A caixa-preta algorítmica e a crise do espaço público : opacidade, explicabilidade e fragmentação discursiva

Abstract

This article examines how algorithmic opacity undermines the conditions under which arguments circulate, are evaluated, and acquire public validity in digital environments mediated by recommendation and content moderation systems. It argues that algorithmic opacity is not the source of discursive fragmentation or of the epistemic injustices produced by these systems; rather, it functions as a structural stabilizer of these harms by hindering their identification, contestation, and correction. To this end, the article examines cases of asymmetric content moderation and the evolution of the Brazilian regulatory framework, from the Brazilian Artificial Intelligence Strategy (EBIA) to the Brazilian Artificial Intelligence Plan (PBIA), drawing on argumentation theory, social epistemology, and AI ethics to analyze the effects of system opacity. The findings indicate that Explainable Artificial Intelligence (XAI) mechanisms and regulatory models based exclusively on the articulation of general principles are insufficient to address the institutional and epistemic factors that sustain system opacity. As its main contribution, the article proposes six ethical guidelines oriented toward transparency, contestability, appropriate distribution of control, plural representation, preservation of the deliberative infrastructure, and the cultivation of institutional epistemic virtues. It concludes that algorithmic transparency should be understood as an epistemic, institutional, and democratic condition indispensable for preserving the legitimacy of the public sphere in societies increasingly mediated by artificial intelligence systems.