Doctoral Dissertation Research: Designing Voice Analysis Technologies for Mental Health Applications in the United States

这个项目对研究生进行基于经验的科学研究方法进行培训，询问人工智能(AI)支持的新技术如何不仅重塑美国精神卫生保健的未来，而且重塑关于语言、思维和大脑之间关系的基本假设。这项研究通过对三所美国大学的跨学科研究团队进行的人种学研究来探索这些问题，这些研究团队正寻求为心理健康应用开发计算机辅助语音分析技术。在这项研究关注的研究团队中，神经学家、精神病学家、心理学家和工程师正在合作开发一种技术，可以通过分析语音的形式声学特性(如音调、音色、语调和语速)来诊断和跟踪精神疾病，完全绕过它的语义内容(单词的意思)。该项目将对心理健康研究人员本身产生影响，因为他们在隐私、监控和对实验技术日益增长的诊断依赖方面进入了未知的伦理领域。贝丝·塞梅尔在麻省理工学院格雷厄姆·琼斯博士的指导下，探索了当专家们对语言和内在心理状态之间的关系持有截然不同的假设时，跨学科边界的专家如何合作设计、开发和测试人工智能启用的技术。这位研究人员假设，围绕这些技术开发的合作不仅在语言如何运作的主流观点中引发了根本的紧张，还反映了美国精神卫生保健领域对权威和专业知识的重新定位。越来越多的精神健康研究人员正在避开传统的精神诊断技术，这种技术依赖于患者对其心理状态的主观、口头描述，以及临床医生的观察和解释技能。相反，他们正在招募计算机工程师的专业知识，这些工程师使用模式识别的人工智能技术来破译行为症状的生物医学意义。通过人种学参与者观察，研究人员将收集有关精神病学家、心理学家、神经科学家和工程师如何合作设计、测试和开发语音分析技术的数据。这项研究聚焦于位于学术、商业和军事舞台交汇处的团队，探索了从科学、公共卫生和国家安全考虑的角度对精神疾病进行概念化的各种方式。通过探索听力练习如何塑造关于言语的假设，以及新的听力技巧和技术的产生如何重塑这些假设，这项研究有助于语言人类学中关于文化如何影响对语言运作方式的理解的持续辩论。

This project, which trains a graduate student in methods of conducting empirically-grounded scientific research, asks how new, artificial intelligence (AI)-enabled technologies reshape not only the future of mental health care in the United States, but also basic assumptions about the relationship between language, mind, and brain. This research explores these questions through an ethnographic study of interdisciplinary research teams at three U.S. universities that are seeking to develop computer-assisted speech analysis technologies for mental health applications. In the research teams that this study focuses on, neuroscientists, psychiatrists, psychologists and engineers are working together to develop technology that can be used to diagnose and track mental illness by analyzing the formal, acoustic properties of speech (such as pitch, timbre, intonation, and speed), bypassing its semantic content (what the words mean) altogether. The project will have implications for the mental health researchers themselves as they move into uncharted ethical domains in regards to privacy, surveillance, and the increased diagnostic reliance on experimental technologies.Beth Semel, under the supervision of Dr. Graham Jones at the Massachusetts Institute of Technology, explores how experts across disciplinary boundaries collaborate to design, develop and test artificial intelligence-enabled technologies when they appear to hold very different assumptions about the relationship between language and inner, psychological states. The researcher hypothesizes that collaborations surrounding the development of these technologies not only enact fundamental tensions within dominant views about how language works, but also reflect a re-working of claims of authority and expertise within U.S. mental health care. Increasingly, mental health researchers are eschewing traditional techniques of psychiatric diagnosis, which depend upon patients' subjective, verbal accounts of their psychological states and clinicians' observational and interpretive skills. Instead, they are enlisting the expertise of computer engineers who use AI techniques of pattern recognition to decipher the biomedical significance of behavioral symptoms. Using ethnographic participant observation, the researcher will collect data about how psychiatrists, psychologists, neuroscientists and engineers work together to design, test, and develop voice analysis technologies. Focusing on teams situated at the confluence of academic, commercial, and military arenas, this study explores the variety of ways in which mental illness is conceptualized in terms of scientific, public health, and national security concerns. By exploring how listening practices can shape assumptions about speech, and how the production of new listening techniques and technologies can reshape such assumptions, this research contributes to ongoing debates in linguistic anthropology about how culture affects understandings of the way language works.