Multiparty Interaction Between Humans and Socially Interactive Agents
Multiparty Interaction Between Humans and Socially Interactive Agents
复制标题
DOI：
10.1145/3563659.3563665
复制
发表时间：
2022-10
期刊：
影响因子：
--
通讯作者：
Sarah Gillet;Marynel Vázquez;C. Peters;Fangkai Yang;Iolanda Leite
中科院分区：
文献类型：
--
作者：
Sarah Gillet;Marynel Vázquez;C. Peters;Fangkai Yang;Iolanda Leite
虽然大多数研究都集中在一对一的交互上，但由于多种原因，用于多方交互的社交交互代理（SIA）在过去几年中受到了越来越多的关注。首先，多方社交互动通常更加非结构化，但也更可能类似于现实世界的情况。当这些代理从实验室环境转移到现实世界（例如博物馆、医院或教室）时，重要的是他们能够处理可能出现的不同类型的社交情况。其次，人类的社会行为在很大程度上取决于社会背景，例如，我们单独的行为方式与我们在群体中的行为方式不同[Zajonc 1965]。因此，我们不能直接假设用户和 SIA 之间一对一交互的实证结果、计算模型等在多方环境中也成立。与一对一交互相比，多方设置给人类带来了额外的挑战。因此，它们对于 SIA 来说也更具挑战性，也就不足为奇了。该领域的核心能力与个人层面的社交能力毫不逊色，但通常与空间管理、编队控制和导航相关。此外，诸如代理人的行为方式和
While most research has focused on one-on-one interactions, Socially Interactive Agents (SIAs) for multiparty interaction have received increasingly more atten tion in the past years for a number of reasons. First, multiparty social interac tions are often more unstructured but also more likely to resemble real-world situations. As these agents move from lab environments to the real world (eg, museums, hospitals, or classrooms), it is important that they can handle the dif ferent types of social situations that may arise. Second, human social behavior is largely dependent on social context, for example, the way we behave alone is different from how we behave in a group [Zajonc 1965]. For this reason, we can not directly assume that empirical findings, computational models, and so on, from one-to-one interactions between users and SIAs will hold in a multiparty setting.When compared to one-on-one interactions, multiparty settings pose addi tional challenges for humans. Thus, it is no surprise that they can also be more challenging for SIAs. The core capabilities in this domain are no less social than at the individual level but typically relate to space management, formation con trol, and navigation. Additionally, questions such as how the agent behavior and