EAGER: RDSV: Collaborative Design of Data and AI Systems for Science and Society

This project aims to build a mechanism for academia, industry, and the public sector to collaborate and co-design research and development (R&D) directions to simultaneously improve data systems and artificial intelligence (AI) systems. Such co-designed systems will be able to better address scientific and societal challenges, while avoiding the harm from the inappropriate use of data and AI, especially the harm on marginalized communities.

The advancement of AI brings forward unprecedented promises for breakthroughs in science and for vastly improved policy-making. But at the current moment, the fast pace of AI development actually poses major challenges to scientists in academia and to public-sector organizations. One can use scientists in academia as the example.

The advancement of AI technology far outpaces the scientists’ individual efforts to adopt AI. Many scientists are also not equipped with sufficient technical skills to adopt AI. As they rush to implement AI in research, many research outcomes have already become questionable and will harm the rigor, reproducibility, validity and trustworthiness of science.

Similarly, any inappropriate use of AI in the government decision making process could result in serious harm. Meanwhile, AI resources and talents are overwhelmingly concentrated in industry, which is developing an increasingly larger number of powerful AI systems but they may not be aligned well with the needs of scientific research and government decision making.

R&D direction co-design, which this project will explore, will help shape the development of emerging technologies, not just AI, so that they can impact science and society positively in more significant and faster ways. It will also strengthen the mentality of placing the needs of scientific research and public interest at the center of future technology development.

In addition, this project will help foster a balanced and vibrant national research and innovation ecosystem, with academia, industry, government and community playing their unique and central roles. Such an ecosystem can effectively leverage emerging technologies and fuel future technologies.

This project will bring together data science and AI methodologists from the University of Michigan and Microsoft, University of Michigan scientists who apply such data science and AI methods across research fields, and the city of Detroit data team. The scientific focus is to develop coordinated research on databases and new AI systems. This is because many AI systems are not yet optimal to deal with specialized data in scientific research and with government data.

Conversely, much of the enormous amount of scientific data and government data are not constructed to leverage the new AI systems. Making data “AI ready” will be a continued priority as novel forms of AI continue to emerge that use diverse types of data representation and preparation. Coordinated database research and AI research will enable data and AI to be more compatible.

Through workshops, presentations and technical demos, structured discussions and deliberations, the project participants will identify: 1) The traditional mechanisms of R&D direction design in industry and academia and their inefficiencies; 2) The next waves of AI systems and how they can advance scientific research and government decision-making; 3) Gaps in database research and AI research that can be filled through research co-design; 4) Possible harms to people and society, especially to marginalized communities, in the implementation of the new AI systems, and ways to mitigate the harms; 5) mechanisms to understand different priorities, perspectives and needs of different organizations and communities and ways to collaborate despite such differences; and 6) a concrete R&D direction based on the above considerations. The project team will employ social science theory and practice to enable equitable participation and deliberation.

The project team will disseminate their work and recommendations to enable the scaling of such effort.

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.