NSF Program on Fairness in Artificial Intelligence in Collaboration with Amazon | 21 585

Opportunity Information: Apply for 21 585

The NSF Program on Fairness in Artificial Intelligence in Collaboration with Amazon (Funding Opportunity No. 21-585) is a National Science Foundation grant program designed to push forward foundational research that makes AI systems more trustworthy and broadly acceptable for real-world deployment. The core idea behind the program is that as AI and machine learning increasingly shape decisions in areas like security, healthcare, employment, finance, education, and public services, the public will only accept these systems at scale if they can be shown to behave fairly and responsibly across different communities and use cases. NSF frames fairness broadly, tying it directly to practical trustworthiness concerns such as transparency, explainability, impartiality, inclusivity, accountability, and broad accessibility and utility, rather than treating fairness as a narrow technical metric.

A key motivation highlighted in the opportunity is the growing use of AI in high-impact domains where biased or uneven performance can cause real harm, including widely deployed technologies such as face recognition, speech processing, and language technologies, especially when these tools are embedded in decision support systems. The program calls for new computational and theoretical advances that go beyond current capabilities, with an emphasis on methods that can assess fairness, demonstrate it credibly, and mitigate unfair outcomes. In other words, the program is not only interested in identifying bias, but also in building the scientific and engineering foundations needed to prevent, measure, validate, and correct it in realistic settings.

NSF and Amazon jointly fund the program, but NSF makes it clear that Amazon does not participate in selecting proposals for award. The partnership is presented as a way to support research that can lead to trustworthy AI systems that society is willing to adopt, while maintaining an independent merit review and award decision process under NSF. The broader national payoff described is that improving fairness and trust can help the United States more fully realize the economic and societal benefits of AI, while reducing the risks that can undermine public confidence and slow adoption.

The solicitation encourages proposals across a wide set of research topics, explicitly including transparency and explainability (methods that help people understand model behavior and decisions), accountability (mechanisms that make responsibility and oversight possible), inclusivity (ensuring systems serve diverse populations and contexts), and the study of adverse biases including social biases and downstream effects. It also calls for mitigation strategies, algorithmic advances, new fairness objectives (how fairness is defined and optimized), validation of fairness (how claims of fairness are tested and supported with evidence), participatory design (engaging affected stakeholders in system design and evaluation), and advances that improve broad accessibility and utility so AI systems work well for more people, not just majority groups or well-resourced settings.

A major theme of the opportunity is interdisciplinarity. NSF emphasizes that progress on AI fairness is not solely a computer science problem, because fairness is tied to human behavior, institutions, social context, and real decision environments. As a result, proposals are expected to draw from or meaningfully integrate perspectives from areas such as computer and information science, engineering, statistics and mathematics, and also cognitive science, psychology, sociology, decision science, and economics. The solicitation explicitly notes that insights from the social, behavioral, and economic sciences can shape better computational approaches to algorithmic and data fairness, which signals an interest in projects that connect technical methods with how systems are actually used, interpreted, governed, and experienced in practice.

The program is structured as a discretionary NSF grant opportunity within the Science and Technology and other Research and Development category, with CFDA numbers 47.070 and 47.075. It anticipated making about 12 awards, and the award ceiling is listed as 0 (typically meaning a formal ceiling is not specified in the summary record and applicants should rely on the solicitation text for budget expectations). The original closing date for submissions was August 3, 2021, and the opportunity was created on May 6, 2021. Eligibility is described broadly as “Others (see text field entitled Additional Information on Eligibility for clarification),” which usually indicates that standard NSF eligibility rules apply with additional details provided in the full announcement.

Finally, the solicitation aligns strongly with NSF’s broader mission of expanding participation in STEM. It specifically encourages submissions from investigators at institutions that serve groups historically underrepresented in STEM disciplines, and it welcomes both single-institution proposals and multi-institution collaborations. This emphasis is positioned as especially fitting for a program centered on fairness and inclusion, reinforcing that the research community building these systems should itself be broad and representative, and that the benefits of the research should reach diverse regions, institutions, and populations.

• The National Science Foundation in the science and technology and other research and development sector is offering a public funding opportunity titled "NSF Program on Fairness in Artificial Intelligence in Collaboration with Amazon" and is now available to receive applicants.
• Interested and eligible applicants and submit their applications by referencing the CFDA number(s): 47.070, 47.075.
• This funding opportunity was created on May 06, 2021.
• Applicants must submit their applications by Aug 03, 2021. (Agency may still review applications by suitable applicants for the remaining/unused allocated funding in 2026.)
• The number of recipients for this funding is limited to 12 candidate(s).
• Eligible applicants include: Others (see text field entitled Additional Information on Eligibility for clarification).

Apply for 21 585

[Watch] Creating a grant proposal using the step-by-step wizard inside the applicant portal: