UVA Engineering’s Hongning Wang Creates AI-powered Technologies Ready to Reduce Stress
Global attention turned to virtual health care in 2020, when some patients were unable to visit their doctors in person during the pandemic. Wired Magazine declared that “artificial intelligence is already powering change in the pandemic’s wake.” Other news outlets weren’t so convinced. They pointed to the newness of the technology and asked whether artificial intelligence, which powers many telehealth platforms, was mature and secure enough.
While the debate unfolded in tech media, one University of Virginia School of Engineering researcher and his collaborators were already years into their research to develop reliable AI that could broaden access to mental health services and serve populations that don’t have ready access to medical care.
Hongning Wang, associate professor in UVA Engineering’s Department of Computer Science, is co-principal investigator with John A. Stankovic, BP America Professor of computer science and director of UVA Engineering’s Link Lab for cyber-physical systems, on a four year, $1.2 million National Science Foundation-funded project to develop a personal assistant for caregivers of dementia patients.
The UVA Engineering researchers are teamed up with Ohio State University nursing professor Karen Rose, a gerontological care expert, and Kristina Gordon, a professor of psychology at the University of Tennessee.
Their AI-powered device detects a caregiver’s rising stress levels, then intervenes to suggest an anxiety reducing strategy. Because dementia care is overwhelmingly provided by family members or unpaid helpers, the researchers believe they can help this underserved population.
“In working with my psychology and nursing colleagues, I learned that little emphasis was being placed on the actual caregivers, who themselves need access to mental health care and support. This was a real opportunity to help a vulnerable population with technology,” Wang said.
Since the grant award in January 2019, the UVA team has made steady progress. The prototype was developed in the first year, and attention turned to testing the device at the start of 2020.
Baseline data would be gathered from test sites with a dementia patient and in-home caregiver. The goal was to monitor vocal patterns in conversations between the caregiver and patient. Vocal pattern datasets could then be analyzed to continuously improve the algorithms that inform the device how to detect rising stress and intervene.
Researchers were about to travel to the test sites and set up data capture equipment just as stay-at-home orders were enacted in response to the pandemic. Even in the face of this challenge, Wang and his team were committed to moving forward.
They opted to safely gather vocal pattern data in their on-Grounds lab by having researchers re-enact agitation states until they could mail out device test kits to the patients’ homes and virtually collect data.
Against the backdrop of an escalating worldwide health crisis, it was becoming obvious how a technology like this could have much broader implications. COVID-19 had millions of people sheltered in place, not knowing when the isolation would end.
By mid-April, the New England Journal of Medicine was sounding an alarm about mental health issues stemming not only from distress surrounding COVID-19 itself, but from the “mass home-confinement directives.”
The journal concluded that patients’ mental wellbeing “may benefit from supportive interventions designed to promote wellness and enhance coping” as a supplement to telemedicine.
Wang’s technology suddenly became the perfect tool for offering supportive interventions to the millions stuck at home and stressed.
“It’s not surprising that breakthroughs in one area lead to evolution in others,” Wang said. “The thought is that these applications can be applied to new situations. This is central to the computer science way of thought. You apply the micro breakthrough to the macro environment. A device like this could help even more people as health care moves toward telemedicine.”
“Hongning is advancing our understanding of how humans and computers interact. This will enable computing to make life better for so many in our society,” said Kevin Skadron, Harry Douglas Forsyth Professor and chair of computer science. “The mental support that might otherwise not have been available to vulnerable populations, but now will be through this device, is another example of why his research is so critical.”
Ultimately, those who are concerned that AI is not ready for this type of demand have a point. Accuracy will be a defining factor in creating AI-powered technologies that can credibly stand in for humans and create a supply of health care services that otherwise might not exist. Wang has been working to make this possible his entire career.
He analyzes human behavior to develop the algorithms that will be able to satisfy the human’s need.
In 2016 Wang earned a CAREER Award from the National Science Foundation to work with companies like Yahoo, Snapchat and Bloomberg. He would analyze, then improve, how accurately the software could infer the user’s intentions to make the person’s experience with technology more intuitive.
The idea is to continuously improve the algorithms by observing and learning from human interactions with technology, similar to the way humans learn to interact with other humans over time. Large data sets of human-computer interactions help create algorithms that enable AI to act human.
This year, Wang received a Google Research Award to analyze human interactions with the Google search engine. He will investigate tradeoffs that exist between users’ privacy and their need for accuracy when using the search engine. The data will improve the underlying mathematical models so that algorithms can deliver the same level of accuracy even when users keep their searches private.
“Things as universal as search engine results being accurate are directly related to smart systems in health care. The research regarding information retrieval from Google gives us knowledge from human behavior on how to program Google’s algorithms to make the smartest suggestions,” Wang said. “There is a similar paradigm in the health domain. The algorithms for AI-powered health devices can be improved over time to make the best recommendations.”
The research will contribute to algorithms that not only provide the smartest suggestions but have built in privacy standards. Both features are critical for AI-powered health care devices that are capable of transforming health care.
“Sophisticated and more advanced algorithms are needed in engineering and computer science to create cyber technologies that authentically interact with humans,” said John A. Stankovic, BP America Professor of computer science and director of UVA’s Link Lab. “It has been a pleasure to collaborate with Hongning on this cyber-physical systems project that will help at-risk populations manage their stress.”