Embodied conversational agents (ECAs) are increasingly used in healthcare and other settings to improve self-management and provide companionship. Their ability to form close relationships with people is important for enhancing effectiveness and engagement. Several studies have looked at enhancing relationships with ECAs through design features focused on behaviours, appearance, or language. However, this evidence is yet to be systematically synthesized. This systematic review evaluates the effect of different design features on relationship quality with ECAs. A systematic search was conducted on electronic databases EMBASE, PsychInfo, PubMed, MEDLINE, Cochrane Library, SCOPUS, and Web of Science in January–February 2019. 43 studies were included for review that evaluated the effect of a design feature on relationship quality and social perceptions or behaviours towards an ECA. Results synthesize effective design features and lay a scientific framework for improving relationships with ECAs in healthcare and other applications. Risk of bias for included studies was generally low, however there were some limitations in the research quality pertaining to outcome measurement and the reporting of statistics. Further research is needed to understand how to make ECAs effective and engaging for all consumers.

体验式对话代理（ECA）越来越多地用于医疗保健和其他场合，以改善自我管理并提供陪伴。他们与人建立亲密关系的能力对于提高效率和参与度很重要。多项研究着眼于通过着重于行为，外观或语言的设计特征来增强与ECA的关系。但是，该证据尚待系统地综合。该系统评价评估了不同设计特征对与ECA关系质量的影响。2019年1月至2019年2月，对电子数据库EMBASE，PsychInfo，PubMed，MEDLINE，Cochrane图书馆，SCOPUS和Web of Science进行了系统的搜索。纳入了43项研究进行评估，以评估设计功能对关系质量和对ECA的社会认知或行为的影响。结果综合了有效的设计特征，并为改善与ECA在医疗保健和其他应用中的关系奠定了科学的框架。纳入研究的偏倚风险通常较低，但是与结果测量和统计报告有关的研究质量存在一些局限性。需要进一步研究以了解如何使ECA有效并使所有消费者参与。但是，与结果测量和统计报告有关的研究质量存在一些局限性。需要进一步研究以了解如何使ECA有效并使所有消费者参与。但是，与结果测量和统计报告有关的研究质量存在一些局限性。需要进一步研究以了解如何使ECA有效并使所有消费者参与。