Inteligencia artificial, opacidad algorítmica y decisiones judiciales

Gabriel Ravelo-Franco; Juan Pablo Sarmiento Erazo; Jesús Eduardo Sanabria Moyano · Dixi (2026) · DOI: 10.16925/2357-5891.2026.03.02

Resumen

This study examines the feasibility of employing artificial intelligence systems in judicial decision-making in light of the structural challenges posed by algorithmic opacity. It contrasts approaches that regard certain levels of functional opacity as tolerable with those that contend that any form of inintelligibility undermines democratic legitimacy, epistemic justice, and fundamental rights. The analysis also incorporates insights from XAI and evidentiary theory to show that existing explanation methods do not enable the reconstruction of the reasoning, facts, or causal links essential for judicial processes. Through an exhaustive review of recent scholarship and an assessment of normative, philosophical, and procedural standards, the study demonstrates that the technical, epistemic, sociotechnical, and institutional limitations of current models preclude meeting the minimum requirements of explainability, auditability, and rational control. In view of these findings and the Inter-American standard affirming the right to understand the reasons underlying a judicial decision, the article proposes a precautionary moratorium on the use of AI in the judicial domain until robust explainability standards are validated through inter-institutional deliberative processes.

Instituciones: Universidad Continental