BEA - Building Engaging Argumentation

BEA - 构建引人入胜的论点

• 批准号: 455911629
• 负责人: Professorin Dr. Elisabeth André
• 金额: --
• 依托单位: Institut für Nachrichtentechnik (NT)
• 依托单位国家: 德国
• 项目类别: Priority Programmes
• 资助国家: 德国
• 项目状态: 未结题
• 来源: https://gepris.dfg.de/gepris/projekt/455911629?language=en
• 关键词: BEA; Building; Engaging; Argumentation

For humans, a natural way of resolving different points of view or forming an opinion is through conversation, i.e., through the exchange of arguments and knowledge. On the other hand, numerous opinions and information on almost every topic are nowadays also directly available online. However, the screening and evaluation of these multiple sources can be a time-consuming and demanding task. Filter algorithms aim to reduce this effort by taking previous user requests into account to serve the users’ interest. As a result, people tend to focus on a biased subset of sources that repeat or strengthen an already established or convenient opinion (so-called filter bubbles). In order to avoid this (often unconscious) process of intellectual isolation, we herein propose an approach to explore a vast amount of different information in a natural and intuitive way – through a conversation. To this end, we propose a virtual agent that engages in a deliberative dialogue with a human user in order to support a fair and unbiased opinion building process. In contrast to persuasive systems, the envisioned agent has no topic-specific agenda, but instead aims for providing a diverse and representative overview - embedded in a conversation with the user.The proposed system will extend technologies investigated and developed in the scope of the EVA project during Phase I of the SPP Priority Program RATIO. Whereas in the EVA project users assumed the (passive) role of the audience, the system envisioned for Phase II will allow the user to actively participate in the argumentation process through a natural language interface. Therefore, the system will be able to react to user utterances, such as opinions, requests and questions, and provide an individualized conversation over the discussed topic. This will be supported by a transparent presentation of pros and cons towards the discussed topic through the agent’s access to recently developed argument search engines and thus contribute to explainability and deeper understanding.

对于人类来说，解决不同观点或形成意见的自然方式是通过对话，即通过争论和知识的交流。另一方面，现在几乎每个主题的大量意见和信息都可以直接在网上获得。然而，对这些多个来源的筛选和评估可能是一项耗时且艰巨的任务。过滤算法旨在通过考虑以前的用户请求来满足用户的兴趣，从而减少这种工作量。因此，人们倾向于关注有偏见的来源子集，这些来源重复或强化了已经确立的或方便的观点（所谓的过滤气泡）。为了避免这种（通常是无意识的）智力孤立过程，我们在此提出了一种通过对话以自然直观的方式探索大量不同信息的方法。为此，我们提出了一种与人类用户进行深思熟虑的对话的虚拟代理，以支持公平和公正的意见构建过程。与说服系统相比，设想的代理没有特定主题的议程，而是旨在提供多样化且具有代表性的概述 - 嵌入与用户的对话中。拟议的系统将扩展 SPP 优先计划 RATIO 第一阶段期间 EVA 项目范围内调查和开发的技术。在 EVA 项目中，用户扮演了观众的（被动）角色，而第二阶段设想的系统将允许用户通过自然语言界面积极参与论证过程。因此，系统将能够对用户的言论（例如意见、请求和问题）做出反应，并就所讨论的主题提供个性化的对话。通过代理访问最近开发的论点搜索引擎，透明地展示所讨论主题的利弊将支持这一点，从而有助于可解释性和更深入的理解。