Even if you don’t consider yourself to be tech savvy, You’re likely to encounter AI on a regular basis, whether it’s scrolling through social media, shopping online, or traveling to new places.
“AI is powerful,” says Yingchen Sun, assistant professor at UNC Greensboro. “It saves effort and money. It’s convenient, but by no means perfect.”
While AI mistakes in some situations may be minor, mistakes in other areas, such as healthcare, can be harmful. Sun is working to reduce some of the errors in AI by leveraging the best of both humans and technology, an area known as human-centered AI.
“Our goal is to improve the AI and avoid repeated mistakes by incorporating people’s feedback throughout the process,” he says.
Early in his career, Sun has already published research results in some of the top publications in his field, including the Journal of Biomedical Informatics and the Journal of the American Medical Informatics Association.
Connecting researchers and clinical trial participants
Sun’s recent research revolves around improving information retrieval in the critical context of clinical trial recruitment.
An estimated 500,000 clinical trials are currently underway. Each year, results from approximately 11,000 clinical trials are published, advancing knowledge and improving treatments.
“When scientists need to develop a new drug or drug, they want to hire or recruit volunteers, but there are many requirements for them to participate in research,” Sun says.
Results from clinical trials are key to advancing science, but researchers often find it difficult to recruit participants. On the other hand, individuals who are willing to participate in research may not know how to do so. One study estimated that fewer than half of those surveyed felt comfortable finding relevant clinical trials.
“Researchers may put flyers in elevators, and patients can see if they’re interested in the flyer and call them,” Sun says. “This approach is highly inefficient.”
Without enough clinical trial participants, science stagnates.
During the COVID-19 pandemic, Sun has launched COVID-19 Trials, an online platform that connects interested people with clinical trial opportunities that fit their background and location. I created a finder.
Potential participants can answer a few questions about themselves, and the platform will generate a list of clinical trial options based on their answers. Additionally, if there are no clinical trials that match a person’s interests, AI provides other similar options.
“If the study is closed or not recruiting new volunteers, we recommend related studies,” Sun says. “this is similar to when You are shopping online and the item is out of stock. The website may recommend related products. ”
The best of both worlds in healthcare
The benefits of Sun’s platform extend beyond matching scientists and clinical trial participants. We also leverage human-centered AI to spot mistakes and improve our platform.
Here’s how it works: After receiving the AI-generated clinical research recommendations, you can review the list and modify your answers to effectively train the AI.
“We want users to be involved in the process. If they feel something is wrong, they can fix it,” Sun said. “Equally important is logging all changes made by users.”
By tracking user feedback, the research team can optimize the platform. In this way, Sun believes it can maximize efficiency and accuracy by combining the best of both humans and AI.
“AI alone is not enough; there is still a lot of room for improvement,” Sun says. “So how do we improve? We collect this feedback and continue to train our AI tools.”
Sun wants to build on these findings.
“In the future, we plan to build on this and develop new tools for other types of public trials, not just for COVID-19 but for other types of diseases,” he says. .
Sun is also active in other areas of research, including building a platform called Evidence Map to help researchers integrate peer-reviewed papers. Sun says he is grateful to have a computer science department where his colleagues are friendly and the students are highly motivated.
“We have a lot of local students from Greensboro. It’s fun working with them,” he said. “Students here really want to learn.”
Rachel Damiani’s story
Photo by Sean Norona, University Press
Dive into computer science