文献类型：期刊文献

中文题名：社会机器人何以可能?——朝向一种具身卷入的人工智能设计

英文题名：How are Social Robots Possible? Toward an Embodied Engagement Approach to Design of Artificial Intelligence

作者：陈巍[1];赵翥[2]

机构：[1]绍兴文理学院脑与社会心智实验室;[2]云南师范大学心理学系

年份：2018

卷号：40

期号：1

起止页码：17

中文期刊名：自然辩证法通讯

外文期刊名：Journal of Dialectics of Nature

收录：北大核心2017、CSSCI2017_2018、北大核心、CSSCI

基金：国家社会科学基金项目"现象学与神经科学对话中他心问题的重构及其超越研究"(项目编号:16CZX015)

语种：中文

中文关键词：社会机器人;人工智能;人机互动;具身认知;卷入性;镜像神经元

外文关键词：Social robots; Artificial intelligence; Human-Robot interaction; Embodied cognition;Engagement; Mirror neuron

中文摘要：当前,AI已经在高级思维与问题解决上取得了长足的进步,但在设计与开发用于人机互动的社会机器人上仍然面临诸多的难题。与各种社会机器人的互动中感知到它是一个可以提供某种互动的自主体究竟意味着什么?我们用什么标准来判断当前的人机互动中包含"真正"的社会性?通过整合社会神经科学、动物行为学、发展心理学等多个学科领域内的研究证据,具身卷入的社会互动观认为社会机器人的AI设计需要应对如下问题:(1)如何利用系统1直接知觉社会行为?(2)机器人的形态与生物学规律如何影响社会互动?(3)如何让社会机器人与人类建构共同任务?这将为AI在设计社会机器人上反思图灵测试的前提假设以及处理"4W/H"问题提供丰富启示。

外文摘要：Over the past decades, Artificial Intelligence （AI） advanced significantly within the realm of abstract reasoning and problem solving. But the flourishing of AI didn＇t expand to the realm of design and developing social robots. Cognitivist AI encounters many formidable problems when applied to social cognition. What differs an interactive agent from a physics object when subject percept environment？ Which conditions qualify human-robot interaction to be truly social interactive？ By synthesizing evidences came from social neuroscience, ethology, developmental psychology, etc. Based on the approach of embodied engagement, we summarize three aspects worth to be considered when design and developing social robots in the future. （1） How to use a low level, automatic system to accomplish the direct perception of social behavior？ （2） What morphology and biological law can be used in social robot to enhance experience of human-robot interaction？ （3） Which pattern of interaction can build a joint task between human and robot？ The inquiry and clarify of those questions may inspire further reflection of Turing test and 4W/H problem faced by social robot and provide an alternative view of AI design.