Effectiveness of Deliberation Using AI-Assisted Online Platform: Comparative study of AI-assisted online versus in-person Deliberative Poll in Japan

The Center for Deliberative Democracy at Stanford University pioneered the Stanford Online Deliberation Platform, which is an AI-assisted automated moderator deliberative platform. This paper introduces the development and features of the Stanford Online Deliberation Platform. Then, to examine the effectiveness of this platform, this paper compares one of the first Deliberative Polls held on the platform, 2020 Tokyo City deliberation on solar energy, and an in-person deliberation in Japan held in 2012. This paper focuses comparing the quality of deliberation in the online and in-person deliberative projects. In the past decades, deliberative democracy scholars empirically analysed the effects of deliberation among citizens using survey data from public deliberation events. However, these studies focused on survey results from pre and post deliberation and did little to investigate what participants actually discussed at events, even though small group discussions among citizens play a key role in the deliberation process. For this paper, we conduct content analysis using transcripts from two deliberative polls in Japan in 2012 and 2020. In 2012, randomly selected Japanese residents discussed future national energy policy face-to-face. Meanwhile, residents in Tokyo deliberated solar energy plans for 2030 via Stanford Online Deliberation Platform featuring an automated moderator in 2020. This rich text data allows us to analyse not only individual but also group attitudes toward energy policy. And, furthermore, allows us to compare if there are differences, if any, in the quality of deliberation between online and in-person deliberations.