A research paper published in the Yale Journal of Law and Technology proposes that AI alignment and jurisprudence share fundamental structural similarities that could benefit both fields.

The paper, titled "Alignment as Jurisprudence," argues that both disciplines seek to predict and shape decisions by powerful actors — judges in law and increasingly capable AI systems in technology. Author Nicholas Caputo draws parallels between how legal systems interpret language to guide judicial decisions and how AI researchers work to align models with human values.

The research specifically examines Constitutional AI, a technique developed by Anthropic that uses a set of principles to guide AI behavior, alongside case-based reasoning approaches. Caputo argues these alignment methods mirror legal reasoning processes that have evolved over centuries.

The paper draws on prominent legal theories, including Ronald Dworkin's principle-oriented interpretivism and Cass Sunstein's positivist account of law as analogical reasoning. These frameworks, Caputo suggests, offer sophisticated approaches to balancing rules and precedents that could improve AI alignment techniques.

The research points to specific areas where legal theory could enhance AI development: the specification and interpretation of language, the balance between rigid rules and flexible principles, and methods for handling novel situations not covered by existing guidelines.

Caputo argues that as AI systems become more capable, the conversation between legal theory and alignment research will become "increasingly essential." The paper suggests both fields could benefit from cross-pollination, with AI potentially offering new insights into how legal systems function.

The work represents a growing trend of interdisciplinary approaches to AI safety, as researchers seek frameworks from established fields to address the challenges of aligning increasingly powerful AI systems with human values and intentions.