ファシリテーション行動の分析によるエージェントの会話参加方法の研究

Discussion is critical to build consensus in our social activities. There are social demands to conduct smooth and effective discussions.  We look at how experienced facilitators control the discussions effectively and smoothly, to uncover the factors which caused the facilitating behaviors of the good facilitator by using measured nonverbal and paralinguistic data [1].  We conducted experiments to obtain the nonverbal and paralinguistic behavior of the participants in face-to-face discussion. Linear discriminant analysis was then applied to the data.  The result suggests that we could achieve 78.4% accuracy in classifying four types of the facilitating behaviors by using seven independent variables. As a result of discriminant analysis, we have found that the facilitator paid attention to not only the context of the discussion but also the individual behavior, and the participants conveyed their requests to the facilitator by using nonverbal behavior.

Reference

[1] Yoshimasa Ohmoto, Yasushi Toda, Kazuhiro Ueda, and Toyoaki Nishida: Analyses of the facilitating behavior by using discussion conditions, participant stances and nonverbal behavior, Social Intelligence Design 2010, 2010.
 
 ディスカッションにおけるファシリテーション行動の分析の研究(図1)は,ディスカッションにおいて上手なファシリテータがどんなファシリテーション行動をするか解明し,予測モデルを構築することを目的にしている.上手なファシリテータを観察して,ファシリテーション行動の要因として,「議論の発散収束」,「参加者の対立同調」,「非言語・パラ言語情報」などを独立変数とし,「発散・不特定」,「収束・不特定」,「対立者」,「具体化」という4種類のファシリテーション行動を従属変数とし,判別分析を行った.この結果,クロスバリデーションで78.6 %の割合で判別できることがわかった.

図1: ファシリテーション行動の分析