Algorithmic Credit and Financial Citizenship: institutional parameters for the use of artificial intelligence in credit underwriting

Abstract

Credit underwriting is progressively ceasing to be mediated by registry information and traditional banking history and coming to be mediated by algorithmic inference. The shift is ambivalent. The same informational infrastructure that makes it possible to assess profiles historically illegible to traditional models, and to that extent to broaden access, also expands the capacity to segment, price and exclude. The article argues that the central problem of algorithmic credit does not lie in the use of artificial intelligence as such, but in the absence of institutional structures capable of rendering its decisions auditable, intelligible, contestable and attributable. Methodologically, the study adopts a legaldogmatic and legal-institutional approach, supported by interdisciplinary literature review, analysis of the applicable
Brazilian rules and a functional examination of foreign and multilateral regulatory references. Drawing on the notion of financial citizenship adopted by the Banco Central do Brasil and on the literature on algorithmic citizenship, it proposes shifting the regulatory centre of gravity from explainability to effective contestability, and sets out a framework of six institutional parameters: data governance, model governance and validation, impact assessment and non-discrimination, proportionate transparency, contestability and review, and accountability with meaningful human oversight. It concludes that financial citizenship in algorithmically mediated markets is not measured by the
quantitative expansion of access alone. It depends on the institutional quality of the decisions that define who obtains credit, on what terms and subject to what possibilities of understanding, review and challenge.