Building BRIDGEs: Coordinating Standards, Diversity, and Ethics to Advance Biomedical AI

总体情况：摘要(项目说明) Bridge2AI是NIH的标志性倡议。它认识到数据科学增长中的挑战和机遇- 用于生物医学和行为研究以及医疗保健提供的科学和数据驱动的方法。我们已经到达 一个关键时刻：随着我们收集和分析数据的能力呈指数级增长，我们必须考虑如何 利用这些信息以公平的方式使每个人受益，提供一条共同的前进道路。数据生成 Bridge2AI内的项目(DGPS)将解决“重大挑战”：将塑造未来科学分布的问题 这可能会对许多人的健康和护理产生影响。集体集结这些力量需要 经验和洞察力，以创建协作、跨学科的努力，将不同的利益联系在一起- 持有者实现Bridge2AI的使命：发现、合作和学习。 从我们成功指导NIH大型倡议和(国际)国家科学研究的集体经验中构建 我们的桥梁协调中心(CC)旨在确保一套响应迅速的核心将支持- 在他们的重大挑战中移植和启用DGPS。代表多个机构(加州大学洛杉矶分校、宾夕法尼亚州立大学- 弗西蒂，佛罗里达大学，密歇根大学，南加州大学，俄勒冈健康与科学 大学，Sage Bionnetworks，EMBL-EBI)，我们提出了多个交互核心。这些核心有盘间- 几个关键领域的专业知识，包括生物医学信息学/数据科学和人工智能(方法、AP- 应用、评估)，以及跨不同领域和数据类型。我们的核心(道德、标准、工具 优化、技能和劳动力发展)已准备好互动，以促进与以下方面相关的交叉活动 道德和值得信赖的人工智能(ETAI)；公平原则(可查找、可访问、可互操作、可重复使用) 跨新出现的数据集和领域；对已开发的人工智能就绪数据集和 工具。在我们的CC中，我们将为不同的学员创建一个基础，让他们不仅了解人工智能在 生物医学/行为研究，但要有意义地参与其中-拥抱实验的异质性- 背景和目标，以最大限度地丰富这种多样性给我们的行动带来的丰富性和力量。 我们计划与团队核心合作，支持将Bridge2AI中不同的团队聚集在一起的活动。 我们的工作是由一个熟练的行政核心组织的，他将为这项工作提供监督和凝聚力- Davor，无论是在核心上，还是与DGPS和NIH。我们的核心被塑造成最大限度地提高 通过动态、现代的组件，将DGPS和Bridge2AI作为一个整体进行思想的整合和共享 交流方法；在这些团体和更广泛的科学部门之间改进和传播最佳做法-- 通过多个场所实现社区一体化；以及对方法和总体效果的评价 Bridge2AI倡议。此CC将为Bridge2AI提供统一的框架，使其参与和教育不同的人 利益相关者，共同开辟生物医学和行为人工智能的集体前进道路-为每个人。

OVERALL: ABSTRACT (PROJECT DESCRIPTION) Bridge2AI is a signature NIH initiative. It recognizes the challenges and opportunities in the growth of data sci- ence and data-driven methods for biomedical and behavioral research and healthcare delivery. We have reached a key moment: with the exponential growth of our ability to collect and analyze data, we must consider how we use this information to benefit everyone in an equitable way, providing a collective path forward. Data Generation Projects (DGPs) within Bridge2AI will tackle “grand challenges”: questions that will shape future scientific dis- covery and can ultimately impact the health and care of many. Marshalling these forces collectively requires experience and insight to create a collaborative, interdisciplinary endeavor that brings together disparate stake- holders to realize Bridge2AI’s mission: discovery, collaboration, and learning. Building from our collective experience in successfully guiding large NIH initiatives and (inter)national scientific consortia, our BRIDGE Coordination Center (CC) is designed to ensure a responsive set of Cores that will sup- port and enable the DGPs in their grand challenges. Representing multiple institutions (UCLA, Penn State Uni- versity, University of Florida, University of Michigan, University of Southern California, Oregon Health & Sciences University, Sage Bionetworks, EMBL-EBI), we propose multiple interacting Cores. These Cores have interdisci- plinary expertise across several key areas, including biomedical informatics/data science and AI (methods, ap- plications, evaluation), as well as across different domains and data types. Our Cores (Ethics, Standards, Tool Optimization, Skills & Workforce Development) are ready to interact to facilitate cross-cutting activities related to ethics and trustworthy artificial intelligence (ETAI); FAIR principles (findable, accessible, interoperable, reusable) across emergent datasets and domains; comparison and benchmarking of developed AI-ready datasets and tools. Across our CC we will create a basis for diverse trainees to not only appreciate the implications of AI in biomedical/behavioral research, but to meaningfully engage with them – embracing the heterogeneity of experi- ences, backgrounds, and objectives to maximize the richness and strength this diversity brings in our actions. We plan to work with a Teaming Core to enable activities that bring together disparate groups within Bridge2AI. Our efforts are organized by a skilled Administrative Core who will provide oversight and cohesion to this en- deavor, both across the Cores as well as with the DGPs and NIH. Our Cores are shaped to maximize the inte- gration and sharing of ideas across the DGPs and Bridge2AI as a whole through dynamic, contemporary com- munication methods; the refinement and dissemination of best practices between these groups and wider sci- entific community through multiple venues; and the evaluation of the effectiveness of the methods and overall Bridge2AI initiative. This CC will provide a unified framework for Bridge2AI to engage and education different stakeholders, and together blaze a collective trail forward for biomedical and behavioral AI – for everyone.