This thesis examines how Catholic priests understand the integration of large language models (LLMs) into homiletic preparation, how they assess whether homiletic texts are human-authored or LLM-generated, how they understand the possibility of the Holy Spirit acting in the context of LLMs, and according to which criteria they define the boundaries of their legitimate use. To analyse this phenomenon, the thesis introduces the concept of computational homiletics. Based on eleven semi-structured interviews incorporating an original adaptation of the Turing test, the thesis finds that the distinction between priest-authored and LLM-generated homilies is not always discernible, with prior experience using LLMs emerging as a key factor in the accuracy of such judgements. Most interviewees do not rule out the possibility of the Holy Spirit acting in the context of LLM use, but make the legitimacy of their use conditional on the priest understanding, critically assessing, and internalising the generated material and taking responsibility for what is preached. Computational homiletics thus allows for the delegation of production, but not of responsibility for preaching.
|
