The Relationship between Semantic-similarity-based Cognitive Diversity and Discussion Effectiveness in an Intelligent Discussion System

Hongli Gao, Ran Ye1, Xiaohan Jiang

Cognitive diversity has been manipulated as the wide range of semantic categories (Nijstad, Stroebe, & Lodewijkx, 2002), low related categories (Baruah & Paulus, 2016) and so on. However, these kinds of cognitive diversity were determined before group discussion and would not change dynamically during the discussion, which could be called static cognitive diversity. Compared to this, adaptive cognitive diversity was defined as the adaptive difference with the previous semantic domain in group discussion. For manipulation, stimuli which were different from the previous dynamic semantic domain were given after automatically identifying the previous semantic domain in the discussion (Gao, Yang, Xu, & Hu, 2019). Results showed that compared with the homogeneous condition, discussion breadth of the participants increased under the diverse condition, participants subjectively considered that the opinions provided by virtual agents were more helpful and the understanding on the discussion questions was more comprehensive under the diverse condition.