Three Laws Blog: Law and Artificial Intelligence

Daly on Artificial Intelligence and Administrative Tribunals

Paul Daly, University of Ottawa, Common Law Section, has published Artificial Intelligence and Administrative Tribunals. Here is the abstract.

In this paper, I will examine the implications of artificial intelligence for administrative tribunals. I will briefly set the scene in Part I by describing the potential advantages and disadvantages of advanced technology for administrative tribunals. Although I have reservations about integrating AI into the daily operations of public administration, administrative tribunals may be a uniquely favoured site for technological experimentation. Since their origins in the 19th century the raison d’etre of administrative tribunals has been more efficient and effective decision-making. The very thing that artificial intelligence promises – efficiency and effectiveness – is exactly that which administrative tribunals have long primarily been designed to achieve. This is not to suggest, however, that administrative tribunals should move blindly forward and embrace AI in all shapes and forms. In Part I, I also suggest that careful, contextual analysis is required before using AI and lay out several typologies that are helpful in conducting the necessary analysis. In Part II, I then discuss the ways in which administrative tribunals can use AI: communicating with the public and enhancing the accessibility of their websites; assigning decisions more effectively to members of the tribunal; triaging files to ensure decisions are made as quickly as possible; drafting decisions; and, conceivably, making decisions. These use cases are set out in increasing order of audacity. Machines making websites more usable for individuals is one thing, making life-defining decisions about individual rights and obligations quite another. In the end, though, I will argue that there is a broad range of potential uses for AI in administrative tribunals.

Download the paper from SSRN at the link.

Ye, Luceri, and Ferrara on Auditing Political Exposure Bias: Algorithmic Amplification on Twitter/X Approaching the 2024 U.S. Presidential Election

Jinyi Ye, Luca Luceri, and emilio Ferrara, all of the University of Southern California, have published Auditing Political Exposure Bias: Algorithmic Amplification on Twitter/X Approaching the 2024 U.S. Presidential Election. Here is the abstract.

Approximately 50% of tweets in X's user timelines are personalized recommendations from accounts they do not follow. This raises a critical question: what political content are users exposed to beyond their established networks, and how might this influence democratic discourse online? Due to the black-box nature and constant evolution of social media algorithms, much remains unknown about this aspect of users' content exposure, particularly as it pertains to potential biases in algorithmic curation. Prior research has shown that certain political groups and media sources are amplified within users' in-network tweets. However, the extent to which this amplification affects out-of-network recommendations remains unclear. As the 2024 U.S. Election approaches, addressing this question is essential for understanding the influence of algorithms on online political content consumption and its potential impact on users' perspectives. In this paper, we conduct a three-week audit of X's algorithmic content recommendations using a set of 120 sock-puppet monitoring accounts that capture tweets in their personalized "For You" timelines. Our objective is to quantify out-of-network content exposure for right-and left-leaning user profiles and to assess any potential biases in political exposure. Our findings indicate that X's algorithm skews exposure toward a few high-popularity accounts across all users, with rightleaning users experiencing the highest level of exposure inequality. Both left-and right-leaning users encounter amplified exposure to accounts aligned with their own political views and reduced exposure to opposing viewpoints. Additionally, we observe a right-leaning bias in exposure for new accounts within their default timelines. We hope this empirical study promotes discussion on the transparency and accountability of social media algorithms, contributing to critical issues on safeguarding election integrity and fostering a more informed digital public sphere.

Download the article from SSRN at the link.