IEEE Trans Neural Syst Rehabil Eng. 2017 Dec;25(12):2295-2304. doi: 10.1109/TNSRE.2017.2753879. Epub 2017 Oct 16.
Humans show a great natural ability at interacting with each other. Such efficiency in joint actions depends on a synergy between planned collaboration and emergent coordination, a subconscious mechanism based on a tight link between action execution and perception. This link supports phenomena as mutual adaptation, synchronization, and anticipation, which cut drastically the delays in the interaction and the need of complex verbal instructions and result in the establishment of joint intentions, the backbone of social interaction. From a neurophysiological perspective, this is possible, because the same neural system supporting action execution is responsible of the understanding and the anticipation of the observed action of others. Defining which human motion features allow for such emergent coordination with another agent would be crucial to establish more natural and efficient interaction paradigms with artificial devices, ranging from assistive and rehabilitative technology to companion robots. However, investigating the behavioral and neural mechanisms supporting natural interaction poses substantial problems. In particular, the unconscious processes at the basis of emergent coordination (e.g., unintentional movements or gazing) are very difficult-if not impossible-to restrain or control in a quantitative way for a human agent. Moreover, during an interaction, participants influence each other continuously in a complex way, resulting in behaviors that go beyond experimental control. In this paper, we propose robotics technology as a potential solution to this methodological problem. Robots indeed can establish an interaction with a human partner, contingently reacting to his actions without losing the controllability of the experiment or the naturalness of the interactive scenario. A robot could represent an "interactive probe" to assess the sensory and motor mechanisms underlying human-human interaction. We discuss this proposal with examples from our research with the humanoid robot iCub, showing how an interactive humanoid robot could be a key tool to serve the investigation of the psychological and neuroscientific bases of social interaction.
人类在相互交流方面表现出了很强的天赋。这种联合行动的效率取决于计划协作和涌现协调之间的协同作用，这是一种基于动作执行和感知之间紧密联系的潜意识机制。这种联系支持了相互适应、同步和预期等现象，这些现象大大缩短了互动中的延迟和对复杂口头指令的需求，从而建立了共同意图，这是社会互动的核心。从神经生理学的角度来看，这是可能的，因为支持动作执行的相同神经系统负责理解和预测他人的观察动作。确定哪些人类运动特征允许与另一个代理进行这种涌现协调对于建立与人工智能设备的更自然和高效的交互范式至关重要，这些设备涵盖了从辅助和康复技术到伴侣机器人等各种应用。然而，研究支持自然交互的行为和神经机制存在着实质性的问题。特别是，涌现协调（例如，无意识的动作或注视）的无意识过程很难——如果不是不可能的话——以一种定量的方式来限制或控制人类代理。此外，在互动过程中，参与者以复杂的方式不断相互影响，导致行为超出了实验控制。在本文中，我们提出机器人技术作为解决这个方法论问题的一种潜在方法。机器人确实可以与人类伙伴建立交互，根据他的动作做出反应，而不会失去实验的可控性或交互场景的自然性。机器人可以作为一种“交互探针”来评估人类-人类交互的感觉和运动机制。我们通过与类人机器人 iCub 的研究示例来讨论这个建议，展示了交互式人形机器人如何成为研究社会交互的心理和神经科学基础的关键工具。
