William & Mary Bill of Rights Journal


This Article addresses some of the limitations of AI as a tool to preselect a long or shortlist of cases for a court at the apex of the judicial system to review. It focuses on the Colombian Constitutional Court, as an example of a court at the apex of the judicial system that has been historically responsive to claims for fundamental rights. Docket selection is an example of a classification problem using supervised learning, in which a machine groups data according to preestablished characteristics.

This Article draws from two different bodies of literature to analyze the consequences of using AI to preselect the docket of Colombia's Constitutional Court. It draws from political science models of the functions of courts and the structure of the judiciary to show the importance of docket selection to provide high courts with complete information about existing conflict in a society. Secondly, it draws from critiques of AI made by philosophers of the mind and cognitive scientists to illustrate the importance of human experience in docket selection. These critiques focus on the ways human experience and our interaction with the world shape our cognitive abilities and processes, to illustrate the differences between human intelligence and AI. The Article uses that literature to show that AI can help in identifying and preselecting cases based on well-defined legal categories. The use of AI, however, can also bring more significant problems. On the one hand, AI classification can sometimes include cases that do not deserve to be selected, producing which are known as false positives. However, the use of AI may produce more difficult problems because a machine cannot help the Court identify emerging patterns of social conflict, or naturalized patterns of social interaction that have recently been perceived as problematic. In other words, it also fails to preselect cases that are worth selecting, which are known as false negatives. Moreover, during times of social and political change, courts can become unresponsive or reactionary, and the costs associated with false negatives tend to be much greater.

This Article is divided into five parts. The first part briefly surveys the relevant AI technologies and elaborates on the criticisms made by philosophers and cognitive scientists. The second part introduces a model that describes the functions of courts, highlighting that such functions vary depending on the position of a court within the structure of the judiciary and showing the importance of information flows within the judiciary for the operation of a well-functioning legal system. The third part describes the role played by the Colombian Constitutional Court and how its responsiveness promoted an increase in fundamental rights litigation. The fourth part describes the current docket selection mechanism and how AI is starting to be used by the Court. The fourth part analyzes the consequences of using AI in preselecting the docket. Finally, the conclusion highlights some of the theoretical implications and makes recommendations for more sensitive ways in which AI can be used to promote greater efficiency and legal certainty.

This abstract has been adapted from the author's introduction.