The National Institutes of Health will invest $130 million over four years, to accelerate the widespread use of artificial intelligence (AI) by the biomedical and behavioral research communities. NIH’s Common Fund program, Bridge2AI, brings together team members from diverse disciplines and backgrounds to generate tools, resources, and richly detailed data relevant to AI approaches. At the same time, the program ensures that its tools and data do not perpetuate any injustice or ethical issues that may arise during data collection and analysis. Through extensive cross-project collaboration, Bridge2AI researchers will create guidelines and standards for the development of ethical, state-of-the-art, AI-enabled datasets that have the potential to help solve some of the most most pressing issues in human health – like discovering how genetics, behavioral and environmental factors affect a person’s fitness throughout their lifetime.
Generating high-quality ethical datasets is critical to enabling the next generation of artificial intelligence technologies that will transform the way we conduct research, said Lawrence A. Tabak, D.D.S., Ph.D., who serves as director of the NIH. Solutions to long-standing human health problems are within reach, and now is the time to connect researchers and artificial intelligence technologies to solve our most challenging research problems and ultimately help improve human health.
Artificial intelligence is both a field of science and a set of technologies that enable computers to imitate the way humans perceive, learn, reason and act. Although artificial intelligence has been used in biomedical research and healthcare, its widespread use is limited in part by the challenges of applying AI techniques to different types of data. This is because routinely collected biomedical and behavioral datasets are often sparse, meaning they lack important contextual information about data types, collection conditions, or other parameters. Without this information, AI technology cannot accurately analyze and interpret data. AI technologies can also inadvertently contain bias or inequality unless careful consideration is given to the social and ethical context in which the data is collected. To harness the power of AI for biomedical discovery and accelerate its use, researchers must first describe well-characterized and ethical datasets, standards, and best practices for generating biomedical and behavioral data that can be used for AI analysis.
As Bridge2AI develops tools and best practices for data preparation for AI, it will also develop a wide range of data types for use by the research community for AI analysis. These categories include sounds and other data that help identify abnormal changes in the body. Researchers will generate data that can be used to establish new connections between complex genetic pathways and changes in cell shape or function. In addition, AI-ready data will be generated to improve decision-making in critical care to facilitate rapid recovery from acute illness and help identify the complex biological processes underlying human recovery from illness.
The Bridge2AI program is committed to fostering research teams with diverse perspectives, experiences, and a diversity of academic and technical disciplines. Diversity is essential for the ethical generation of datasets and the training of future AI methods to reduce bias and improve efficiency for all populations, including those underrepresented in biomedical and behavioral research. Bridge2AI will develop ethical practices for data generation and use, addressing key issues of privacy, data reliability, and bias reduction.
NIH awarded four awards to data generation projects and three awards to connecting centers that form integration, dissemination, and evaluation activities. The data generation project will generate new biomedical and behavioral datasets that can be used to advance artificial intelligence technologies, while creating data standards and tools to ensure data availability, accessibility, interoperability and reusable, known as the FAIR principle. In addition, the data generation project will develop training materials to promote the use of different cultural and ethical practices throughout the data generation process. The Bridge Center will be responsible for integrating activities and knowledge into data generation projects and the dissemination of products, best practices and training materials.
