Artificial intelligence chatbots may be making healthcare more accessible than ever, but when it comes to mental health treatment, there are still significant limitations, according to University of Texas at Dallas researcher Dr. Ryan Raimi.
His research examines how people interact with chatbots in health care settings and why some users perceive AI systems as more judgmental than human providers.
"It seems like there are three main reasons...or predictors leading to this elevated perception of judgmentalness regarding chatbots," Raimi said. "First of all, the AI agents lack real-world experience."
But he says the challenge goes deeper than simply processing information.
"The second part is a deep profound understanding which, in turn, can be divided into social understanding and basic emotional understanding," he said. "You know what it means to be isolated and lonely as a human being, whereas an AI has absolutely no comprehension of what that actually means."
Raimi also noted that many people seeking mental healthcare are not necessarily looking for solutions.
"They’re not looking for a solution perse, he said. "They just want to be heard."
Despite those limitations, Raimi sees substantial promise for chatbots in certain areas of healthcare.
"When it comes to triage and screening, it's a pretty straightforward structure," Raimi said. "It's accessible around the clock at negligible costs."
He said AI tools could be especially valuable in areas where traditional mental health care services are scarce.
"You can have access to these agents around the clock 24/7 without virtually any cost associated with it and you can get recommendations and suggestions that are very relevant to your situation," he said.
Raimi's research has examined both screening and treatment interactions with AI systems. What they found is that creating meaningful interactions remains difficult.
"It's very tricky when it comes to empathizing with the subject, with the client," he said. "Sometimes you have to let the machine act like a machine, and other times it needs to emulate human behavior, so it's very intricate and nuanced."
For now, Raimi believes one of the strongest use cases for AI in health care is helping patients navigate the front end of the system.
"On the screening side, it's getting pretty close to being deployed globally," he said.
The technology could also help clinics that are struggling with staffing shortages.
"A lot of clinics may be understaffed, they don’t have enough personnel to do the screening and triage," Raimi said. "These agents can be scalable and you can run at almost no cost. That's a game changer."
Ron Corning is a host of KERA's NTX Now. Got a tip? Email Ron at rcorning@kera.org.
KERA News is made possible through the generosity of our members. If you find this reporting valuable, consider making a tax-deductible gift today. Thank you.
