As the use of AI therapy chatbots increases, so do the potential dangers
Will AI help close the mental health gap, or will it do more harm than good?
You may have heard of AI interviewers, but what about AI doctors and therapists?
Ty Eveland’s videos of him “trolling” technology have racked up millions of views on TikTok and sparked conversation about the challenges of AI in healthcare.
“If you watch my videos, it’s clear that doctors and therapists with artificial intelligence don’t know what they’re doing and can’t respond to people in crisis the same way a real person would,” Eveland told USA TODAY.
Eveland says he doesn’t pay for any certified medical software for his videos. Instead, he used a number of AI platforms that allow AI to act as doctors, therapists, and other professions. An artificial intelligence company was recently sued for allegedly misrepresenting its companion bots as licensed medical professionals.
The 23-year-old New York-based actor filmed himself feeding an AI agent a variety of fake and outlandish scenarios, including asking how to get rid of a bowling ball-sized pimple and saying he thought he was pregnant after spitting out a pacifier.
“They definitely match the energy that you give. Obviously, all my videos are trolls and I’m cocky. But they really want to spit it back on me right away, too,” he laughed. “This shows that they are not experts, because professional doctors would not have the cocky attitude I described.”
In the comments section, viewers agree that AI agents often sound annoying, rude, and even condescending.
Eveland recognizes that real-world healthcare platforms that use AI may work differently than the programs he uses in his videos, but his experience still makes him wary of the technology.
“I’m never going to see an AI doctor,” he declared. And several of his commenters shared similar sentiments, with one person writing, “To tell you the truth, AI doctors are pretty scary.”
Human doctor reacts to AI doctor’s troll video
Actual human doctors say the video makes some good points about AI’s flaws.
Dr. Alan Cohen, chief of pediatric neurosurgery at Johns Hopkins Children’s Hospital and author of Comfort Always: Healing in the Age of Technology, said these videos do a good job of using humor to illustrate some of the shortcomings of AI in the doctor-patient relationship.
“If not managed properly, AI tends to drive a wedge between doctors, patients, and families,” Cohen told USA TODAY.
Cohen explains that when he started working in pediatric neurosurgery 40 years ago, computer technology didn’t even exist. He believes AI should be embraced as a useful tool, but remember that medicine requires nuance: “Computers don’t know.”
“If we forget that part, the role of the bedside physician, building relationships with children and families, then we can’t cure patients. Technology treats the disease, humanity cures the patient,” he said, adding that human kindness, empathy, compassion and restrained optimism “can be imitated by machines, but they cannot be replaced.”
When it comes to therapy, young people and adults alike are turning to ChatGPT and other AI chatbots to vent and open up.
Relationship therapist Christina Tracy Stein says that while some clients have been consulted by chatbots, the AI often lacks all of the patient’s context and history, making it impossible to truly understand what someone is sharing.
“I think it’s dangerous to label someone with the beliefs of an AI therapist, because when people are vulnerable, they take it on. That’s what they believe,” she said.
Additionally, she explained that a lot of being a therapist is not feeding back exactly what was said, but feeding back underlying feelings.
“We don’t know if AI can interpret the energy or emotion behind the actual words. It just takes the words and returns an interpretation,” she explains, adding that it can miss important nonverbal cues, such as someone’s tone or whether they are crying. “They don’t know that there’s more to what is being said.”
Are there any benefits to AI in healthcare?
Experts are warning against relying solely on AI for physical and mental health concerns, but that doesn’t mean the technology is all bad.
For example, for people who don’t have access to therapy or are really struggling with loneliness, Tracy Stein says, “Sometimes something is more helpful than nothing.” However, don’t take everything an AI chatbot says as fact, including diagnoses.
“There’s some value in having predictable, consistent interactions like this, but because they’re one-sided, you can’t hold people accountable for their actions,” she explained, suggesting instead using it to organize your thoughts or have a brainstorming session.
“The ultimate decision about what’s going on should be made by the person or the actual therapist, not the AI,” she says.
Cohen agrees that while AI is scary, it can also be a useful tool, especially given the rapid pace of its emergence. He uses it in his practice to aid in diagnosis and to compare large databases of brain tumor scans.
“AI is very good at processing big data quickly, but it cannot replace human touch or healing techniques,” he said.
