Democracy out-of-the-box: analysis of compliance with constitutional principles in tax policies that use Artificial Intelligence

• Autor(a) principal: da Costa Nunes, Erik Henrique
• Data de Publicação: 2024
• Outros Autores: Oliveira, Junior Cezar Bezerra de, Melo, Laura Maria de Queiroz, Feitosa, Carlos Eduardo A., Monteiro, Ingrid Teixeira
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Journal on Interactive Systems
• Texto Completo: https://sol.sbc.org.br/journals/index.php/jis/article/view/3879
• topic: Democracy; AI; Taxation; Law; Systematic Mapping Study

The Democratic State of Law emerged with the objective of improving everyone’s life, restricting the power of tyrants, who, among other arbitrary acts, unfairly taxed the people. While taxes remain crucial for the State’s upkeep, modern rules prevent individuals from enduring excessive burdens. Although new technologies leveraging Artificial Intelligence (AI) hold the promise of enhancing lives, the extent of this improvement raises questions. This study delves into the relationship between individuals and the State, specifically exploring the use of AI in tax-related scenarios. Conducting a comprehensive three-stage investigation, the first stage involved surveying the reasons behind State-imposed limitations on taxation. The second stage identified criteria from computer literature addressing potential AI challenges and strategies for achieving explainability. Additionally, a discussion emphasized the importance of technological models aligning with principles that underpin our society. Then, to understand how actions have been carried out in Public Administration, a Systematic Mapping Study (SMS) was carried out, analyzing works from the ACM Digital Library, SBC OpenLib (SOL) and the ENAP (National School of Public Administration)’s repository (a Gray Literature repository). On the ENAP website, works relating to the Creativity and Innovation Award from the Federal Revenue of Brazil were specifically consulted. In the end, 10 works that met the inclusion criteria were selected. It was concluded that none of them has an AI explainable model, considering the interpretability criteria of models (Rules and logic, Data visualization, and Model documentation). In light of this, it is recommended that studies addressing the use of AI in Public Administration incorporate a dedicated section discussing the explainability of models.