Equivocation and Erosion: How LLMs Undermine Catholic Religious Discourse

Notre Dame research students, Matthew Lad and Matthew Karr, have kindly agreed to present a paper “Equivocation and Erosion: How LLMs
Undermine Catholic Religious Discourse” on Thursday 6th November @ 6.00 p.m. in the Library – ahead of presenting the findings of their research to a conference in England. All welcome.

Abstract

Large Language Models (LLMs) offer opportunities for information dissemination,

yet present challenges with upholding the distinct theological practices of the Catholic faith.

By training on vast datasets, LLMs can generate responses that equivocate or blend together diverse perspectives.

While this tendency can be beneficial for providing broad access to information, it can dilute the distinct theological tenets foundational to Catholicism.

While these challenges may affect various faiths, we conduct a case study to investigate them within the Catholic tradition.