[4R2-OS-22a] グループインタラクションとAI

[4R2-OS-22a-04] Discussion Visualization and Feedback to the User based on Persuasiveness Estimation

Atsushi Ito¹, 〇Yukiko Nakano¹, Fumio Nihei², Tatsuya Sakato¹, Ryo Ishii², Atsushi Fukayama², Takao Nakamura² (1. Seikei University, 2. NTT Human Informatics Laboratories)

Keywords:persuasiveness, Group Discussion, feedback

Persuasiveness is one of the important skills for smooth communication. This study proposes a machine learning model that estimates participants' persuasiveness in group discussions, and implements a meeting browser that visualizes persuasiveness as well as plays discussion videos. We added a function to visualize the level of persuasiveness estimated by the persuasiveness estimation model. We conducted an experiment to examine whether our meeting browser can grasp the contents of the entire discussion in a shorter time compared to a browser without the visualization function. We found the difference in the degree of accomplishment of the task between with and without the persuasiveness visualization function. Furthermore, we will discuss a method of giving feedback to the user. Since it is difficult to understand the features used in the estimation model, we extract features that people can understand, and propose a feedback method using the features with statistical significance.