CAREER: Tracking correlation or inferring causation: How human language processing adapts to the environment

职业：跟踪相关性或推断因果关系：人类语言处理如何适应环境

• 批准号: 2336304
• 负责人: Rachel Ryskin
• 金额: $ 67.42万
• 依托单位: University of California - Merced
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2024
• 资助国家: 美国
• 起止时间: 2024-05-01 至 2029-04-30
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2336304&HistoricalAwards=false
• 关键词: CAREER; Tracking; correlation; inferring; causation

When people converse face-to-face, on the phone, or by video conference, the signal is often noisy and incomplete. Luckily, communication usually still proceeds smoothly because listeners can use other sources of information (the context, visual cues, etc.) to make rational inferences about what the speaker said. This CAREER award aims to expand our theoretical understanding of this process and refine existing computational accounts, by investigating how listeners flexibly adapt their inferences and expectations in different contexts. For example, how do adults adapt their expectations when speaking with a child rather than with another adult or when speaking with other adults with different accents? Results of this work will advance knowledge of the cognitive processes that enable human communication and may have important implications for the development of technologies that support human-to-human or human-to-AI conversations. The PI is in a Hispanic-Serving Institution and plans to broaden participation in STEM science on multiple scales, from one-on-one mentorship of undergraduates to an upper-level undergraduate course and an outreach talk series.The project investigates how listeners flexibly adapt to different contexts (e.g., different speakers/environments), within the perspective of human communication as rational inference under uncertain input. Two possible mechanisms are that listeners track correlations in the environment or that they deploy and update a causal model of the speaker and transmission process. A series of eye-tracking experiments, paired with computational simulations, will capture the inferences that listeners make “on the fly,” when listening to speech. How listeners learn and adjust their representations to the context will be probed by varying the nature and quantity of errors and noise.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

当人们面对面、通过电话或视频会议匡威时，信号往往是嘈杂和不完整的。幸运的是，沟通通常仍然顺利进行，因为听众可以使用其他信息来源（上下文，视觉提示等）。对演讲者所说的话做出合理的推断。这个CAREER奖旨在扩大我们对这一过程的理论理解，并通过调查听众如何在不同的背景下灵活地调整他们的推断和期望来完善现有的计算账户。例如，当与孩子而不是与另一个成年人交谈时，或者与其他有不同口音的成年人交谈时，成年人如何调整他们的期望？这项工作的结果将促进对人类沟通的认知过程的了解，并可能对支持人与人或人与人工智能对话的技术的开发产生重要影响。PI是在西班牙裔服务机构，并计划扩大参与STEM科学在多个尺度上，从一对一的本科生导师到高层次的本科课程和外联讲座系列。该项目调查听众如何灵活地适应不同的环境（例如，不同的说话人/环境），在人类交流的角度下，不确定输入下的理性推理。两种可能的机制是，听众跟踪环境中的相关性，或者他们部署和更新扬声器和传输过程的因果模型。一系列的眼动追踪实验，再加上计算机模拟，将捕捉听众在听演讲时“在飞行中”做出的推断。听众如何学习和调整他们的陈述，以不同的性质和数量的错误和噪音的情况下，将探讨。这个奖项反映了NSF的法定使命，并已被认为是值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估的支持。