HCC: Large: Collaborative Research: Always-On Relational Agents for Social Support of Older Adults

该项目将创造一种新的智能、自主的虚拟或机器人代理，它可以长时间持续运行并与人类互动，其主要动机是与人类建立和维持长期的社会关系。本研究最初的应用重点是为独居老年人提供陪伴和社会支持，并促进他们的健康。例如，在一个典型的日子里，一个永远在线的关系代理可能会和成年人一起玩社交纸牌游戏，充当锻炼教练，并帮助安排与成年人的家人和朋友的访问或电话。一种新的社会代理综合理论，称为共享计划关系理论，将被发展为这些关系代理的原则基础。该理论将建立在一个永远在线的关系软件体系结构上，该体系结构将作为开源分发，供其他人使用和扩展。使用参与式设计过程，包括家庭和实验室研究，目标用户群体将帮助开发关系代理的规范，然后使用理论和软件体系结构构建关系代理，并将其放置在用户家中进行长期（一个月或更长时间）纵向评估。该项目将对关系、社会性、互动参与和社会支持模型做出基本的理论贡献。这项工作还将对一般人，特别是老年人如何在关系、活动和微观行为的分析层面上实施社会支持产生新的见解。新的永远在线的关系软件体系结构将是当前代理体系结构的根本进步，而当前代理体系结构只支持围绕明确指定的任务进行简短、集中的交互。该体系结构还将支持代理功能的增量扩展，并能够控制虚拟和机器人代理实施例。在现代社会中，社会孤立是一种普遍令人不安的趋势。永远在线的关系代理有可能直接抵消这种隔离，通过提供陪伴，并作为中介，通过电子和物理方式使孤立的人与其他人接触。陪伴和社会支持也被认为是疾病恢复和死亡率的重要积极因素，特别是对老年人而言。因此，本项目的应用重点具有帮助控制医疗保健成本的潜力。

Abstract This project will create a new category of intelligent, autonomous virtual or robotic agent, which is continuously operating and interacting with humans - always on - for long periods of time and whose primary motivation is building and maintaining long-term social relationships with humans. The initial application focus of this research is to provide companionship and social support and to promote wellness for older adults who are living alone. For example, during a typical day, an always-on relational agent might play a social game of cards with the adult, act as an exercise coach and help arrange visits or phone calls with the adult's family and friends. A new integrated theory of social agency, called SharedPlans Relationship Theory, will be developed to serve as a principled foundation for these relational agents. The theory will be grounded in an always-on relational software architecture, which will be distributed as open-source for others to use and extend. Using a participatory design process, including home and laboratory studies, the target user population will help develop the specifications for a relational agent, which will then be constructed using the theory and software architecture, and placed in users' homes for long-term (month or more) longitudinal evaluation. This project will make fundamental, theoretical contributions to models of relationship, sociality, interactional engagement and social support. The effort will also produce new insights into how people in general, and older adults in particular, enact social support at the relational, activity and micro-behavioral levels of analysis. The new always-on relational software architecture will be a fundamental advance over current agent architectures, which only support brief, focused interactions around a well-specified task. This architecture will also support incremental extension of agent capabilities and be able to control either virtual and robotic agent embodiments. Social isolation is a broadly troubling trend in modern society. Always-on relational agents have the potential to counteract this isolation both directly, by providing companionship, and as intermediaries, by putting isolated people in contact with other people, both electronically and physically. Companionship and social support are also known to be significant positive factors in disease recovery and mortality, especially for older adults. The application focus of this project therefore has the potential for helping with health care cost control.