Analysis of the Relationship Between Subject Characteristics and Preferred Reactions Using a Dialogue Model Reflecting Tsundere Reaction Statistics
JADH (Japan Association for Digital Humanities)- 期間
- September 2026
- キーワード
- Subject Characteristics, ser Personality Traits, Character Analysis, Dialogue Model, Tsundere
In recent years, attempts to implement specific personality traits in conversational AI systems with the aim of enriching human–computer interaction have attracted considerable attention. The primary objectives of this study are twofold: first, to reproduce the responses of characters with specific personality traits using a mathematical model; and second, to investigate how dynamic adjustment of a character's personality traits via parameters affects users' psychological and evaluative responses. To achieve these objectives, the personality trait known as 'tsundere,' which is highly distinctive and amenable to quantitative data collection. In our previous work, we collected reaction data from well-known tsundere characters, parameterized these data, and extracted statistical features through analysis. Based on these statistical findings, we have already constructed a tsundere reaction inference probabilistic model and a dialogue model. The purpose of the present study is to conduct an evaluation experiment using this dialogue model and to clarify the correlation between user personality traits and preferred tsundere reactions. Through this analysis, we aim to examine how user characteristics influence the perception and evaluation of dialogue AI.
準備中