Abstract
The study aims to understand consumer preferences for information about depression in South Korea, using the online Q&A platform Knowledge-iN on Naver for analysis. It explores this through the lens of persuasion, focusing on ethos (credibility), pathos (emotional appeal), and logos (logical appeal). The methodology involves comparing the most upvoted answers to less upvoted ones on the same questions, analyzing the differences in the use of rhetorical principles. The study involves collecting 248 sets of answers (most upvoted and less upvoted) from the platform, with 164 sets used for training and 84 for prediction by a large language model (LLM), specifically ChatGPT 3.5. Researchers initially code each sentence of the answers for rhetorical principles, which then assists in fine-tuning the LLM. The goal is to enhance the model’s ability to identify these rhetorical elements automatically using the LLM. The research will evaluate the effectiveness of fine-tuning on the model’s predictive capabilities and its impact on improving F1 scores compared to the base model. Additionally, it seeks to compare the analysis of content by researchers and the LLM, to determine the model’s accuracy in identifying persuasive elements based on Aristotle’s Rhetoric. The study’s significance lies in its potential to revolutionize efficiency in humanities and social science research by automating the analysis of persuasive content, traditionally a labor-intensive process. It also contributes to advancements in AI text extraction, highlighting the practical applications of LLMs in understanding consumer preferences and effective health message dissemination.
Presenters
Yong Jeong YiAssociate Professor, MetaBioHealth, Sungkyunkwan University, Seoul Teugbyeolsi [Seoul-T'ukpyolshi], South Korea Beom Bae
Associate Professor, Communication Arts, Georgia Southern University, Georgia, United States GyeongCheol Shin
Student, BA, Sungkyunkwan University, Seoul Teugbyeolsi [Seoul-T'ukpyolshi], South Korea Sojeong Bae
Student, Bachelor's degree, Sungkyunkwan University, Seoul Teugbyeolsi [Seoul-T'ukpyolshi], South Korea Hyunwoo Moon
Student, MetaBioHealth, Sungkyunkwan University, Seoul Teugbyeolsi [Seoul-T'ukpyolshi], South Korea June Yoon
Student, Ph.D, Sungkyunkwan University, Gyeonggido [Kyonggi-do], South Korea Sang Hyuk Lee
Student, bachelor's degree, Sungkyunkwan University, Gyeonggido [Kyonggi-do], South Korea
Details
Presentation Type
Paper Presentation in a Themed Session
Theme
KEYWORDS
Persuasion, Large language model, Depression,Health communication, Artificial intelligence
Digital Media
This presenter hasn’t added media.
Request media and follow this presentation.