Avaliação de impacto algorítmico: desafios éticos e jurídicos na identificação e mitigação do viés de gênero

Abstract

This dissertation investigates the social, ethical, and legal impacts of AI, with an emphasis on gender bias, proposing guidelines for algorithmic impact assessment and risk mitigation to ensure responsible technological development. Biases and discriminations present in human decision-making, often unconsciously, are also reflected in software and Artificial Intelligence systems, contradicting the initial expectation of an impartial and objective digital environment. The recognition that algorithmic biases can reinforce inequalities and threaten human rights makes it essential to implement legal, technological, and ethical regulations to mitigate these issues. Considering that AI operates through data and machine learning, rigorous oversight of the stages of data collection, analysis, and interpretation is indispensable to prevent discrimination based on gender, ethnicity, or social class. The research adopts a qualitative methodology, being descriptive and explanatory in its objectives, with a critical and interdisciplinary approach that combines the analysis of ethical and legal theories with a review of specialized literature. To support the reflections, books, scientific articles, applicable legislation, and official documents are used. In the first chapter, the concept of gender is discussed from the perspective of feminist theories, analyzing how social norms shape symbolic representations and perpetuate discriminatory stereotypes, in addition to exploring the relationship between gender, power, and inequality in contemporary society. The second chapter examines the processes of developing and training AI systems, highlighting intuitive heuristics and their impacts on automated decision-making, with examples of recurring biases and their consequences. The third chapter analyzes the ethical challenges in AI system development, aiming to prevent violations of human dignity and strengthen the Democratic Rule of Law within the context of international regulation. Strategies to mitigate ethical and legal risks associated with AI algorithms are discussed, with an emphasis on algorithmic governance and impact assessments that incorporate a critical gender analysis. Proposed practices include independent audits, robustness testing, and database diversification to minimize discriminatory biases, ensuring that technology promotes equity.