Words or code first? Is the legacy document or a code statement the better starting point for complexity-reducing legal automation?

Abstract

Law is a critical tool that humans have created to assist them in managing complex social interactions. Computational Law holds the potential to significantly enhance our capacity to express and manage legal complexity, and a number of advantages can result from restating public and private legal rules in computable form. Capturing that potential depends in part on the approaches taken to automation. One set of choices involves whether to translate directly into code from existing natural language statements of laws, regulations and contracts or whether to step back, envision the basic structure underlying those statements and build a software approach that reflects that structure in a code-native manner. We argue that many advantages can flow from the second approach, and we present a specific use case of a simplified insurance policy as an example of this approach. Large language models may assist in this process, but are not yet a replacement for a code-native utility. This article is part of the theme issue 'A complexity science approach to law and governance'.