Studying change and continuity in the precedents of the CJEU – a machine learning approach

This paper presents a new approach to studying continuity and change in precedent-setting by high courts and international courts, with a focus on the Court of Justice of the European Union (CJEU). Like other precedent-setting courts, the judgments of the CJEU include holdings specifying rules and principles. A core methodological challenge in the field lies in identifying changes to the authority of these holdings—their precedential value—without being dependent on resource-intensive doctrinal legal research. We suggest a methodological innovation that builds on the latest development in artificial intelligence. By relying on the CJEU’s own references to previous decisions, and the semantic similarity between strings of judgment text, we train a machine learning model to predict relevant citations to precedents for any legal proposition relating to EU law. These “expected citations” are used to assess—over time—the precedential value of court decisions, i.e. their life cycles as authoritative expressions of rules and principles. Such trajectories, in effect, signify continuity and change in specific precedents. While our approach falls short of capturing what the law “is” in a doctrinal sense, by identifying the life cycles of precedents we can capture where and when change occurs. Our methodological innovation opens for theoretical advancement of theories of policy change close to the core of political science and law.

• Register
• Add to Calendar