SLATE VII: Problems with Probability

Professor Anthony Niblett with the University of Toronto Faculty of Law, Toronto, Canada, delivered the seventh seminar in the Seminars on LAw and Tech (“SLATE”) series. In his seminar, drawn from his paper “Problems with Probability” which was jointly published with Professor Casey from Chicago Law School, Anthony explored the issues relating to the use of artificial intelligence (AI) systems to help triage the backlog of cases and facilitate the resolution of civil disputes in courts as well as for arbitration proceedings. Anthony noted that AI can accomplish this by establishing the facts of cases and by way of predicting the outcomes of disputes. In Anthony’s view, as to the former, which he termed “algorithmic fact finding”, no real problems would arise because these involve the probabilistic predictions of some ground truth for which there is a correct answer. However, when AI is used to predict answers to questions of legal liability, such as whether a given plaintiff is vicariously liable, by for instance, comparing the facts of the instant case to a rich dataset of previously decided cases, Anthony questioned the significance of such a prediction or “algorithm legal prediction” by the algorithm.  Focusing on the issue of how these predictions, which are probabilistic in nature, might determine if legal liability would be established in any given case, Anthony opined that important questions remain as to how these probabilistic predictions should be converted into legal decisions. Anthony posited the impact which the AI-generated probability of the plaintiff or defendant winning would have on the likelihood of increases in settlement between the parties as well as the quantum of damages payable, and iterated through the various scenarios and alternatives for converting AI predictions to outcomes. He also noted that this is premised on the degree to which algorithms are used to make legal decisions, access by parties to such algorithms, the extent to which the judges accept such algorithmic decisions and even the parties perception as to whether the judges have access to such decisions. Finally, to better understand these issues, Anthony also shared with the participants an ongoing research project that he is currently conducting, which involves surveying participants who are simulated adjudicators to assess if their views of a hypothetical party’s chances of winning a case will change if they, in addition to the parties, have access to the AI predictions.

The seminar ended with a series of insightful questions from the audience, comprising academics, practitioners, students as well as visitors from Nankai University Law School, Shangdong University School of Law and Sichuan University Law School. The participants and the moderator of the seminar, A/P Daniel Seng, also exchanged information about the extent and degree of use of AI for adjudication and its acceptance by courts in the US, China and in ASEAN. The consensus from everyone who participated is that more research is required to address many of the pressing questions raised in this seminar before AI can be adopted for adjudication purposes by the courts.