S[ai]nts - Exploring the Use of GPTs for Spiritual Conversation in Catholicism

Using Open.AI’s ChatGPT, I am creating three separate chatbots with three unique specializations and personalities: one St. Francis of Assisi, one St. Thérèse of Lisieux, and one St. Thomas More. I am training these GPTs on information about their respective saint's lives, works, and beliefs using a mix of primary and secondary academic sources. The s[ai]nts will be made accessible through a web app. The user will engage in conversation with the chatbot and will hopefully find the experience meaningful to their religious experience. My paper will detail the development of the s[ai]nts, investigation of their reception within religious communities, and explanation of results of this project.

One of the problems for such a project is that the information a large language model has access to may lack the specialization of a specific saint’s life and will be dependent on the saint’s popularity. An LLM may be aware of a saint’s general biography, but most likely not primary sources regarding any writings or specific sayings. Additionally, while ChatGPT attempts to avoid controversial subjects, intentional prompting can reveal biased perspectives. As Robert Geraci showed on his blog, ChatGPT has a theory of religion that excludes non-Western religions as religions and prioritizes Western religions. Though I am working with Catholicism (a Western religion), this is important to note regarding lack of access to information and limitations concerning data collection. Thus, any attempt to use ChatGPT to discuss saints with members of a religious community must have both academic articles and primary source material, not just drawing from the results of a general ingestion of the internet. While good information is essential to an authentic conversation, the information is useless if the user finds the conversation boring. For the user to be engaged, the AI must have interesting, dynamic, explorative responses. ChatGPT has this ability, frequently responding with more words and enthusiasm than necessary. However, just like in person-to-person interactions, sometimes conversations can fade and come to a dead end. To minimize this possibility I needed to develop prompts that would encourage consistent conversation, keep the conversation on topic, and make the user’s experience require minimal effort other than responding to an AI that was enthusiastically ready to converse about its life.

Next, this project will involve a survey where users describe their experience with the various GPTs, outlining what they liked, disliked, found comforting, found disturbing, and more immediately after usage. I plan to utilize focus groups to gather data over time from a parish in Boone, North Carolina, and a parish in Durham, North Carolina. I plan to utilize the youth groups of each community as well as the campus catholic group at a local public university. I will examine differences in engagement and explore correlations with general political, theological, and social perspectives in general. I suspect in more liberal churches there may be more openness towards AI and contrarily, in more conservative communities, more resistance to the use of AI in religion and wariness towards the chatbots. My goal is, through a combination of surveys and focus groups of participants who use the chatbots and engage in consistent traditional prayer, to try and gauge their respective enjoyment and understand participants’ preferences and their general feelings about AI in Religion.

I believe my experiment will allow a glimpse into just how much of an impact AI could have on religious communities and how responsive parishes may be to the incorporation of AI into their worship practices. Saint worship and intercessional prayer are personal practices for many parishioners; this project may provide more insight and understanding into how saints intimately function in believers' lives. I hope to gather responses from participants, recognize patterns regarding group members' interactions and responses, and receive feedback on how to improve the chatbots to cater to Catholic believers. Regarding potential future research, there are saintly types in many other religious traditions, such as Islam, Buddhism, and Hinduism. This project is a prototype and trial run of the idea before possibly extending the model to other religious communities. I believe that if users find value in these interactions, this form of AI could be an avenue for church communities' recruitment of younger parishioners and encouragement of forward progression for the church.