Ph.D. fellows from GW’s Designing Trustworthy AI Systems program are gaining firsthand knowledge about convergent problems in artificial intelligence and machine learning.
By Kristen Mitchell
A group of George Washington University Ph.D. fellows are spending the summer exploring what it means to tackle some of the world’s biggest challenges at the intersection of computer science and systems engineering, where questions about artificial intelligence and machine learning implementation could affect the lives of everyday people.
Seven fellows from the GW School of Engineering and Applied Science are participating in a 10-week research formulation summer boot camp. The boot camp is part of the Co-Design of Trustworthy AI Systems (DTAIS) program, supported by a $3 million grant awarded by the National Science Foundation Research Traineeship Program last year. The program brings together faculty from the SEAS Department of Engineering Management and Systems Engineering (EMSE) and the Department of Computer Science (CS) to develop a transformative model for graduate education training that combines research in computer science and systems engineering to support the next generation of leaders and researchers exploring how artificial intelligence (AI) is used in real-world contexts.
Zoe Szajnfarber, DTAIS principal investigator, professor and chair of the Department of Engineering Management and Systems Engineering, said today’s students must be prepared to take advantage of the power of AI while mitigating potential risks and embedded bias.
“To tackle the messy, in-between problems that are so critical for deploying systems that can take full advantage of the opportunities for AI to transform society for good, without accidentally introducing some of the scary risks that can come out of that, we want the fellows to both dive deep in their chosen field and also have enough appreciation of what the other discipline(s) do to bridge them” she said. “That's a really big challenge, both to have the requisite knowledge to even speak across the disciplinary boundaries, but also to figure out what these problems are.”
Erica Wortham, director of the SEAS Innovation Center, is the boot camp’s lead instructor. Robert Pless, computer science department chair and the Patrick and Donna Martin Professor of Computer Science, Rachelle Heller, director of the SEAS Center for Women in Engineering and professor emerita of computer science, and Ekundayo Shittu, EMSE associate professor, are co-investigators of the DTAIS program. Additional affiliated faculty include Tim Wood, CS associate professor; Arkady Yerukhimovich, CS assistant professor; Erica Gralla, EMSE associate professor; and David Broniatowski, EMSE associate professor.
The boot camp was designed to take the program’s first cohort of fellows out of their comfort zone to explore areas of critical importance and delve into the types of exciting in-between challenges that could spark ideas for future research, Szajnfarber said. It is inspired by human-centered design, which places a strong focus on solving the right problem. For that reason, fellows and faculty mentors met with experts from three distinct organizations, representing three pillars of research areas the team is interested in contributing to. This provided an opportunity to explore concrete examples demonstrating how individuals and entities are wrestling with decisions about how AI and machine learning tools can transform work in different application areas.
The fellows met with representatives from Comcast, who came to the Foggy Bottom campus to talk about the AI applications used in technology like voice-assisted TV remotes and home security cameras. The fellows also visited the MITRE Corporation’s federally funded research and development site in Tysons Corner, Virginia, where they were able to tour labs focused on autonomous vehicle and air traffic control experimentation.
Lastly, the fellows visited with members of the Virginia Task Force 1, a domestic and international disaster response resource sponsored by the Fairfax County Fire and Rescue Department. They spoke to experts, including GW associate professor Joseph Barbera, to learn about the different elements that go into a successful search and rescue situation and potential opportunities for AI to support their operations.
“Having these three different contact points with industry, we had time with researchers who are considering how or whether to introduce AI into their systems, as well as experts who are already using it and can talk about the problems they're experiencing,” said Cody Dunn, program manager for DTAIS.
Following those exposures, fellows met to discuss the challenges presented and what they learned. They spent hours drawing connections between the different interactions and mulled over the potential solutions they might be able to uniquely contribute to.
Dan Koban, a Ph.D. student in systems engineering, said getting exposed to diverse use cases for AI and talking through the different challenges with other fellows has been an insightful experience.
“It's great to see practical applications of all of these things that we study,” he said. “It gives you a really useful context for understanding what's really going on.”
David Balash, a Ph.D. student in computer science, said the NRT program has been helpful to facilitate a space for bouncing around ideas with fellow students and to think about new research areas to explore.
“Having that cohort or that community has had a really good impact,” he said. “It's given me an opportunity to expand the people that I'm working with as well.”
Fellows will spend the remainder of the boot camp honing their research area and working collaboratively to refine their ideas with each other and faculty. Ideally, the fellows will finish the summer with a set of practices for conducting interdisciplinary research that will serve them throughout the rest of their time at the university. The boot camp, like the overall DTAIS program, emphasizes the importance of joining diverse perspectives to build a research ecosystem centered around solving convergent challenges.
The tipping point of AI
AI tools have been developed at an unusually rapid pace over the past decade. They’ve been deployed into environments where value maximization precedes regulation, and these tools impact every realm of life.
“In the last 10 years we've gone from thinking that computing and AI will solve many of the world's problems to realizing also that it's creating a whole new set of problems,” Pless said. “We need to build the engineering to try to address and attack those problems.”
This current moment is a tipping point, Szajnfarber said, representing both huge opportunities and significant risks.
Researchers in computer science and systems engineering tend to confront challenges in different ways. By bringing these disciplines together, the Co-Design of Trustworthy AI Systems program seeks to inspire fellows to confront questions with non-traditional solving approaches. Different ways of thinking can be critical to minimizing harm and ensuring systems are trustworthy, equitable and fair.
The school’s “engineering and” philosophy emphasizes the societal context in which engineering and computing are practiced—a key understanding that will help prepare students to become leaders and cultivators of trust in an increasingly AI-driven world. Through collaboration across SEAS departments and GW at large, and by forging connections with external research and industry leaders only possible in Washington, D.C., the program is well-positioned to meet the challenge of preparing tomorrow’s leaders in the field, Pless said.
“D.C. is a place where there's so many people trying to work hard to solve problems,” Pless said. “That’s what this is.”