Joining forces with Deloitte to tackle wicked problems
As artificial intelligence (AI) permeates more of daily life — from online searches to hiring decisions to medical diagnoses — it generates greater social and scientific challenges than it ever resolves. How do scientists keep bias from creeping into AI algorithms, for instance? How do they build an AI pipeline that people trust? Do humans correct for bias or unintentionally insert more?
That devilish knack for defying easy answers makes AI what researchers and policy makers often refer to as a "wicked problem" — a challenge so complex that there is no single, simple solution.
Wicked problems are still worth addressing. One approach is Virginia Tech’s aptly named Problem Day, a semiannual gathering under the auspices of the Data and Decisions destination area that brings together Virginia Tech researchers with employees from the multinational professional services firm Deloitte. Representing disciplines as diverse as agriculture, business, computer science, economics, engineering, political science, sociology, and statistics, the group met virtually on Nov. 11 to discuss thorny issues related to artificial intelligence.
Problem Day originated in a desire to increase collaboration between Virginia Tech and Deloitte, a longtime corporate partner and contributor to the university. Five years ago, then–dean of the College of Science Sally Morton posed a question to Virginia Tech alumnus Robert Torpey, senior manager in the analytics and cognitive practice within Deloitte Consulting LLP. “Have you heard of the concept of a wicked problem?” she asked.
Torpey had. In the corporate world, to turn seemingly intractable wicked problems from challenges into opportunities required getting the right people at the table. That became the model for the first Problem Day, held in April 2018. “The whole intent was, 'How do we bring Deloitte and Virginia Tech thought leaders together to solve common problems?'” Torpey said.
While other Problem Days have focused on the wicked problems of blockchain or health care analytics, the sixth Problem Day, on Nov. 11, centered on artificial intelligence. Daniel Sui, vice president for research and innovation, welcomed the group of more than 40 participants. Then Deloitte researchers presented their findings on several key questions related to wicked AI problems:
- How can we investigate AI creation and decisions while establishing safeguards for evaluating whether that use reflects the lived realities of stakeholders?
- How do we resolve the changing rules and regulations of AI technologies, such as data privacy, facial recognition, and decision transparency?
- Can we use explainable AI to bridge the gap between interpretability and power among AI solutions?
Each question prompted robust conversation.
“Models don’t have bias,” asserted Tom Woteki, director of the Academy of Data Science and head of the Data Analysis and Applied Statistics master’s degree program at the Virginia Tech Research Center in Arlington. “The choices humans make introduce bias.”
In addition, bias can be baked into the data, so that AI algorithms unintentionally amplify that bias, said Hoda Eldardiry, associate professor in the Department of Computer Science.
Other researchers shared their own relevant work. For instance, Sylvester Johnson, director of the Center for Humanities, and Bill Ingram, assistant dean of University Libraries, discussed some of the challenges they faced using AI to search newly digitized records at the National Archives.
Torpey hopes the discussion leads to ongoing collaborations between Deloitte and Virginia Tech faculty, like one that developed after the spring 2020 Problem Day. A discussion around health analytics prompted Amy Pruden, a professor of civil and environmental engineering, to partner with Deloitte on a $3 million National Science Foundation grant proposal, which was awarded in 2021.
Deloitte hosts one of the largest corporate alumni chapters developed at Virginia Tech, with a passionate base of more than 1,000 Hokie employees, called “D’Hokies” for Deloitte Hokies. Their contributions across campus include a partnership between Deloitte’s Artificial Intelligence Institute and the Ted and Karyn Hume Center for National Security and Technology on a program that aims to prepare graduate students for AI careers. Direct access to Deloitte professionals and hands-on research experience equips them with enhanced professional and technical skills.
Problem Day is just one example of how Deloitte is engaging on campus. “From graduate assistantships with the Hume Center, to support of GBAC, to next frontiers in research, Deloitte is a highly valued strategic partner,” said Brandy Salmon, Virginia Tech’s associate vice president of innovation and partnerships. “We look forward to using activities such as Problem Day as a launch pad for achieving shared goals around technology and talent to make a global impact.”