Hermenêutica e o problema da vigilância algorítmica: aproximação do conceito ricoeuriano de instituições justas com o EU AI Act

Abstract

Faced with the implications of algorithmic surveillance and the centrality of narrative identity in Paul Ricoeur, this dissertation analyzes the EU AI Act: a pioneering AI regulatory framework in economic and social relations. A scrutiny was conducted on certain provisions of the EU AI Act originally conceived to address the opacity problem in artificial intelligence systems. The methodology employed is anchored in philosophical hermeneutics, particularly in the Ricoeurian semiotic dimension of the sign. The corpus consists of the EU AI Act (Regulation EU 2024/1689), whose methodological distancing of its text (the transition from the explanation of the sign to comprehension), in Ricoeurian terms, allows for the investigation of mediations regarding the phenomenon of algorithmic surveillance. The guiding hypothesis, corroborated over the course of the investigation, demonstrated a presumed convergence between Ricoeur’s approach and certain provisions of the EU AI Act — such as the categorical ban on predictive policing based on personality profiles. In this convergence, a presumed safeguard of identities is observed, insofar as the European legislator imposes limits on artificial intelligence systems: on algorithmic prediction and data opacity. The research results further specify the support for this latter circumstance in some of the provisions formulated by the legislator, whose intent was to prevent the conversion of man into a predictable metric — such as the prohibition of predictive policing (EUROPEAN UNION, 2024, art. 5, no. 1, letter g) — and to protect the right to obtain clear explanations resulting from algorithmic decisions (EUROPEAN UNION, 2024, art. 86, no. 1), a prerogative through which the EU AI Act can touch upon the Ricoeurian horizon. The caveat, however, lies in the fragility of the purely positive norm in safeguarding just institutions.