Artificial intelligence systems can generate text that presents the outward appearance of a reasoned judicial decision while lacking the genuine justification required to sustain it, according to a theoretical framework published in Frontiers in Artificial Intelligence. The study highlights this vulnerability as an immediate practical concern because automated systems are already drafting judicial documents that judges subsequently validate in assisted judicial action, while prevailing legal frameworks lack precise criteria to distinguish valid legal reasoning from superficial imitation.
To resolve this regulatory gap, researchers Roberto Leonardo Cruz-Núñez, Hugo Armando Aguilar-Brito, Andherson J. Aguirre-Lanegra, and Nuccia Seminario-Hurtado reconstructed the constitutional duty to give reasons into an explicit set of formal correctness conditions. Their methodology synthesizes legal-dogmatic analysis with formal reconstruction, formalizing these judicial requirements within models of structured argumentation and defeasible deontic logic under a framework designated as computable due process.
The investigative method grounds its formal specifications in Peruvian constitutional jurisprudence, identifying it as the continental tradition’s most articulated typology of defective judicial reasoning. The analysis maps six recognized judicial reasoning flaws onto five formal correctness conditions that any admissible argument must fulfill: the argument must demonstrably exist, its logical inference must be formally valid, its underlying premises must be justified, it must satisfy the legally applicable standard of proof, and it must directly answer every claim raised in the proceeding.
The researchers emphasize that the required justification threshold is not uniform but varies according to the gravity of the judicial act, establishing that a simple procedural order and a custodial criminal conviction do not demand the same evidentiary or inferential threshold. Furthermore, the underlying structure extends beyond Peru, as Brazilian statutory law enacts six equivalent grounds of defective reasoning, Spanish law specifically grades the judicial duty to give reasons across decision types within its regulation of assisted judicial action, and both European and Inter-American courts enforce analogous requirements.
When applied to a decided case, the formal model demonstrated that existing Peruvian legal doctrine regarding circumstantial evidence already defines the necessary structural components of a formal argument.
The authors demonstrate that current large language models fail to satisfy these constitutional conditions, citing evidence that specialized legal tools utilizing source retrieval generate false information between 17% and 33% of the time. The authors note that language models are particularly vulnerable when a special legal rule clashes with a general norm, as statistical models tend to apply the norm that appears with greater frequency in their training corpus rather than the hierarchically appropriate rule demanded by the specific case.
From these formal and empirical findings, the study derives an architectural design rule for assisted judicial technology: the burden of producing inspectable justification must reside on a symbolic reasoning layer, while final qualitative assessment remains reserved for the human judge. The researchers conclude that computable due process provides an objective criterion to separate artificial intelligence systems that legitimately sustain the judicial function from those that merely simulate it, providing a verifiable standard for the constitutional review of assisted court rulings.
The hypothesis and theory study, published in the Technology and Law section of Frontiers in Artificial Intelligence, was developed through an institutional collaboration between the Universidad Autónoma de Chiapas in Tuxtla Gutiérrez, Mexico, and the Technological University of Peru in Lima, Peru. Their work addresses central requirements in explainable artificial intelligence and human oversight by establishing explicit boundaries for automated legal reasoning within assisted judicial action.
Subject of Research: Formal logical criteria and computable due process standards for artificial intelligence in assisted judicial decision-making
Article Title: Computable due process: the guarantee of a reasoned decision as a specification of admissible legal reasoning in assisted judicial action
Article References: Cruz-Núnez, R. L., Aguilar-Brito, H. A., Aguirre-Lanegra, A. J., & Seminario-Hurtado, N. (2026). Computable due process: the guarantee of a reasoned decision as a specification of admissible legal reasoning in assisted judicial action. Frontiers in Artificial Intelligence, 9, Article 1957004. https://doi.org/10.3389/frai.2026.1957004
Image Credits: AI Generated
DOI: 10.3389/frai.2026.1957004
Keywords: assisted judicial action, computable due process, defeasible deontic logic, structured argumentation, judicial reasoning, legal artificial intelligence, constitutional case law, legal-dogmatic analysis, formal reconstruction, standard of proof, inspectable symbolic layer, human oversight, retrieval error rates, normative conflict, circumstantial evidence, judicial justification, constitutional review
News Source: Denise Maddox. (November 11, 2026). Computable Due Process Model Establishes Legal Reasoning Standards for Judicial AI. Scienmag.



