Integrative mechanisms in human-human and human-robot communication across the adult lifespan.

成人一生中人与人以及人与机器人沟通的整合机制。

• 批准号: RGPIN-2020-07214
• 负责人: Chambers, Craig
• 金额: $ 3.42万
• 依托单位: University of Toronto
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2022
• 资助国家: 加拿大
• 起止时间: 2022-01-01 至 2023-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=764192
• 关键词: Integrative; mechanisms; human; robot; communication

Spoken language is produced and understood at a striking pace. For example, words in fluent speech are often generated at a rate of 2 or 3 per second, and many core linguistic distinctions (such as the acoustic properties that differentiate individual speech sounds) hinge on timing differences at the level of tens of milliseconds. How does the human mind cope with the rapid and dynamic nature of spoken language? The broad goals of our research program are to understand how the speed and efficiency of language processing result from the particular ways that humans unconsciously combine visual and auditory signals with mentally-stored knowledge. One theme in our past research explored how real-time comprehension draws on a listener's implicit expectations about the form and content of what will be said next. Another theme explored the extent to which listeners rapidly draw on fine-grained acoustic details in the auditory signal. Further, by making comparisons between younger and older (65+) adults, we have been working to understand how communicative abilities change over the course of the adult lifespan.  Going forward, one objective of our current research is to understand how the speed and efficiency of language processing arises from the information beyond the speech signal itself. Communicative requirements vary from context to context, and this may entail conscious or unconscious changes in the way that listeners attend to and draw on potentially helpful information sources. A second objective is to develop insights into the way that language processing abilities change as we age.  Given the complexity of the underlying mechanisms, this goal requires a clear programmatic approach involving converging evidence from many different angles. Finally, given recent advances in technology, one increasingly important topic involves humans' interactions with artificial agents equipped with spoken language interfaces. To date, comparatively little is known about how humans adapt or adjust their speaking and listening strategies in these scenarios. To address this issue, we have begun to use human-robot communication to systematically identify which processes occur automatically as people engage in communicative interaction, and which are under more conscious control. This work is conducted with a programmable social robot whose communicative behaviours (e.g., speech rate, intonation, accent, pauses, eye gaze patterns) can be controlled with fine-grained precision. In all of these studies, we employ sensitive and adaptable eye-tracking techniques to reveal the character of comprehension processes as language is encountered in real time at the millisecond level.  The outcomes of this research will provide important insights into the cognitive architecture that supports core language processing abilities as we communicate with other humans or with artificial agents.

口语的产生和理解速度惊人。例如，流利语音中的单词通常以每秒 2 或 3 个的速度生成，并且许多核心语言区别（例如区分各个语音的声学特性）取决于数十毫秒级别的时间差异。人类思维如何应对口语的快速和动态特性？我们研究计划的总体目标是了解人类无意识地将视觉和听觉信号与心理存储的知识相结合的特定方式如何导致语言处理的速度和效率。我们过去研究的一个主题探讨了实时理解如何利用听众对接下来要说的内容的形式和内容的隐含期望。 另一个主题探讨了听众快速吸收听觉信号中细粒度声学细节的程度。此外，通过比较年轻人和老年人（65 岁以上），我们一直在努力了解沟通能力在成年人的一生中如何变化。  展望未来，我们当前研究的目标之一是了解语言处理的速度和效率是如何从语音信号本身之外的信息中产生的。沟通要求因语境而异，这可能需要有意识或无意识地改变听众关注和利用潜在有用信息源的方式。第二个目标是深入了解语言处理能力随着年龄的增长而变化的方式。  考虑到潜在机制的复杂性，这一目标需要一种明确的程序化方法，涉及来自许多不同角度的汇聚证据。最后，鉴于最近技术的进步，一个日益重要的主题涉及人类与配备口语界面的人工智能体的交互。迄今为止，人们对于人类如何在这些场景中适应或调整他们的口语和听力策略知之甚少。为了解决这个问题，我们已经开始使用人机通信来系统地识别当人们进行交流互动时哪些过程会自动发生，哪些过程受到更有意识的控制。这项工作是通过一个可编程社交机器人进行的，该机器人的交流行为（例如语速、语调、口音、停顿、眼神模式）可以进行精细控制。在所有这些研究中，我们采用敏感且适应性强的眼球追踪技术来揭示在毫秒级实时遇到语言时的理解过程的特征。  这项研究的结果将为我们与其他人或人工智能体交流时支持核心语言处理能力的认知架构提供重要的见解。