Quantitative Modeling of Polarization in Online Intelligent Argumentation and Deliberation for Capturing Collective Intelligence

Abstract

Massive online argumentation deliberation has the potential to capture collective intelligence and crowd wisdom. However, certain observable phenomena, such as polarization, often emerge in online argumentation and deliberation, preventing constructive discourse. Accurately detecting the presence and intensity of polarization is important to determining if collective intelligence and crowd wisdom can be captured from online deliberation. In this paper, an innovative method of measuring polarization quantitatively in online argumentation is presented. Important polarization attributes such as homogeneity in groups, heterogeneity across groups, the number of poles, and the size of poles are identified to measure polarization in online argumentation. This new method uses our argumentation tool, Intelligent Cyber Argumentation System's (ICAS) cognitive computing component, a fuzzy logic engine, to derive the participant's agreement distribution. Then we apply an income polarization measurement from the field of economics [7] that we have modified and expanded for argumentation, on the agreement distribution to produce a polarization index value. We discuss why our method to measure argumentation polarization is a significant improvement over existing measurements for online argumentation polarization in terms of these identified attributes. We conducted empirical studies using ICAS that demonstrate that our method outperforms others that exist on our empirical data.