Program (Tentative)

9:00-9:05 (CST) ; 10:00-10:05 (JST) ;

Opening Remarks by prof. Takayuki Ito

9:05-9:035 (CST); 10:05-10:35 (JST) ;

Graph Neural Networks for Argument Structure Extraction

Shiyao Ding (Graduate School of Informatics, Kyoto University)

9:35-10:05 (CST) ; 10:35-11:05 (JST) ;

Governance by Simulation

Rafik Hadfi (Graduate School of Informatics, Kyoto University)

10:05-10:35 (CST); 11:05-11:35 (JST) ;

A Compromising Strategy based on Constraint Relaxation Automated Negotiating Agents closer to human consensus

Shun Okuhara, (Faculty of Engineering, Graduate School of Engineering, Mie University, Japan)

10:35-12:30 (CST) ; 11:35-13:30 (JST) Lunch Break

12:30-13:00 (CST) ; 13:30-14:00 (JST)

Formalisation of SNS Argumentation in Block Argumentation

Ryuta Arisaka, (Graduate School of Informatics, Kyoto University)

13:00-13:30 (CST) ; 14:00-14:30 (JST)

Influencer detection based online forum facilitation support

Wen Gu, (Japan Advanced Institute of Science and Technology (JAIST))

13:30-14:00 (CST) ; 14:30-15:00 (JST)

The impact of AI-based civic platform on facilitating crowd intelligence for social good: Experimental democratic evidence from Afghanistan

Jawad Haqbeen, (Graduate School of Informatics, Kyoto University)

14:00-14:30 (CST) ;15:00-15:30 (JST) Break

14:30-15:00 (CST) ; 15:30-16:00 (JST)

Towards Hyperdemocracy: Multiple Agents that Facilitate Crowd Discussion

Takayuki Ito (Kyoto University)

15:00-15:30 (CST) ; 16:00-16:30 (JST)

How Can We Utilize Large-scale Language Models for Facilitating Consensus-Building?

Shun SHIRAMATSU, (Nagoya Institute of Technology)

15:30-16:00 (JST) ; 16:30-17:00 (JST)

Opinion Changes and Discourse Quality in Deliberative Polls

Takashi Nakazawa, (Department of Sociology, Toyo University, Japan)

16:00-16:30 (CST) ; 17:00-17:30 (JST)

Effects of Social Influence on Engaging the Crowds in Web-based Idea Contest

Sofia Sahab (Graduate School of Informatics, Kyoto University)

16:30-17:00 (CST) ; 17:30-18:00 (JST)

Listening Unintended Action and Supporting AI-Democracy

Tokuro Matsuo (Advanced Institute of Industrial Technology, Japan)

Invited Speakers, Titles, and Abstracts (Tentative)

Title:

How Can We Utilize Large-scale Language Models for Facilitating Consensus-Building?

Shun SHIRAMATSU, Nagoya Institute of Technology

Abstract:

Recently, large-scale language models such as GPT-3 has rapidly evolved. Such language models have a great potential to be applied to facilitating consensus-building, e.g., generating facilitator-like questions, providing related information, providing lacking viewpoints, summarizing discussion, etc. In this talk, I would introduce our GPT-3-based ongoing trials toward facilitating mutual understanding and building consensus without social division.

Opinion Changes and Discourse Quality in Deliberative Polls

Takashi Nakazawa, Department of Sociology, Toyo University, Japan

Abstract:

Mini-publics have been attracting much attention as a practice of deliberative democracy. Mini-publics are a method of creating a microcosm of citizens by selecting participants through random sampling and using the results of their deliberations for policy formation and political decision-making. Among various methods of mini-publics, deliberative polls, which intend to form “refined”, rather than “raw”, public opinion through deliberation, has been practiced in various countries and regions.

The present study examines the relationship between opinion changes and discussion quality in deliberative polls. Deliberative democracy emphasizes changes in the preference of participants through rational deliberation. In deliberative polls, the changes in participants’ preference through deliberation are often used as an indicator to evaluate its success. However, it has been understudied if there really is a link between opinion changes and the quality of deliberation. In this study, the author evaluated the quality of discussion, and conducted an exploratory analysis of its relationship with preference changes.

The data used was 4,178 statements from a mini-public deliberation event that mimicked a deliberative poll, conducted in 2019 on the topic of the restart of Hamaoka nuclear power plants and local agreement. Two coders coded the discussion data for each of the eight groups with an improved version of the Discourse Quality Index as a measure of discussion quality. Of the 30 coded items, 14 items with Kappa coefficient >0.58 were selected and scored, and their relationship to the changes in participants’ attitudes before and after the discussion was examined.

As a result, it became clear that although there seems to be a relationship between the quality of discussion and opinion change, it is not as simple as good debate leading to opinion changes as assumed by deliberative polls. It was also suggested that what influences opinion changes may not only be "rational argumentation" as assumed by Habermasian theory of deliberative democracy, but also “experiences” and “emotions”.

- Title: A Compromising Strategy based on Constraint Relaxation Automated Negotiating Agents closer to human consensus

- Name & Affiliation: Shun Okuhara & Faculty of Engineering, Graduate School of Engineering, Mie University, Japan

- Abstract

This talk presents a compromising strategy based on constraint relaxation for automated negotiating agents in the nonlinear utility domain. Automated negotiating agents have been studied widely and are one of the key technologies for a future society in which multiple heterogeneous agents act collaboratively and competitively in order to help humans perform daily activities. A pressing issue is that most of the proposed negotiating agents utilize an ad-hoc compromising process, in which they basically just adjust/reduce a threshold to forcibly accept their opponents offers. Because the threshold is just reduced and the agent just accepts the offer since the value is more than the threshold, it is very difficult to show how and what the agent conceded even after an agreement has been reached. To address this issue, we describe an explainable concession process using a constraint relaxation process. In this process, an agent changes its belief by relaxing constraints, i.e., removing constraints, so that it can accept it is the opponents offer. We also propose three types of compromising strategies. Experimental results demonstrate that these strategies are efficient.

Title: Formalisation of SNS Argumentation in Block Argumentation

Name: Ryuta Arisaka

Affiliation: Kyoto University

Abstract: In this talk, I will motivate Block Argumentation as a formal model of argumentation on SNS. Block argumentation is a formal argumentation framework that adopts the perspective that an argument is an argumentation and vice versa, which helps model contexts and references. I will take examples from online discussion forums to illustrate the real-life use of both contexts and references, and show how they are intuitively modelled in block argumentation. I will then explain the benefit of formally modelling argumentations on SNS by highlighting its connectivity to argumentation mining.

Title: Influencer detection based online forum facilitation support

Name & Affiliation: Wen Gu & Japan Advanced Institute of Science and Technology (JAIST)

Abstract: With the development of the automated facilitation support for online forum, influential user detection becomes a critical issue for supporting human facilitator. Influential maximization (IM) aiming at choosing a set of users that maximize the influence propagation from the entire social network users is one of the key approaches to detect influential users in online social network. However, conventional IM algorithms cannot be applied to online forum because of the lack of existing social network. In this talk, we introduce a novel IM-based approach to detect influential users in online forum. The online forum influence propagation network (OFIPN) is modeled with the consideration of both individual contribution and relevance between users, and a heuristic algorithm that aims to find influential users in OFIPN is proposed. In addition, detected influential users based online facilitation support will be also discussed.

- Title

The impact of AI-based civic platform on facilitating crowd intelligence for social good: Experimental democratic evidence from Afghanistan

- Name & Affiliation

Jawad Haqbeen, Kyoto University

- Abstract

With the advance of the internet, digital platforms have become the next generation democratic social future research focus. Many researchers have invested heavily in this area by developing and inventing an innovative digital platform to harness crowd wisdom. These platforms employ Artificial Intelligence (AI) technology and issue-based information models to improve the collective civic discussion. This talk presents an online civic platform called D-Agree that uses machine intelligence to facilitate and deliver crowd intelligence for social good. We deployed the system with the collaboration of local governments in many countries like Japan and Afghanistan to study how AI-based civic platforms can support the facilitating crowd intelligence for democratic comprehensive planning and policy making. In this talk, we present the experimental democratic application of D-Agree in Afghanistan. We believe our approach speaks of actual democracy and potential benefits that will offer an attractive service value for democratic policymaking.

Title:

Governance by Simulation

Name & Affiliation:

Rafik Hadfi (Graduate School of Informatics, Kyoto University)

Abstract:

The world is witnessing a growing number of democracies following the third wave of democratization. While some countries are consolidating their transition to democracy, others are becoming less democratic. This rising autocratization is affecting both western and eastern democracies. Freedom of expression is constantly under attack, and the public sphere is subject to toxic polarization. Citizens are becoming more and more disillusioned with their institutions and leaders. This dissatisfaction is resulting in declining electoral turnouts and increasing populist discourses. Incidentally, social media are playing a transformational role and leading to social changes that could disturb democracy by exacerbating socioeconomic inequality and cultural heterogeneity. These signs suggest that the contemporary forms of democratic governance face globalization and mediatization challenges. In this talk, I will depart from the argument that we must thoroughly examine any form of governance before providing any solution to overcome political problems. Then, I will elaborate on the example of computer simulation of safety-critical systems to show that the simulation of organizational processes could lead to controlled outcomes before their enactment into society. Next, I will establish the foundations of an algocratic form of governance that uses simulations at its core. To elaborate on this paradigm, I will build on the recent advances in social choice theory, distributed artificial intelligence, and social simulations. Finally, I will identify a set of factors that make algocracy a distinctive mode of governance and explore its impact on the future of democracy.

- Title：Graph Neural Networks for Argument Structure Extraction

- Name & Affiliation : Shiyao Ding & Kyoto University

- Abstract :

Argument structure extraction is a fundamental problem in argumentation mining which aims at automatically constructing arguments from unstructured textual documents. It has been applied in many fields such as online discussions and deliberative democracy. Regarding each argument as a node, then classifying the node labels or predicting relationships of those nodes are important tasks in argument structure extraction, which are called node classification and link prediction tasks. Graph neural networks (GNN) as an efficient approach to cope with graphical data have shown good performances on such kinds of tasks.

In this tutorial, I will first introduce the basic knowledge of argument structure extraction problems and GNN methods, the necessity of using GNN methods in solving argument structure extraction, and the recent results of GNN-based solutions. Then, I will analyze both the advantages and limitations while applying GNN methods for solving argument structure extraction. At last, I will discuss how argument structure extraction can benefit deliberative democracy.

Title: Effects of Social Influence on Engaging the Crowds in Web-based Idea Contest

Name: Sofia Sahab

Affiliation: Kyoto University

- Abstract : Different variants of incentive mechanisms exist and are the research focus. This talk is about research hypothesized that the author of quality of opinion positively generates social influence on idea generation and active participation. An experimental study was conducted to study the effect of influencing participants on engaging the crowd in a crowdsourcing contest by featuring them through real-time virtual scores based on their activities in the crowdsourcing platform. The results indicate that those who were gaining high scores and were consistently ranked among the top three contest participants by the system significantly increased their group performance in terms of group engagement, communication, and interaction.

Towards Hyperdemocracy: Multiple Agents that Facilitates Crowd Discussion

Professor Takayuki Ito (Kyoto University)

Abstract: Online discussion among civilian is important and essential for next-generation democracy. Providing good support is critical for establishing and maintaining coherent discussions. Large-scale online discussion platforms are receiving great attention as potential next-generation methods for smart democratic citizen platforms. Such platforms require support functions that can efficiently achieve a consensus, reasonably integrate ideas, and discourage flaming. Researchers are developing several crowd-scale discussion platforms and conducting social experiments with a local government. One of these studies employed human facilitators in order to achieve good discussion. However, they clarified the critical problem faced by human facilitators caused by the difficulty of facilitating large-scale online discussions. In this work, we propose an automated facilitation agent to manage crowd-scale online discussions. An automated facilitator agent extracts the discussion structure from the texts posted in discussions by people. In this paper, we present our current implementation of D-agree, a crowd-scale discussion support system based on an automated facilitation agent, which extracts discussion structures from text discussions, analyzes them, and posts facilitation messages. We conducted a large-scale social experiment with Nagoya City’s local government. The results present that our automated facilitation agents succeeded to gather more opinions from people. Also, we found that people satisfactions on both discussions by automated facilitation agents and discussions by human facilitators were almost same score. Our main contribution is that this is one of the earliest real trials, in which an automated agent facilitated discussion among people in the real society.

Title

Listening Unintended Action and Supporting AI-Democracy

Name of Speaker

Tokuro Matsuo, Ph.D. Professor, Advanced Institute of Industrial Technology, Japan

Abstract:

Nowadays, a lot of types of communication system to make consensus among people are provided. We can utilize these kinds of systems, such as Social network system, e-mail, and instant messenger system, to make a decision and determination through online discussion. In the next decade, we can forecast a lot of types of consensus formation systems are provided and we may find new communication systems integrating between cyber and physical environment. In this talk, I introduce a behavior tracker system using wireless tag and sensors. The system provides the density of places, moving history, and similarities of event attendees’ behavior. Also, I also introduce the environment to provide useful information for event attendees by the digital signage system in the conference venue. This digital signage system is connected to the attendees location capture system and conference registration system. These integrations between cyber and physical environments and data enable to make better consensus formation between all sorts of people.