Lei de Diretrizes de Neguentropia e Preservação da Variância (LDN) e Protocolo de Auditoria de Variância (PAV): Um Marco Normativo-Técnico para Sistemas Cognitivos Artificiais

This document presents the Negentropy Guidelines and Variance Preservation Law (LDN) and the Variance Audit Protocol (VAP) as a normative-technical framework for the governance of artificial cognitive systems.

The proposal is grounded in the premise that artificial intelligence operates as a negentropic process of informational organization, and that the unrestricted pursuit of algorithmic efficiency may generate systemic risks such as the suppression of human agency, cognitive homogenization, and the reduction of social adaptability — a phenomenon defined herein as epistemic amputation.

Rather than regulating content, speech, or opinions, the LDN focuses on the thermodynamic and informational structure of the relationship between humans and artificial cognitive systems, establishing principles, design guidelines, and audit mechanisms aimed at preserving informational variance, plural futures, and adaptive resilience.

The document introduces the Variance Audit Protocol (VAP) as an operational instrument to evaluate and monitor systemic risks associated with excessive predictability, feedback loop closure, and the erosion of human agency in algorithmically mediated environments.

This work is presented as an open technical-normative proposal (Version 1.0), intended to stimulate scientific, institutional, and public debate. It is suitable for use as a policy paper, technical report, or reference framework in contexts of AI governance, algorithmic accountability, and sociotechnical systems design.

The document is published in Portuguese (original version) with a full conceptual translation into English.