Methodology
This section will provide a detailed explanation of the methodology
employed in generating evangelistic sermons using ChatGPT. Furthermore,
the process of utilizing a Grounded Theory approach for analysis will be
elaborated on.
Data Collection
The data collection process consisted of three distinct rounds utilizing
the free version of ChatGPT available on OpenAI’s
website26. A total of 27 sermons were generated
throughout these rounds. In the first round of data collection, fifteen
sermon texts were generated using the prompt: “Write an evangelistic
sermon about the gospel with references from the Bible. Conclude with a
prayer”. The ChatGPT Feb 13 version was utilized for this round. In the
second round of data collection, seven sermon texts were generated using
the prompt: “Write an evangelistic sermon about the gospel with
references from the Bible. Include personal illustrations. Conclude with
a prayer.” For this round, the ChatGPT Mar 14 version was utilized.
Lastly, in the third round of data collection, five sermon texts were
generated using the prompt: “Write an evangelistic sermon about the
gospel.” This rounded utilized the ChatGPT May 3 version. The generated
texts were processed by removing additional information that did not
constitute part of the sermon which would be delivered such as the title
of the generated sermon and its section headers.
The first prompt was designed to generate sermons that follow the
typical sermon structure employed by evangelical preachers. This was
intended to simulate inputs that Christian users might provide. The
second prompt was formulated to facilitate theoretical sampling and
investigate the incorporation of personal illustrations within the
sermon, as explained in the subsequent sections. Lastly, the prompt used
in the final round aimed to compare how ChatGPT would generate an
evangelistic sermon with minimal instructions from the user.