Generative Misinterpretation

I have a new article draft, Generative Misinterpretation, with Ben Sobel and David Stein, forthcoming in the Harvard Journal on Legislation. We argue against recent proposals to use LLMs in the judicial process. We combine an empirical critique, showing that experimental results in using LLMs to perform interpretation are brittle and arbitrary, with a jurisprudential critique, explaining why commonly offered justifications for using LLMs don’t work in the context of judging. Here is the abstract: