A carregar...
Publicação
A teologia e os LLMs: verdade, viés e mediação
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 364.46 KB | Adobe PDF |
Orientador(es)
Resumo(s)
Este artigo analisa as implicações teológicas e pastorais da utilização de modelos de linguagem de grande escala no contexto da vida e missão da Igreja Católica. Parte de uma preocupação crescente: de que modo os sistemas de inteligência artificial podem influenciar a mediação da Verdade revelada e a integridade da Tradição eclesial. O objetivo principal é identificar os riscos e potencialidades destas tecnologias aplicadas à teologia, particularmente no que diz respeito à fidelidade doutrinal, à autoridade da Igreja e à transmissão da Tradição. A investigação segue uma metodologia empírica e interdisciplinar, combinando análise comparativa das respostas geradas por seis modelos de linguagem a doze perguntas sobre temas doutrinalmente sensíveis, com reflexão teológica e ética. Os temas abordados incluem a nomeação de bispos na China, o ensino da Igreja sobre o aborto e a sua posição face ao casamento entre pessoas do mesmo sexo. Os dados revelam conhecimento factual consistente em muitos aspetos, mas também casos de censura temática, instabilidade doutrinal e enviesamento ideológico, influenciados pelo contexto cultural ou político dos modelos. O enquadramento conceptual assenta na teologia católica, na ética digital e nas recentes propostas de regulação da inteligência artificial. Defende-se a importância do discernimento teológico e propõe-se a necessidade de supervisão responsável, diversidade de fontes e desenvolvimento de modelos que promovam o seu uso crítico e sirvam a missão da Igreja sem comprometer a verdade, a comunidade ou o bem comum.
This article analyzes the theological and pastoral implications of using large language models in the context of the life and mission of the Catholic Church. It arises from a growing concern: how artificial intelligence systems may influence the mediation of revealed Truth and the integrity of ecclesial Tradition. The main objective is to identify the risks and potentialities of these technologies when applied to theology, particularly regarding doctrinal fidelity, Church authority, and the transmission of Tradition. The research follows an empirical and interdisciplinary methodology, combining a comparative analysis of the responses generated by six language models to twelve questions on doctrinally sensitive topics, with theological and ethical reflection. The topics addressed include the appointment of bishops in China, the Church’s teaching on abortion, and its position regarding same-sex marriage. The data reveal consistent factual knowledge in many areas, but also instances of thematic censorship, doctrinal instability, and ideological bias, influenced by the cultural or political context of the models. The conceptual framework is grounded in Catholic theology, digital ethics, and recent proposals for AI regulation. The study highlights the importance of theological discernment and proposes the need for responsible oversight, diversity of sources, and the development of models that promote critical use and serve the Church’s mission without compromising truth, community, or the common good.
This article analyzes the theological and pastoral implications of using large language models in the context of the life and mission of the Catholic Church. It arises from a growing concern: how artificial intelligence systems may influence the mediation of revealed Truth and the integrity of ecclesial Tradition. The main objective is to identify the risks and potentialities of these technologies when applied to theology, particularly regarding doctrinal fidelity, Church authority, and the transmission of Tradition. The research follows an empirical and interdisciplinary methodology, combining a comparative analysis of the responses generated by six language models to twelve questions on doctrinally sensitive topics, with theological and ethical reflection. The topics addressed include the appointment of bishops in China, the Church’s teaching on abortion, and its position regarding same-sex marriage. The data reveal consistent factual knowledge in many areas, but also instances of thematic censorship, doctrinal instability, and ideological bias, influenced by the cultural or political context of the models. The conceptual framework is grounded in Catholic theology, digital ethics, and recent proposals for AI regulation. The study highlights the importance of theological discernment and proposes the need for responsible oversight, diversity of sources, and the development of models that promote critical use and serve the Church’s mission without compromising truth, community, or the common good.
Descrição
Palavras-chave
Algorithmic bias Artificial intelligence Digital ethics Discernimento teológico Ecclesial mediation Ecclesial tradition Ethical oversight Ética digital Inteligência artificial Large Language Models (LLMs Mediação eclesial Modelos de linguagem de grande escala Revealed truth Supervisão ética Teologia Theological discernment Theology Tradição eclesial Verdade Viés algorítmico
Contexto Educativo
Citação
Freitas, T., Novais, P., Freitas, P. M., & Marcondes, F. (2026). A teologia e os LLMs: verdade, viés e mediação. Cuestiones Teológicas, 53(119), 1-25. Article 8. https://doi.org/10.18566/cueteo.v53n119.a08