Federal Administrative Adjudication, adopted at the most recent plenary, as well as Recommendation 2017-1,  Adjudication Materials on Agency Websites, Recommendation 2011-1, Legal Considerations in e-Rulemaking, and Recommendation 2011-8, Agency Innovations in E-Rulemaking. These recommendations focus on widely accepted practices (i.e. electronic case management and e-rulemaking). However, the adoption of machine-learning algorithms remains controversial and will likely continue to be at the center of debate.

Machine-learning algorithms make inferences about data without being explicitly programmed. Essentially, the algorithm “learns” from the data to produce a prediction. This process is referred to as a “black box” because humans only see the inputs and outputs. Machine-learning is not synonymous with artificial intelligence. The goal of artificial intelligence is to remove human error, whereas machine-learning algorithms produce a prediction (output) through pattern recognition. Machine-learning can help agencies make better decisions by processing larger data sets faster than humans.

In Cary Coglianese and David Lehr’s article, Regulating by Robot: Administrative Decision Making in the Machine-Learning Era, they acknowledge the benefits of machine-learning in administrative law and argue that the use of machine-learning does not circumvent the principles of nondelegation, due process, equal protection, or transparency in agency action.

Agencies, large cities, and private companies are beginning to use machine-learning. In the context of administrative law, machine-learning can be split into two categories: 1) adjudication by algorithm and 2) regulation by robot. Adjudication by algorithm can be appropriate when quantifiable data determines an outcome, such as eligibility for benefits. The City of Los Angeles uses regulation by robot to improve traffic flow and reduce delays. The algorithm synthesizes large quantities of data and adapts traffic lights accordingly. At a personal level, email filters, Netflix suggestions, and dating app matches are a few examples of machine-learning already a part of everyday life.

Technology should not be blindly adopted and each agency must evaluate its specific needs to determine if it should adopt machine-learning. Fairness and efficiency are the cornerstones of our system of laws and adopting machine-learning, when appropriate, complements those values.

This post is part of the ACUS Intern Blog Series.

Share on FacebookShare on TwitterShare on Linkedin