In a clinic exam room, the conversation between a doctor and patient is often brief. A physician has minutes to explain HPV, why it matters and whether a patient should get vaccinated. 

For some, that window is not long enough to untangle fear, misinformation or hesitation. At the Viterbi School of Engineering, researchers are trying to stretch that moment by bringing in a new kind of assistant, one that doesn’t require additional provider time.

“Chat-HPV: A Conversational Health AI Tool for HPV” is a project developed by USC faculty to address common misconceptions about the HPV vaccine while easing the burden on clinicians. Designed as an interactive tool, Chat-HPV aims to simulate a back-and-forth conversation, allowing patients to ask questions and receive information regarding HPV vaccines.

“We’re working with clinicians and experts in the HPV vaccine and doing a combination of a chatbot language model and structured questions and answers,” said Jonathan May, one of the creators of Chat-HPV and a research associate professor of computer science. 

May said ensuring accuracy is critical, especially as artificial intelligence becomes more involved in health communication. 

“We’re trying to be extra careful and guide the patients initially toward kind of common questions that they might have, even prewriting the answers, so we’re not even using a chatbot, then it’s just basically a script. That’s the safest of all,” May said, “We’re going to be monitoring the outputs … we’re going to test it extensively and try a bunch of different kinds of questions.”

In a clinical setting, patients would be given a tablet that introduces information about the HPV vaccine, along with suggested prompts and the option to ask their own questions.

HPV is the most common sexually transmitted infection in the United States and is linked to several types of cancer, with over 40 types of HPV that can infect someone, according to the Centers for Disease Control and Prevention.

However, despite the availability of vaccines, uptake remains uneven, particularly with limited access to consistent care. According to Dr. Mona Guo, the lead researcher and a gynecologic oncologist at Keck Medicine of USC and Los Angeles General Medical Center, the issue is often tied not just to availability but to gaps in education and awareness.

Angie Ghanem-Uzqueda, an epidemiologist at the Engemann Student Health Center, said HPV presents a unique challenge — the virus often appears without obvious symptoms and can take years to develop into more serious conditions. This makes it harder to tell younger populations how to prevent it. 

The lack of knowledge contributes to common misconceptions among students, which may cause people to miss the window to get the vaccination. Ghanem-Uzqueda said that regardless of whether or not someone is currently sexually active, the vaccine is always a good preventative measure for the future.

Guo said her experience working with underserved patients at L.A. General Medical Center helped shape the motivation behind Chat-HPV, particularly in clinics where physicians are often stretched thin.

“One of the things that is the best predictor of uptake of [the] HPV vaccine is physician recommendation,” Guo said, “But the clinics at L.A. General are just so busy that it’s really hard … this was a way of thinking [about] ‘how do we leverage technology and leverage AI to try to take some of that burden off of the physician and the resident …  but still give that counseling piece to the patient.’”

Researchers said the goal is not only to inform patients, but also to make clinical conversations more efficient. 

“We’re trying to lower the time that physicians spend counseling on the HPV vaccine by 25%,” May said. “We’re also hoping to get an increase of 10% of vaccine initiation.”

Guo said that the tool is meant to support, instead of replace, the relationship between patients and providers.

“It’s not deciding anything for you. It is providing background information and answering common questions that someone may just ask their physician.” Guo said. “The ultimate decision to get the vaccine is still going to be a conversation between the patient and the provider … but this will hopefully help smooth over and make [that conversation] more efficient.”

As the project moves toward clinical testing and is expected to start working with real patients in the next one or two weeks, researchers hope it can serve as a model for how artificial intelligence can be integrated into patient education without compromising trust or care.

“We want to see … ‘Can we integrate these kinds of technologies into the patient experience to improve their overall health journey?” May said. “If it’s the middle of the night and something happens, you can have a reliable source to ask questions to, even if you can’t get a human on the phone right away.”