Banca de DEFESA: João Sérgio dos Santos Soares Pereira

Uma banca de DEFESA de DOUTORADO foi cadastrada pelo programa.
STUDENT : João Sérgio dos Santos Soares Pereira
DATE: 09/07/2025
TIME: 09:00
LOCAL: https://teams.microsoft.com/l/meetup-join/19%3ameeting_NDc2MmZlNDQtMTdmYy00N2EyLTlhMzYtZDA2NTQzMzYyO
TITLE:

CUSTOMIZATION OF DESIGNS FOR DISPUTE RESOLUTION, BASED ON GENERATIVE TECHNOLOGICAL APPLICATIONS AND THE USE OF LLM'S (LARGE LANGUAGE MODELS) IN SUPPORT OF JUDICIAL DECISIONS: Case study from the perspective of developers, collaborators, and internal users of the Judiciary

KEY WORDS:

Generative Artificial Intelligence; Judiciary; Large Language Models; Decision Support Systems; AI Governance.

PAGES: 1300
BIG AREA: Ciências Sociais Aplicadas
AREA: Direito
SUMMARY:

This thesis investigated the application of Generative Artificial Intelligence (GAI), especially Large Language Models (LLMs), in supporting judicial decision-making, proposing an operational governance structure that integrates ethical and normative guidelines, theoretical foundations of law, and empirical field data. The research adopted a mixed methodology: qualitative (interviews with developers, internal users, and institutional collaborators) and quantitative (structured questionnaire for civil servants working in court offices and magistrates), organizing the findings around seven units of empirical analysis: familiarity, frequency of use, purposes, perception of risks and measures to mitigate them, transparency, quality of results, and training. It started from the observation that the digital transformation of the judiciary is not only technical, but also cultural and epistemic, implying changes in the structure of jurisdiction, in the standards of reasoning, and in the interactions between humans and machines. The thesis was structured in six chapters. The first presented the judicial process and decisions (including binding decision-making standards) from the perspective of interpretive rationality and legal language, articulated with a theoretical-normative triad based on Robert Alexy (validation), Ronald Dworkin (integrity), and Jürgen Habermas (quality of reasoning), refounded for the contemporary times of hyperconnectivity. The second dealt with the digital transformation of the judiciary and the emergence of so-called hybrid courts. The third chapter analyzed national and international ethical and normative guidelines, presenting twelve empirically tested hypotheses derived from conceptual bases previously formulated by the author. The fourth detailed the empirical methodology of the research. The fifth described the data collection and analysis process, with the concrete limitations encountered. The sixth presented the proposed operational framework, with practical recommendations for the safe and legitimate use of AI in courts. As a limitation, the study faced institutional obstacles to quantitative collection in some courts and difficulties in accessing key subjects in early-stage AI projects. The diversity of realities required methodological adaptations and reinforced the choice of qualitative analysis as the main axis of depth. The interviews were organized based on a matrix of convergences, similarities, and discursive divergences, respecting the local specificities of each court. The final model resulted in a proposal for a practical checklist aimed at three groups (internal users, institutions, and collaborating companies), with a view to ensuring the ethical, safe, and legally appropriate use of IAG in courts. Among the main partial conclusions, it was observed that there are significant institutional inequalities in the digital maturity of courts and in the levels of adherence to IAG solutions, with low standardization in development processes, a lack of systematic training, and overreliance on automated outputs without critical review. In contrast, good practices were identified, such as layered review, prompt engineering with discursive safeguards, on-site training strategies, and the formation of interdisciplinary teams. The importance of communicational engagement, internal transparency, and shared governance with collaborating companies was also revealed. The research reinforced that the use of these technologies should not replace the human deliberative process, but act as its Socratic opponent, promoting rationality, critical review, and responsible legal reasoning.

COMMITTEE MEMBERS:
Externo à Instituição - ANDERSON DE PAIVA GABRIEL - UERJ
Externo à Instituição - DIERLE JOSÉ COELHO NUNES - PUCMINAS
Presidente - 2250715 - FABIANO HARTMANN PEIXOTO
Interna - 1150035 - FERNANDA DE CARVALHO LAGE
Externo à Instituição - SERGIO RODRIGO DE PADUA - UNIBRASIL