Actor map
The system identifies affected people, institutional roles, interests, dependencies, and conflicting positions.
Normative Universe Engine
JudgeAI builds a computational object for normative choice: a system of actors, dependencies, admissible legal worlds, and consequences through which a decision changes the position of affected people.
Technical foundation
Law as Computation
The central task is to evaluate the consequences of a normative choice in a situation where a decision affects many actors, changes dependencies between them, and triggers probabilistic causal chains.
The main computational object is the Normative Universe: a formal space of possible legal worlds for a specific situation. Each world describes affected people, an admissible decision option, legal constraints, allocation of duties, causal consequences, the probability of those consequences, and the limiting risk for each affected position.
Actors are connected through contract, debt, subordination, access to a resource, care, control, procedure, public service, information asymmetry, or an enforcement mechanism.
The backend builds a map of actors and dependencies. Institutions, companies, platforms, public bodies, and legal roles are resolved into people and typological classes of people whose capacity to act changes through the decision.
The LLM layer serves language: fact extraction, candidate-world generation, critique, calibration notes, and drafting. The computational core is the Normative Universe Engine: structured schemas, actor ontology, admissibility gates, probability and reliability protocol, memory of prior threat patterns and consequence comparison.
Normative choice itself happens through consequence evaluation: the system compares possible legal worlds by how they change the position of affected actors, while the public document becomes a legal-language translation of the selected world.
Legal worlds
For each position, the system builds materially distinct worlds: regulatory architectures in lawmaking and disposition scenarios in arbitration.
Consequence graph
For each world, the system records the first consequence of the legal mechanism and the chain of secondary consequences for each actor.
Option comparison
The computational layer accounts for probability, reliability, causal-chain length, background threats, and risk distribution between actors.
Program paper
Law as Computation
The Technology section is based on this working paper. It sets the formal definition of autonomous normative choice and explains why a legal decision must be computed through actors, admissible worlds, constraints, probabilities, and consequences.
The paper is the reference point for the Normative Universe Engine: the system architecture on this page is the product implementation of that formal research program.
Application modes
One engine for different normative tasks
Normative Universe Engine is used for action review, AI Arbitration, and AI Lawmaking.
The system evaluates who an agent’s action affects and what risk moves between positions.
Case materials are translated into legal worlds where admissible outcome options are compared.
A normative problem is unpacked through actors, risks, duties, safeguards, and legislative output.