Viabilidade da aplicação de sistemas de inteligência artificial aos atos administrativos discricionários

  • Jaime Paulino Maia e Silva (Student)

Student thesis: Master's Thesis


In recent decades algorithms have evolved exponentially and their implementation in all sectors of activity results simultaneously from a will – a voluntary act – and from an inevitability. However, the more evolved and powerful they become, the more data they can analyze and the more they can work with unstructured databases, the greater the effects resulting from their forecasts and decisions. There are many risks associated with databases that are biased or not representative of society, since they disadvantage risk groups or minorities, causing discrimination, decision-making bias, exponentiated by the scalability and massification of outputs produced by automated systems and not giving guarantees of equity. On the other hand, there are issues associated with algorithmic opacity or the “black-box” effect, in which there is a lack of transparency, explainability and intelligible interpretability by all, but fundamentally by intermediate or end users, of decisions and the reason for them, a right to state reasons. To address this challenge, we propose our own legal-administrative methodology, structured in 4 dimensions, to approach the topic of automating administrative acts by decision levels, according to the degree of complexity and discretion. This methodology is based on a process of collaborative and co-responsible automation, between the public agent and the AI systems, through a model of collaborative human-machine intelligence, supported by the latest in the field of AI - not just algorithms Machine and Deep Learning self-learning, but others that are much more evolved and responsive to the requirements that are imposed in the implementation of these tools in the field of Law and, in particular, Administrative Law, in terms of transparency, explainable AI, causal AI, interactive algorithms , or cognitive computing algorithms.
Date of Award25 Sept 2023
Original languagePortuguese
Awarding Institution
  • Universidade Católica Portuguesa
SupervisorPedro Cerqueira Gomes (Supervisor)


  • Discretionary administrative act
  • Artificial intelligence
  • Causal artificial intelligence
  • Deep learning
  • Machine learning
  • Human-in-the-loop
  • Algorithmic bias
  • Explainability
  • Accountability
  • Discretionary
  • Transparency


  • Mestrado em Direito

Cite this