%0 Conference Paper %B IJCAI 2021 Workshop on AI for Social Good %D 2021 %T AI-based Mediation Improves Opinion Solicitation in a Large-scale Online Discussion: Experimental evidence from Kabul Municipality %A Jawad Haqbeen %A Takayuki Ito %A Sofia Sahab %X We present a large-scale case study using agent plat- form that facilitated and gathered public opinions on internet-based town discussion. The hypothesis set to test how agent-mediated argumentative messages leads the discussion structure in a “Issue-giving” and “Issue-solving” themes involving human precipitants. The agent facilitation’s mechanism set to dynamically react to participants by moderating and supporting on the bases of “issue-solving” stance in both discussion types. We conducted two large- scale experiments to evaluate the influence of agent mediation while looking at elements of both discus- sion themes. The first experiment themed as a “is- sue-giving” with 188 participants, and the second experiment set as a “issue-solving” with 1076 citizens from Afghanistan. The goal of the first experiment is to contribute insights about the scale of the issues the residents facing at districts 1 and 2. The goal of second experiment is to contribute insights about the scale of issues and their solutions. In the first experiment, we found that the due to participants started by taking part with theme stance “is- sue-giving” the first post of submitters were issues, hence the themed “issue-giving” increased the number of issues but when agent started posting facilitation messages, the participants stance changed from issue-giving to issue-solving stance, while in second experiment the participants stance remain the same as the theme type. %B IJCAI 2021 Workshop on AI for Social Good %G eng