Advanced Computing in the Age of AI | Thursday, March 28, 2024

NSF Gets Pushback on AI Ethics Research 

The National Science Foundation is taking some heat for collaborating with Amazon to fund research on AI fairness. With proposals due at the end of June (“letters of intent” were due on May 10), critics are warning that AI leaders like Amazon are waging a campaign to influence the AI fairness debate in their favor.

At first glance, the agency’s notice, “NSF Program on Fairness in Artificial Intelligence (AI) in Collaboration with Amazon,” raised some eyebrows with the inclusion of a corporate sponsor (NASDAQ: AMZN) in the program title. After all, Amazon’s industry leading cloud division already provides services across the federal government. It clearly has a stake in forging AI fairness guidelines.

“Although Amazon provides partial funding for this program, it will not play a role in the selection of proposals for award,” NSF officials noted in their program description.

“NSF's ability to bring together multiple scientific disciplines uniquely positions the agency in this collaboration, while building AI that is fair and unbiased is an important aspect of Amazon’s AI initiatives,” the synopsis continued.

Those provisions have done little to reassure skeptics who worry Amazon, Facebook (NASDAQ: FB), Google (NASDAQ: GOOGL) and others are exercising undue influence over guidelines designed to ensure, according to the NSF solicitation, “transparency, explainability, accountability, potential adverse biases and effects, mitigation strategies, validation of fairness and considerations of inclusivity.”

Among those pushing back is Yochai Benkler, a law professor at Harvard University and co-director of Harvard’s Berkman Klein Center for Internet and Society. In a commentary published in the journal Nature, Benkler cited efforts by technology industry stakeholders to shape the AI ethics debate, including Facebook’s investment in an AI ethics center at the Technical University of Munich and Google’s defunct AI ethics board.

“Organizations working to ensure that AI is fair and beneficial must be publicly funded, subject to peer review and transparent to civil society,” Benkler asserted. “And society must demand increased public investment in independent research rather than hoping that industry funding will fill the gap without corrupting the process.”

NSF has stated its AI fairness research will reflect the fact that “AI is a highly interdisciplinary endeavor” cutting across computer, information and cognitive sciences as well as engineering, statistics, mathematics and—particularly with regard to AI ethics—psychology.

Nevertheless, Benkler noted that growing mistrust of large technology companies, especially among European regulators, provides government agencies like NSF with leverage in the AI ethics debate. Indeed, U.S. lawmakers have taken tentative steps toward developing guidelines aimed at promoting ethical AI development.

Therefore, critics sense an opportunity to push back against tech giants as the AI ethics debate heats up. “Governments should use that leverage to demand that companies share data in properly-protected databases with access granted to appropriately insulated, publicly-funded researchers,” Benkler wrote in Nature. “Industry participation in policy panels should be strictly limited.”

Update: The National Science Foundation emailed the following response to Yochai Benkler's commentary in Nature:

"NSF's 'Program on Fairness in Artificial Intelligence in Collaboration with Amazon,' like NSF’s other public-private partnerships, is focused on ensuring American leadership in science and engineering research by leveraging this nation’s unique government-university-industry R&D ecosystem.

Public-private partnerships that support fundamental research drive technology development and economic growth by leveraging investments and expertise in areas of mutual interest to government, industry, and academia. They bring pressing, real-world, “use-inspired” challenges to university researchers where fundamental, pre-competitive research may be pursued and contribute to open advances. Such partnerships also build research and workforce capacity by linking faculty and students with industry representatives, industry settings, and industry jobs. Partnerships create the opportunity to leverage industry expertise to accelerate the transition of open and published research results into the marketplace.

NSF and Amazon are providing funding for this research, but there are many other agencies, companies, and other organizations that have listed fairness in AI as a priority, and that stand to benefit from this work. Research funded under the joint Amazon-NSF fairness program is expected to result in open-source tools, publicly available datasets, and publications that are widely accessible.

We reiterate that the NSF-Amazon collaboration protects the integrity of the research and the research teams. For example, as noted in the solicitation:

    • Although Amazon provides partial funding for this program, it will not play a role in the selection of proposals for award.
    • Awards through this program will support fundamental, pre-competitive research with open results. It is expected to result in open-source tools, publicly available datasets, and publications that are widely accessible.

Amazon researchers may be involved with the funded projects as advisors or to host student interns wishing to gain further industry experience — but they may do so only at the request of an awardee or with the awardee's consent. Amazon does not seek, and the awardee should not share with Amazon, any information that the awardee does not intend to share widely with other parties."

--Editor's note: This story has been updated to include NSF's response.

About the author: George Leopold

George Leopold has written about science and technology for more than 30 years, focusing on electronics and aerospace technology. He previously served as executive editor of Electronic Engineering Times. Leopold is the author of "Calculated Risk: The Supersonic Life and Times of Gus Grissom" (Purdue University Press, 2016).

EnterpriseAI