JudgeAI

JudgeAI Technology

JudgeAI is an integrated system for research and decision-making. It brings together normative theory, a model of human relationships, empirical knowledge, mathematical computation and AI that orchestrates work on a problem.

The system investigates which action should be taken when different options affect people in different ways. The result may be the resolution of a dispute, a regulatory framework or a decision for an autonomous agent — together with the grounds that can be examined and refined.

A shared model of reality

At the foundation of JudgeAI is a normative ontology: a structured model of people, their roles, capabilities and dependencies.

One person can simultaneously be an employee, a parent, a property owner and a recipient of medical care. A change in one role can affect their other roles and the lives of other people. Companies and institutions are represented through relationships between employees, owners, customers, citizens and other participants.

The ontology provides a common language for describing different situations and comparing the consequences of decisions. Empirical data fills this structure with information about the circumstances: available resources, existing constraints, participants’ behaviour and causal relationships.

Possible worlds and the choice of action

For each action under consideration, JudgeAI constructs a distinct possible world. The options start from a shared state but change it in different ways: reallocating duties and resources, opening or restricting access, and creating new risks and opportunities.

The mathematical framework computes the immediate changes and how they propagate among connected participants.

For example, closing a factory may reduce pollution while also affecting workers’ incomes, their families and access to essential goods. Transition periods, compensation and alternative production methods change the overall picture.

The basis for comparison is preserving people’s capacity to live and act.

Health, safety, income, access to resources and legal mechanisms are considered through their effect on this capacity. The system compares the computed consequences, giving priority to the most severe threats facing the participants.

Law as Computation

Normative theory, the mathematical model and the architecture of computable choice.

First page of Law as Computation

Research and computation in one process

AI works with the normative model throughout the task. It interprets the request and documents, connects facts to the ontology, develops alternatives, researches sources and prepares inputs for computation. The comparison results then inform the next stage of work.

In a complex task, this process repeats at several levels. When designing regulation, the system first investigates the overall approach, then the related questions of funding, duties, exceptions and enforcement.

Each measure is considered in relation to the others: permitted activities must be feasible, and new duties must be supported by the necessary resources. Research, computation and preparation of the final document form the continuous development of a single solution.

Traceability as a basis for self-improvement

We have built a computational theory of law that connects evidence about a situation, the calculation of consequences and normative choice. The production of the decision itself is fully traceable: which circumstances are represented, which alternatives are compared and which computations determine the result.

This connection makes the decision an object of further investigation. An autonomous AI agent can examine how the choice depends on particular data, which disagreements between sources matter and whether the result holds when assumptions change. The ontology connects these questions to specific actors, relationships and consequences.

Further research refines the representation of the situation or expands the alternatives. Recalculation shows whether the decision has changed and why. Full traceability allows this work to continue while preserving the connection between each result and its empirical grounds.