Collaborative Research: FW-HTF-RL: Understanding the Ethics, Development, Design, and Integration of Interactive Artificial Intelligence Teammates in Future Mental Health Work

这个研究项目是针对全国缺乏精通研究支持治疗方案的精神卫生工作者的问题作出的反应。研究人员试图了解人工智能（AI）的最新创新如何有效和合乎道德地解决和减轻对精神健康治疗的未满足需求。精神卫生工作者包括临床心理学家、社会工作者和咨询师等相关职业。这一规模不足的劳动力迫切需要可扩展和有效的技能提升，以促进研究支持的治疗方案的广泛和常规实施。提高工作人员的技能受到限制，因为没有足够的专家培训人员使精神卫生工作者熟练掌握现有的最佳做法。这种劳动力主要依赖于最初的人与人之间的培训（例如，研究生院），随后是在整个职业生涯中相对较少的后续观察和反馈。因此，数百万有精神健康问题的美国人很难获得有效的、有研究支持的护理。精神卫生工作人员将受益于帮助临床医生学习和维持他们使用研究支持的治疗方案的技术。重要的是，现代人工智能系统已经发展到这样一个程度，即该技术可以被视为高技能工作环境中的队友，而不仅仅是数据处理工具。综合人工智能的最新进展，跨学科研究团队将开发一个交互式人工智能系统，该系统可以快速评估精神卫生工作者对患者的表现，向工作者提供可操作的反馈，并接收工作者的输入，以便根据个人工作者需要学习的内容提供反馈。这个计算系统被称为可信赖的、可解释的、自适应的人工智能团队监控机器（TEAMMAIT），它将作为一个客观的、无判断力的、保密的同事，在一段时间内提供个性化的反馈。这种类型的工人-人工智能团队有可能通过减少对成本高昂且难以获得的人与人之间培训的依赖来改变技能提升过程。虽然该项目侧重于心理健康工作，但由于关键的未满足需求，该项目的见解可以推广到其他医疗保健和教育环境。这个项目汇集了几个学科，包括临床心理学、工业组织心理学、人机交互和信息科学。团队的结构是为了实现多个聚合目标。首先，研究人员旨在更好地了解引入工人-人工智能团队将如何影响精神卫生工作者的预期能力，包括如何与人工智能合作和应对风险。其次，研究人员旨在深入了解如何在心理健康工作中设计人工智能队友，以促进道德和有效的工人-人工智能团队。第三，调查人员的目标是学习如何开发和部署人工智能队友，以提高精神卫生工作人员的技能。TEAMMAIT的原型将在不同的环境、不同的工作人员和不同的患者群体中进行评估。从原型用户收集的数据将为精神卫生工作中的工人-人工智能团队制定一套开发指南，以及为开发和使用这些系统制定一套可推广的道德指南。对用户的访谈将深入了解心理健康工作场所如何为工人-人工智能团队做好最好的准备，并优化其使用，同时保持工人的福祉和高质量的临床护理。该研究计划将提供一些见解，帮助精神卫生工作者在现实世界的诊所中提高技能，使其更具可扩展性和有效性，为美国各地不同的患者群体提供最佳实践。该项目由人类-技术前沿跨部门计划的未来工作资助，旨在通过推进与人类工人和谐运作的智能工作技术的设计，促进对工作环境中相互依赖的人类-技术伙伴关系的更深层次的基本理解。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This research project is a response to the national shortage of mental health workers who are skilled in research-supported treatment protocols. The investigators seek to understand how recent innovations in artificial intelligence (AI) can effectively and ethically address and mitigate unmet demands for mental health treatment. Mental health workers include several related professions including clinical psychologists, social workers, and counselors. This undersized workforce is in dire need for scalable and effective upskilling in order to facilitate widespread and routine implementation of research-supported treatment protocols. Upskilling the workforce has been constrained because there are insufficient numbers of expert trainers to keep mental health workers proficient in the best available practices. This workforce has primarily relied on initial human-to-human training (e.g., graduate school) followed by relatively minimal follow-up observation and feedback throughout one’s career. As a result, millions of Americans with mental health conditions have restricted access to effective, research-supported care. The mental health workforce will benefit from technology that helps clinicians learn and sustain their use of research-supported treatment protocols. Important to this need, modern AI systems have developed to such a point where the technology can be considered a teammate in highly skilled work contexts, not simply a data processing tool. Integrating recent advancements in AI, the interdisciplinary team of investigators will develop an interactive AI system that can quickly evaluate a mental health worker’s performance with a patient, provide actionable feedback to the worker, and receive input from the worker so that feedback is based on what that individual worker needs to learn. This computational system, called the Trustworthy, Explainable, and Adaptive Monitoring Machine for AI Teams (TEAMMAIT), will function as an objective, nonjudgmental, and confidential colleague who can provide individualized feedback over a period of time. This type of Worker-AI Teaming has potential to transform the upskilling process by reducing the reliance on cost-prohibitive and scarcely available human-to-human training. While this project focuses on mental health work due to critical unmet demands, insights from this project can generalize to other healthcare and educational contexts.This project brings together several disciplines including clinical psychology, industrial-organizational psychology, human-computer interaction, and information science. The team is structured to achieve multiple convergent goals. First, the investigators aim to better understand how introducing Worker-AI Teams will impact the expected competencies of mental health workers including how to collaborate with AI and respond to risks. Second, the investigators aim to gain insights regarding how to design AI Teammates in mental health work that facilitate ethical and effective Worker-AI Teaming. And third, the investigators aim to learn how to develop and deploy AI Teammates that can upskill the mental health workforce. A prototype of TEAMMAIT will be evaluated in diverse settings and with diverse workers and diverse patient populations. Data collected from prototype users will result in a set of development guidelines for Worker-AI Teaming in mental health work, as well as a set of generalizable ethical guidelines for developing and using these systems. Interviews with users will provide insights into how mental health workplaces can best prepare for Worker-AI Teaming and optimize its use while maintaining worker well-being and high-quality clinical care. The research plan will provide insights that will help make mental health worker upskilling more scalable and effective in real-world clinics, improving access to best practices for diverse patient populations across the United States. This project has been funded by the Future of Work at the Human-Technology Frontier cross-directorate program to promote deeper basic understanding of the interdependent human-technology partnership in work contexts by advancing the design of intelligent work technologies that operate in harmony with human workers.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.