Collective intelligence in Massive Online Dialogues

Collective intelligence refers to the intelligence that emerges from local interactions among individual people. These local interactions in human collectives can be either direct or indirect. In addition, the interactions of both forms can take place through verbal communication or non-verbal action. As illustrated in Figure 1, following these two binary distinctions, interactions among humans can be classified into four groups. Meanwhile, the emergence and ongoing development of Web 2.0 technologies have enabled advanced forms of collective intelligence, allowing large numbers of individuals to act collectively and create high quality intellectual artifacts. To understand the current status in this research and development area, we conducted a systematic review of the studies that used automatic tools to analyze the large-scale Web-based platforms and identified the research gaps. We focused on the platforms in which intelligence may arise from dialogue exchange and discussion among participants (left upper quadrant in Figure 1). Therefore, online reviews, microblogs, and news headlines are excluded since these text units are independent pieces of writings. Chat and email tools are also excluded because they are primarily designed to support small-scale interactions. To classify the existing literature, the McGrath’s framework of group studies [McGrath 1984] is used. Although this framework has been originally intended to address relatively small groups, it is not merely a model about small group activities, but a framework for studying groups systematically. Hence, here, it is applied to research on larger online collectives. In this framework, group interaction process is considered the essence of a group and it refers to “patterned relations among the behaviors of individuals” [McGrath 1984]. The framework encompasses several other elements that are member’s properties, group structure, environment, and tasks. In addition, McGrath describes interaction processes at a micro level, viewing it in terms of three different stages: communication pattern among interacting people, content of communications, as well as impact of different group interaction processes on each other and on participants. The content itself is then viewed in terms of two different aspects: interpersonal relationship and task performance. Current research on the analysis of collective intelligence commonly treats members’ properties, group structure, environment, and task as the context of the study and then facilitates computational tools to analyze the interaction processes. As such, the prior research is categorized into three main groups: communication pattern analysis, content analysis, and impact analysis. Content analysis is then divided into two categories of research. Due to the different nature of virtual spaces, interpersonal relationships component is seen at a higher level, focusing on the studies that analyze sentimental aspects of the content. The task performance component is called task/purpose performance since such settings might be less task-oriented.