Transparência pela cooperação

Como a regulação responsiva pode auxiliar na promoção de sistemas de machine-learning inteligíveis

Autores

Palavras-chave:

regulacao responsiva, inteligencia artificial, vies discriminatorio, transparencia, explicabilidade

Resumo

[Propósito] Analisar a aplicabilidade da teoria da regulação responsiva para promoção da inteligibilidade de sistemas de machinelearning sob o enfoque da Lei Geral de Proteção de Dados.

[Metodologia/abordagem/design] Este artigo tem a teoria da regulação responsiva como marco teórico e se baseará, inicialmente, em uma análise comparada da LGPD e do RGPD para identificar como essa teoria pode auxiliar reguladores brasileiros, mais especificamente a Autoridade Nacional de Proteção de Dados, a abordar a inteligibilidade de sistemas de inteligência artificial.

[Resultados] A partir de uma análise comparativa prévia entre como LGPD e RGPD lidam com o tema da explicabilidade de sistemas de decisão automatizada (incluso machine-learning), identificou-se que o racional de cooperação entre regulador e regulado, um sistema de governança em rede e a existência de uma pirâmide regulatória permitem a aplicação da teoria da regulação responsiva para a promoção da inteligibilidade desses sistemas.

[Implicações práticas] Sistemas de IA têm sido frequentemente acusados de possuírem vieses discriminatórios. Isso faz com que pessoas negras sejam mais frequentemente identificadas do que brancas por tecnologias de reconhecimento facial ou que afrodescendentes tenham menor chance de conseguir crédito, potencializando o abismo racial no Brasil. Garantir que a tecnologia seja compreensível para humanos identificarem melhor como endereçar essas falhas é primordial para promover o uso de sistemas mais justos. O presente estudo pretende, por meio da identificação das estratégias regulatórias mais adequadas a lidar com a opacidade algorítmica, auxiliar reguladores a endereçar a discriminação promovida por esses sistemas.

Downloads

Não há dados estatísticos.

Biografia do Autor

José Renato Laranjeira de Pereira, Universidade de Brasília

Diretor do Laboratório de Políticas Públicas e Internet - LAPIN, Mestrando em Direito Regulatório pela Universidade de Brasília - UnB e Bacharel em Direito pela UnB com intercâmbio na Università degli Studi di Roma Tre. É German Chancellor Fellow da turma 2021-2022 pela Fundação Alexander von Humboldt.

Referências

ANTUNES LIMA DA FONSECA CARVALHO, J. P. The Legal Status of the National Data Protection Authority in light of the Regulatory State Theory: Is there any room for the adoption of the material concept of administrative decentralization in Brazil?. Law, State and Telecommunications Review, [S. l.], v. 12, n. 2, p. 118–132, 2020. DOI: 10.26512/lstr.v12i2.34714. Disponível em: https://periodicos.unb.br/index.php/RDET/article/view/34714. Acesso em: 4 apr. 2021.

ARANHA, M. I. Manual de Direito Regulatório: Fundamentos do Direito Regulatório, 5a ed. rev. ampl, London: Laccademia Publishing, 2019.

Article 29 Working Party (A29WP). Guidelines on Automated individual decision-making and Profiling for the purposes of Regulation 2016/679, Adopted on 3 October 2017, As last Revised and Adopted on 6 February 2018, WP251rev.01, 29.

ARYA, V. et al. One Explanation Does Not Fit All: A Toolkit and Taxonomy of AI Explainability Techniques, 2019, p. 1. Available at: https://arxiv.org/abs/1909.03012. Accessed on 27 August 2020.

AYRES, Ian; BRAITHWAITE, John. Responsive Regulation: Transcending the Deregulation Debate. Oxford University Press, USA, 1992.

BAYAMLIOGLU, E. Contesting Automated Decisions: A View of Transparency Implications. European Data Protection Law Review, Volume 4, Issue 4, 2018, pp. 433 - 446. Available at: doi: https://doi.org/10.21552/edpl/2018/4/6. Accessed on 27 August 2020.

BILGIC, M. & Mooney, R.J. Explaining recommendations: satisfaction versus promotion. Beyond Personalization Workshop, IUI, vol. 5, 153, 2005.

BLACK, Julia & MURRAY, Andrew. Regulating AI and Machine Learning: Setting the Regulatory Agenda. European Journal of Law and Technology, Vol 10, Issue 3, 2019.

BOZDAG, E. Bias in algorithmic filtering and personalization. Ethics Inf Technol, n. 15, 23 Jun 2013. Available at: DOI: 10.1007/s10676-013-9321-6. Accessed on 13 December 2020.

BRAITHWAITE, J. To Punish or Persuade: Enforcement of Coal Mine Safety. Albany: State University of New York Press, 1985.

BRAITHWAITE, John. Responsive Regulation and Developing Economies. World Development, v. 34, n. 5, p. 884 – 898, 2006.

BUCHER, T. If… then: algorithmic power and politics. Oxford University Press, New York, 1st edition, 2018.

CARVALHO, D.V.; PEREIRA, E.M.; CARDOSO, J.S. Machine Learning Interpretability: A Survey on Methods and Metrics. Electronics, n. 8, 832, 2019.

COLLINGRIDGE, David. The social control of technology. Pinter, 1st ed., 1980.

COWAN, N. The magical mystery four: How is working memory capacity limited, and why? Curr. Dir. Psychol. Sci., n. 19, 2010, pp. 51–57.

DOSHI-VELEZ, F. et al.. Accountability of AI Under the Law: The Role of Explanation. Berkman Klein Center Working Group on Explanation and the Law, Berkman Klein Center for Internet & Society working paper, 2017. Available at nrs.harvard.edu/urn-3:HUL.InstRepos:34372584. Accessed on 11 December 2020.

DOSHI-VELEZ, F. & KIM, B. Considerations for Evaluation and Generalization Interpretable Machine Learning. Explainable and Interpretable Models in Computer Vision and Machine Learning, Springer: Berlin, Germany, 2018; pp. 3–17.

DOSHI-VELEZ, F. & KIM, B. Introduction to Interpretable Machine Learning. Proceedings of the CVPR 2018 Tutorial on Interpretable Machine Learning for Computer Vision, Salt Lake City, UT, USA, 18 June 2018.

DOSHI-VELEZ, F. & KIM, B. Towards a rigorous science of interpretable machine learning. arXiv 2017, arXiv:1702.08608.

FRIEDMAN, David. Does technology require new law. Harvard Journal of Law & Public Policy, v. 25, p. 71, 2001. Available at https://digitalcommons.law.scu.edu/facpubs/22/. Accessed on 7 December 2020.

GARCIA, Renata Cavalcanti de Carvalho. Proteção de dados pessoais no Brasil: Uma análise da Lei nº 13.709/2018 sob a perspectiva da Teoria da Regulação Responsiva. Journal of Law and Regulation, [S. l.], v. 6, n. 2, p. 45–58, 2020. Disponível em: https://periodicos.unb.br/index.php/rdsr/article/view/28490. Acesso em: 4 abr. 2021.

GONÇALVES, M. E. The risk-based approach under the new EU data protection regulation: a critical perspective. Journal of Risk Research, 2019, DOI: 10.1080/13669877.2018.1517381.

GILPIN, H.; BAU D.; YUAN, B. Z.; BAJWA, A.; SPECTER, M.; KAGAL, L. Explaining Explanations: An Overview of Interpretability of Machine Learning. 2019. Available at: arXiv:1806.00069. Accessed on 5 December 2020.

IRAMINA, A. GDPR v. GDPL: Strategic Adoption of the responsiveness approach in the elaboration of Brazil´’s General Data Protection Law and the EU General Data Protection Regulation. Law, State and Telecommunications Review, [S. l.], v. 12, n. 2, p. 91–117, 2020. DOI: 10.26512/lstr.v12i2.34692. Disponível em: https://periodicos.unb.br/index.php/RDET/article/view/34692. Acesso em: 13 may. 2021

KAMINSKY, Margot E. & MALGIERI, Gianclaudio. Algorithmic Impact Assessments under the GDPR: Producing Multi-layered Explanations (September 18, 2019). International Data Privacy Law, 2020, forthcoming., U of Colorado Law Legal Studies Research Paper No. 19-28, Available at: https://ssrn.com/abstract=3456224 or http://dx.doi.org/10.2139/ssrn.3456224.

KOLIEB, Jonathan. When to Punish, When to Persuade and When to Reward: Strengthening Responsive Regulation with the Regulatory Diamond. Monash University Law Review, v. 41, n. 1, p. 136-162, 2015.

MCGOUGH, M. How bad is Sacramento’s air, exactly? Google results appear at odds with reality, some say.” Sacramento Bee. 2018 August 7.

MITTELSTADT, Brent. Auditing for Transparency in Content Personalization Systems. International Journal of Communication 10(2016), 4991–5002. Disponível em https://www.ijoc.org/index.php/ijoc/article/view/6267. Acesso em 11 ago 2020.

MOLNAR, C. Interpretable Machine Learning. 2019. Disponível em: https://christophm.github.io/ interpretable-ml-book/. Accesso em 30 ago 2020.

RUDIN, C. Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Instead. Disponível em: https://arxiv.org/abs/1811.10154. Acesso em 28 ago 2020.

SEARCH ENGINES IN COLOMBIA: LEGAL REVIEW AND STUDY OF THE MUEBLES CAQUETA VS. GOOGLE INC CASE. Law, State and Telecommunications Review, [S. l.], v. 12, n. 2, p. 1–13, 2020. DOI: 10.26512/lstr.v12i2.34688. Disponível em: https://periodicos.unb.br/index.php/RDET/article/view/34688. Acesso em: 4 apr. 2021.

SELBST, Andrew D. & POWLES, Julia. Meaningful Information and the Right to Explanation. International Data Privacy Law, vol. 7(4), 2017, pp. 233-242. Available at SSRN: https://ssrn.com/abstract=3039125. Accessed on 26 August 2020.

STORINO, F.; SENNE, F.; PORTILHO, L.; BARBOSA, A. Unequal Inclusion: An Analysis of the Trajectory of Inequalities in Access, Use and Appropriation of the Internet in Brazil. Law, State and Telecommunications Review, [S. l.], v. 12, n. 2, p. 187–211, 2020. DOI: 10.26512/lstr.v12i2.34718. Disponível em: https://periodicos.unb.br/index.php/RDET/article/view/34718. Acesso em: 4 apr. 2021.

UK Government House of Lords. AI in the UK: Ready, Willing and Able? 2017. Available at: https: //publications.parliament.uk/pa/ld201719/ldselect/ldai/100/10007.htm. Accessed on 21 November 2020.

WACHTER, S; MITTELSTADT, B; FLORIDI, L. Why a Right to Explanation of Automated Decision-Making Does Not Exist in the General Data Protection Regulation.. IDPL, n. 76, 2017.

WEXLER, R. When a Computer Program Keeps You in Jail: How Computers are Harming Criminal Justice”. New York Times. 2017 June 13.

Downloads

Publicado

2021-06-23

Como Citar

LARANJEIRA DE PEREIRA, José Renato. Transparência pela cooperação: Como a regulação responsiva pode auxiliar na promoção de sistemas de machine-learning inteligíveis. Journal of Law and Regulation, [S. l.], v. 7, n. 1, p. 194–223, 2021. Disponível em: https://periodicos.unb.br/index.php/rdsr/article/view/37976. Acesso em: 24 abr. 2024.