The algorithm as text: an analysis of textuality mechanisms in an experiment on racial bias in ML-based credit scoring in Brazil

Authors

DOI:

https://doi.org/10.11606/issn.2236-4242.v39i1p150-173

Keywords:

Algorithmic textuality, Performativity, Racial Bias, Machine Learning, Algorithmic discourse

Abstract

In this article, we argue that the algorithm is a text, understood as a communicative event that articulates linguistic, cognitive, and social dimensions. We investigate how algorithmic textuality manifests in a credit scoring system and which discursive and social practices it reveals. The objective is to analyze the performativity of a credit algorithm, highlighting mechanisms of racial exclusion masked by the appearance of technical neutrality. Our theoretical framework is based on three axes: the conception of algorithms as performative texts (Araújo, 2025a), the factors of textuality (Beaugrande; Dressler, 1997), Marcuschi’s (2008) contributions regarding text and Hill’s (2016) concerning the discursive nature of the algorithm. Methodologically, we conducted a textual analysis based on a Python simulation, inspired by Vilarino and Vicente (2020). The results indicate that algorithmic textuality consolidates itself as a socio-cognitive practice that produces interactions and exclusions. We conclude that recognizing the algorithm as a text is fundamental to unveiling racial biases.

Downloads

Download data is not yet available.

Author Biographies

  • Júlio Araújo, Federal University of Ceará

    Doutor em Linguística pela Universidade Federal do Ceará, Brasil (2006). Professor na Universidade Federal do Ceará, Brasil.

  • Lineker Luque, Federal University of Ceará

    Doutorando em Linguística pela Universidade Federal do Ceará, Brasil.

References

ARAÚJO, J.; ARAÚJO, J. C. D. Racismo algorítmico e inteligência artificial: uma análise crítica multimodal. Revista Linguagem em Foco, v. 16, n. 2, p. 89-109, 2024a. Disponível em: https://doi.org/10.46230/lef.v16i2.13108. Acesso em: 9 out. 2025.

ARAÚJO, J. Racismo algorítmico e microagressões nas redes sociais. Domínios de Lingu@gem, Uberlândia, v. 18, e1849, 2024b. Disponível em: https://doi.org/10.14393/DLv18a2024-49. Acesso em: 10 out. 2025.

ARAÚJO, J. O algoritmo é um texto. Texto Livre, Belo Horizonte, v. 18, e58505, 2025a. DOI: https://doi.org/10.1590/1983-3652.2025.58505.

ARAÚJO, J. Necroalgoritmização: notas para definir o racismo algorítmico. Campinas: Mercado das Letras, 2025b.

ARAÚJO, J. Instâncias textuais do racismo algorítmico. Redis: Revista de Estudos do Discurso, [S. l.], n. 17, p. 18-57, 2025c. DOI: https://doi.org/10.21747/21833958/red17a1.

ADIB-MOGHADDAM, A. Racist Artificial Intelligence: Where does it come from? International Studies Journal = ISJ = Faṣlnāmah-i muṭālaʻāt-i bayn al-milalī, v. 20, n. 2, p. 139–159, 2023. DOI: https://doi.org/10.22034/isj.2023.182419.

BALCH, A. Why Algorithms Remain Unjust: power structures surrounding algorithmic activity. Philosophy & Technology, v. 36, n. 3, p. 1-23, 2023. DOI: https://doi.org/10.48550/arXiv.2405.18461.

BEAUGRANDE, R; DRESSLER, W. Introducción a la linguística del texto. Barcelona: Editorial Ariel, 1997.

BENJAMIN, R. Race After Technology: abolitionist tools for the New Jim Code. Medford: Polity Press, 2019, p. 285.

BUCHER, T. If...Then: Algorithmic Power and Politics. New York: Oxford Academic, 2018. DOI: https://doi.org/10.1093/oso/9780190493028.001.0001.

BÜCKER, M.; SZEPANNEK, G.; GOSIEWSKA, A.; BIECEK, P. Transparency, auditability and explainability of machine learning models in credit scoring. Journal of the Operational Research Society, 2020. DOI: https://doi.org/10.48550/arXiv.2009.13384.

FAIRCLOUGH, N. Discurso e mudança social. Trad. Izabel Magalhães. Brasília: Editora Universidade de Brasília, 2001 [1992].

GARCÍA-CANCLINI, N. Cidadãos substituídos por algoritmos. São Paulo: EdUSP, 2021.

GENETTE, G. Palimpsests: Literature in the Second Degree. Trad. Channa Newman; Claude Doubinsky. Lincoln: University of Nebraska Press, 1997 [1982].

HALLIDAY, M.; HASAN, R. Cohesion in English. London: Longman, 1976

HILL, R. What an Algorithm Is. Philosophy & Technology, v. 29, n. 1, p. 35–59, 2016. Disponível em: https://link.springer.com/article/10.1007/s13347-014-0184-5. Acesso em: 26 set. 2025.

HOSTINGER. O que é Python: guia completo da linguagem de programação. Hostinger Tutoriais, 17 dez. 2024. Disponível em: https://www.hostinger.com/br/tutoriais/python-o-que-e. Acesso em: 29 out. 2025.

IBM Brasil. O que é uma árvore de decisão (decision tree)? IBM Think, [2025?]. Disponível em: https://www.ibm.com/br-pt/think/topics/decision-trees. Acesso em: 29 out. 2025.

KAVLAKOGLU, E.; RUSSI, E. What is XGBoost? IBM Think, [2025]. Disponível em: https://www.ibm.com/think/topics/xgboost. Acesso em: 29 out. 2025.

KOCH, I. Introdução à Linguística Textual. São Paulo: Martins Fontes, 2006.

LOYA, J. Racial stratification among latinos in the mortgage market. Race Soc Probl, v. 14, p. 39–52, 2022. DOI: https://doi.org/10.1007/s12552-021-09326-3.

KRISTEVA, J. Sēmeiōtikē: recherches pour une sémanalyse. Paris: Éditions du Seuil, 1969.

MARCUSCHI, L. Produção textual, análise de gêneros e compreensão. 2. ed. São Paulo: Parábola, 2008.

MBEMBE, A. Necropolítica: biopoder, soberania, estado de exceção, política da morte. São Paulo: N-1 edições, 2018.

NOBLE, U. Algoritmos da opressão: como o Google fomenta e lucra com o racismo. Trad. Felipe Damorim. Santo André: Rua do Sabão, 2021.

O’NEIL, C. Algoritmos de destruição em massa: como o Big Data aumenta a desigualdade e ameaça à democracia. Trad. Rafael Abraham. Santo André: Rua do Sabão, 2020.

ONWUEGBUZIA, A. Digital disparities: how Artificial Intelligence can facilitate anti-black racism in the U.S. Healthcare Sector. International Relations and Diplomacy, v. 12, n. 1, p. 40-50, 2024. DOI: https://doi.org/10.17265/2328-2134/2024.01.005.

SILVA, T. Racismo algorítmico: inteligência artificial e discriminação nas redes digitais. São Paulo: Edições SESC, 2022.

STRYKER, C; KAVLAKOGLU, E. O que é inteligência artificial (IA)? IBM Think, 2024. Disponível em: https://www.ibm.com/br-pt/think/topics/artificial-intelligence. Acesso em: 06 out. 2025.

TOLEDO, L. Cliques, tempo em reunião e monitoramento de tela ao vivo: os programas que vigiam funcionários em home office. G1, 19 set. 2025. Disponível em: https://g1.globo.com/trabalho-e-carreira/noticia/2025/09/19/cliques-tempo-em-reuniao-e-monitoramento-de-tela-ao-vivo-os-programas-que-vigiam-funcionarios-em-home-office.ghtml. Acesso em: 08 out. 2025.

VILARINO, R; VICENTE, R. An experiment on the mechanisms of racial bias in ML-based credit scoring in Brazil. arXiv preprint arXiv:2011.09865, 2020. DOI: https://doi.org/10.48550/arXiv.2011.09865.

TU, J; WU, Z. Inherently interpretable machine learning for credit scoring: Optimal classification tree with hyperplane splits. European Journal of Operational Research, v. 322, n. 2, p. 647–664, 2025. DOI: https://doi.org/10.1016/j.ejor.2024.10.046.

Downloads

Published

2026-04-30

Issue

Vol. 39 No. 1 (2026): Digital Discourse – Contemporary Challenges and Problems

Section

Articles

License

Copyright (c) 2026 Júlio Araújo, Lineker Luque

Creative Commons License

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

The Editorial Board authorizes free access to and distribution of published contentes, provided that the source is cited, that is, granding credit to the authors and Linha D'Água and preserving the full text. The author is allowed to place the final version (postprint / editor’s PDF) in an institutional/thematic repositor or personal page (site, blog), immediately after publication, provided that it is available for open access and comes without any embargo period. Full reference should be made to the first publication in Linha D'Água. Access to the paper should at least be aligned with the access the journal offers.

As a legal entity, the University of São Paulo at Ribeirão Preto School of Philosophy, Sciences and Languages owns and holds the copyright deriving from the publication. To use the papers, Paidéia adopts the Creative Commons Licence, CC BY-NC non-commercial attribution. This licence permits access, download, print, share, reuse and distribution of papers, provided that this is for non-commercial use and that the source is cited, giving due authorship credit to Linha D'Água. In these cases, neither authors nor editors need any permission.

Partial reproduction of other publications

Citations of more than 500 words, reproductions of one or more figures, tables or other illustrions should be accompanied by written permission from the copyright owner of the original work with a view to reproduction in Linha D'Água. This permission has to be addressed to the author of the submitted manuscript. Secondarily obtained rights will not be transferred under any circumstance.

How to Cite

ARAÚJO, Júlio; LUQUE, Lineker. The algorithm as text: an analysis of textuality mechanisms in an experiment on racial bias in ML-based credit scoring in Brazil. Linha D’Água, São Paulo, v. 39, n. 1, p. 150–173, 2026. DOI: 10.11606/issn.2236-4242.v39i1p150-173. Disponível em: https://revistas.usp.br/linhadagua/article/view/242277. Acesso em: 9 may. 2026.