(coauthored by Harry Surden)
At least since Jeremy Bentham, legal formalists in the Anglo-American tradition have dreamed of making the law clearer, more precise, and predictable in its application, with the ultimate goal of limiting or eliminating the human subjectivity of judging. The strongest versions of this "legal formalist" project envisioned a system of laws that could be applied by a machine, with perfect reliability and no trace of personal, political, or other bias—and no exercise of the fallible faculty of human judgment.
Enter ChatGPT and other LLMs—a new form of generative artificial intelligence that has received tremendous attention since the public launch of ChatGPT 3.5 in late 2022. In just two short years, these models have improved and proliferated at an astonishing pace. With some important caveats, they are now capable of outperforming most humans at many complex cognitive tasks, including the bar exam and medical licensing exams.
But using AI to interpret the Constitution (or decide other legal questions) does not eliminate the need for normative judgment. It simply shifts those judgments to different stages of the decision-making process. Like matter or energy, judgment in constitutional interpretation can be shifted around, dispersed or concentrated. It might be transferred from one decision-maker or one stage in the decision-making process to another. But when it is squeezed out of one part of the interpretive process, it inevitably pops up somewhere else. We call this the law of conservation of judgment.
For a fuller explanation, you can read our new paper, “Artificial Intelligence and Constitutional Interpretation.”