JudgeAI
JudgeAI Research develops law as a computable procedure of normative choice. We study how the computable connection between empirical evidence and normative choice enables autonomous AI to progressively develop a decision and examine its robustness. This process is examined in Empirical Self-Improvement in Normative Reasoning.
JudgeAI Research publications describe the development of one research program: from algorithmic dispute resolution to the formal conditions of autonomous normative choice.
This paper sets the initial frame: court automation requires procedural inference that moves through facts, evidence, applicable law, and legal reasoning.
It describes JudgeAI as a system where the algorithmic layer matters because of predictability, transparency, and reduced arbitrariness.
The second paper shifts the question from automating individual court operations to a fully automated judicial procedure.
At the center is a combination of LLMs and logical algorithms: claim, defenses, evidence, and decision structure are analyzed as one process.
The current work generalizes judicial and legislative modes into a single theory of autonomous normative choice. A normative decision is defined as the result of an internal logic where the admissibility of a rule depends on empirically observable consequences of its application.
This study examines how JudgeAI's computable, traceable normative decisions enable autonomous research to identify consequential questions and discover preferable regulatory alternatives. The expanded UAE analysis connects broad threat exploration, empirical refinement and the design of implementable interventions.
Controlled experiments examine how consequential relationships, missing information and additional alternatives affect the choice. A separate experiment follows 16 threat directions into local research tasks and recalculates the shared model after contextual refinement. The results support the potential for better-grounded decisions while preserving the distinction between conditional findings and established intervention effects.