[1L3-GS-9a] Human interface

2:25 PM - 2:40 PM JST(5:25 AM - 5:40 AM UTC)

[1L3-GS-9a-04] Analysis of the Impact of Large Language Model Presentations on Audience Affective Evaluation in Bibliobattle

Yusuke Sasai¹, 〇Momoha Hirose², Shoichi Hasegawa¹, Kenta Oku³, Tadahiro Taniguchi^1,2 (1. Ritsumeikan University, 2. Kyoto University, 3. Ryukoku University)

Keywords:

Large Language Models,Bibliobattle,Affective Evaluation

Bibliobattle is a casual book review game widely practiced in Japan at libraries and educational institutions. While various functions of Bibliobattle have been substantiated by previous research, it remains unclear to what extent these effects depend on the presenter. In this study, we introduced Large Language Models (LLMs)—which are qualitatively different from humans—as presenters to examine whether audience affective evaluations differ between human and LLM presenters. We conducted a Bibliobattle with both human and LLM participants and surveyed audience evaluations under two conditions: face-to-face and online. The results indicated no significant differences in metrics such as "motivation to read" or "enthusiasm" based on the type of presenter. Conversely, while evaluations of human presenters varied significantly between face-to-face and online settings, evaluations of LLM presenters showed no significant differences across the experimental environments.