This year, the Stanford Institute for Human-Centered AI (HAI) has awarded $2 million in seed grants to 25 research teams that support early, innovative, and interdisciplinary research into future AI technologies designed to benefit the human condition. HAI also has contributed $500,000 toward a first-time grant partnership with the Stanford Center for Artificial Intelligence in Medicine and Imaging (AIMI) that will provide an extra $1 million to six teams for health care research.
The teams funded this year include researchers from all of Stanford’s seven schools — approximately 20 departments — focused on AI that is inspired by human intelligence, augments people rather than replaces them, and provides widespread social benefit that can be broadly shared.
Each of this year’s 25 winning seed grant proposals are novel and highly promising, says James Landay, Stanford professor of computer science and an HAI associate director.
“We’re trying to get new, innovative research off the ground,” he says. “Larger funding sources usually require researchers to show that some work has already been done to prove the viability of an idea. These seed grants help people get their research to this next stage.”
2021 Trends: Bias, Health Care
This year’s winning seed grant proposals focus largely on bias/fairness research, cognitive science, and health care. They include the following:
- An investigation by researchers in communication and computer science into the disparities in race and gender representation in online advertising. Biases in AI systems can result in advertisers irresponsibly targeting particular groups, which can perpetuate or exacerbate existing social inequity.
- A project by researchers in engineering, psychology, and computer science to develop computational models of how humans combine touch, visual, and auditory information to make causal inferences, which could lay a foundation for human-inspired learning techniques in robotic systems.
- A study by researchers in medicine and engineering involving the use of AI in ultra-fast MRI for precision radiotherapy for cancer patients.
In addition to HAI’s standard project evaluation, each submitted research proposal underwent an ethics review by HAI’s newly established Ethics and Society Review (ESR) Board. Piloted in 2020, the ESR requires researchers to evaluate their proposals for any potential negative impact on society before being green-lighted for funding.
“We want researchers thinking about how they’re going to protect society at the earliest stages of research, which is when you’re writing a proposal,” Landay says. “So far the process has been quite successful. The ESR is now looking to see how it can scale up to apply to all kinds of other research at Stanford beyond HAI.”
Focusing on Interdisciplinary Work
The winning proposals represent research collaborations across a wide spectrum of disciplines, says HAI Interim Research Director Vanessa Parli.
“It’s becoming increasingly apparent that one researcher in a lab doesn’t always have the expertise to make the progress we need right now, so these teams are often made up of people from different schools or perhaps members of the same department who have very different research focus areas,” she says. “The proposals we look for are ones that create partnerships among researchers that are maybe unexpected or surprising.”
The seed grant program has operated since 2018 and to date has distributed $8 million to early-stage AI research.
The HAI-AIMI grants stem from a new program that launched just this year.
“HAI and AIMI are a natural fit,” says AIMI Director and HAI Associate Director Curtis Langlotz, MD. “The hallmark of medical AI research is its interdisciplinary nature. Our co-investment in this program enables us to support six ambitious research projects that reimagine artificial intelligence in health care and address questions from all corners of medicine by talented and diverse teams of scientists. AIMI and HAI working together have enabled the two organizations to multiply our impact on human health.”
The Stanford Department of Pathology has also contributed to the 2021 HAI-AIMI grant program.
Creating a Community
Prior seed grantees have seen successful results, including a 2019 project conducted by researchers in biomedical data science, law, and computer science that developed some of the first efficient algorithms capable of deleting individual data without needing to retrain an entire model. Another 2019 team of researchers in psychology, computer science, and engineering won a seed grant for their work to develop computational models of infant development. They forged such a strong collaboration that they went on to receive a prestigious Hoffman-Yee Research Grant from HAI to continue their research.
It’s that type of interdisciplinary cooperation that can bring remarkable results, Landay says.
“This program helps us create a community of researchers who are all working on this idea of a human-centered AI from different angles and directions, not just in computing but in the humanities, social sciences, and medicine,” he says. “To me, the key marker of success here is having faculty collaborating on ideas and working with people they wouldn’t have had a chance to work with before. That’s definitely going on here.”
See all the recipients.
Learn more about Stanford HAI grants programs.