Measuring Group Interaction in Online Discussions and Application to Autonomous Agent Deliberation

测量在线讨论中的群体互动及其在自主代理审议中的应用

• 批准号: 20K11936
• 负责人: Rafik Hadfi
• 金额: $ 2.41万
• 依托单位: Kyoto University
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Scientific Research (C)
• 财政年份: 2020
• 资助国家: 日本
• 起止时间: 2020-04-01 至 2024-03-31
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-20K11936/
• 关键词: Agents; Conversational AI; Interaction; Deliberation; NLP; Interdependence; Online Discussion; Decision-making; Online Debates; Agent Deliberation; Predictive Deliberation; Conversational Agents; Augmented Democracy; Mutual Information; Natu

The research resulted in the discovery and evaluation of various patterns of interaction in group discussions involving humans and conversational agents. The patterns where analyzed using a combination of techniques relying on time series analysis, information theory, natural language processing, and natural language generation. The results lead to an enhanced interpretation of the links between online debates and the underlying social dynamic. Moreover, the results helped in improving the implementation of artificial agents that interact with humans in various domains, including linguistic (conversational) and economic (negotiation) ones. These results come as an extension of the findings reported previously. The core findings were presented and published in various national and international conferences, workshops, and tutorials (AAMAS, PRICAI, KICSS, IEEE ICA, etc.). The key achievement of the research is laying out the foundation of platforms that allow humans to work effectively while interacting with conversational agents. This published form of "Augmented Democratic Deliberation" (AAMAS 2022) sets the future of online augmented democracies and addresses in a principled way key questions such as ethics, autonomy, and fairness. These notions will be investigated by the PI in other related research projects that exploit the present findings.

这项研究的结果是发现和评估了涉及人类和会话代理的小组讨论中的各种互动模式。使用依赖于时间序列分析、信息理论、自然语言处理和自然语言生成的技术组合来分析模式。研究结果增强了对在线辩论与潜在社会动态之间联系的解释。此外，研究结果有助于改进在不同领域与人类互动的人工智能体的实现，包括语言（对话）和经济（谈判）领域。这些结果是先前报告的结果的延伸。核心研究结果在各种国内和国际会议、研讨会和教程（AAMAS、PRICAI、KICSS、IEEE ICA等）上发表。这项研究的关键成果是奠定了平台的基础，使人类能够在与会话代理交互时有效地工作。这种出版形式的“增强民主审议”（AAMAS 2022）设定了在线增强民主的未来，并以原则性的方式解决了道德、自治和公平等关键问题。PI将在利用本研究结果的其他相关研究项目中对这些概念进行调查。

项目成果

Recent Advances in Agent-Based Negotiation: Applications and Competition Challenges, Springer

基于代理的谈判的最新进展：应用和竞争挑战，施普林格

• 发表时间: 2023
• 作者: Zongcan Li;Rafik Hadfi and Takayuki Ito
• 通讯作者: Rafik Hadfi and Takayuki Ito

Society 5.0: Challenges, Implementations, and Implications

社会 5.0：挑战、实施和影响

• 发表时间: 2022
• 作者: Rafik Hadfi

Agenda-Based Automated Negotiation Through Utility Decomposition

通过效用分解进行基于议程的自动谈判

• DOI: 10.1007/978-981-99-0561-4_7
• 发表时间: 2023
• 期刊: Recent Advances in Agent-Based Negotiation: Applications and Competition Challenges, Springer
• 作者: Zongcan Li;Rafik Hadfi and Takayuki Ito
• 通讯作者: Rafik Hadfi and Takayuki Ito

Scalable Agent-based Discussion Platforms: Foundations, Development, and Applications

可扩展的基于代理的讨论平台：基础、开发和应用

• 发表时间: 2022
• 作者: Rafik Hadfi;Takayuki Ito;Sofia Sahab
• 通讯作者: Sofia Sahab

Identifying Collaborative Editing Traits and Phases in Good Wikipedia Articles

识别优秀维基百科文章中的协作编辑特征和阶段

• 发表时间: 2021
• 作者: Hui Chen Chou;Rafik Hadfi;Donghui Lin and Takayuki Ito
• 通讯作者: Donghui Lin and Takayuki Ito