It's scary to share how you feel. Friends and family can judge, and therapists can be expensive and difficult to get. That's why some people are turning to ChatGpt for mental health help.
Some praise AI services for saving lives, but others say the lack of regulations around it can pose a risk. A psychology expert at Northeastern said there are safety and privacy issues brought about by someone open to an artificial intelligence chatbot like ChatGpt.
“I think AI is truly exciting as a new tool with a lot of promise and there are many applications for delivering psychological services,” says Jessica Hoffman, professor of applied psychology at Northeastern University. “It is exciting to see how things are unfolding and explore the possibilities of supporting psychologists and mental health providers in our work.
“However, when you think about the current situation, there are major concerns about the limitations of ChatGPT to provide psychological services. There are real safety concerns that people need to recognize. ChatGPT is not a trained therapist. It does not comply with the legal and ethical obligations at which the mental health service provider works.
Cons
It's easy to see the appeal of being confident in a chatbot. Northeast experts say therapists are expensive and difficult to find.
“There's a shortage of experts,” says Hoffman. “There are barriers to insurance. There are real issues in rural areas where there are more shortages. You can reach out to your computer and get support.”
The chatbot also functions as a listening ear.
“People are lonely,” says Josephine Au, clinical professor of applied psychology at Northeastern University. “People don't just look at ChatGpt (like a universally-generated AI tool) for treatment. They are looking for dating, so sometimes they naturally evolve into a cure-like conversation.
However, Au says that these forms of artificial intelligence are not designed to be therapeutic. In fact, these models are often set up to validate the user's thinking. This is a problem that poses serious risks to those dealing with delusions and suicide ideas.
There have been cases of people who died from suicide after getting guidance on how to do this from an AI chatbot. One of them prompted a lawsuit. In addition, there are an increasing number of reports of hospitalizations due to “AI psychosis.” In this report, there are mental health episodes where people were triggered by these chatbots. Openai added guardrails to ChatGpt after realising it encourages unhealthy behaviour.
The American Psychological Association warned against using AI chatbots for mental health support. A Northeastern study found that people can bypass the guardrail of the language model and use it to get details on how to harm themselves and how to die by suicide.
“I don't think it's a good idea at all for people to resort to non-treatment platforms as a form of treatment,” says Au. “We're talking about interactive tools designed to agree and validate. There's a risk of preferring what data is generated through such conversational patterns. Many LLM tools are designed to reinforce problematic beliefs about themselves.”
This is particularly relevant when it comes to diagnostics. The AU says people think they have a particular condition, they ask ChatGpt and may get a “diagnosis” from their self-reported symptoms thanks to the model's mechanisms.
However, experts in the Northeast say many factors depend on getting diagnosed, such as examining a patient's body language and looking at their lives more overall as they develop relationships with the patient. These are things that AI can't do.
“It feels like a slippery slope,” says Joshua Curtis, assistant professor of applied psychology at Northeastern University. “ChatGpt has five of these nine depression symptoms and what a human diagnoser does is say, “OK, it sounds like depression,” a structured clinical assessment. They ask many follow-up questions to support (you've had one). This justifies the fact that this is suffering serious interference in your life, as it takes hours from your work.
Next, there are privacy concerns. Although clinicians are bound by HIPAA, chatbots do not have the same restrictions in terms of protecting the personal information people may share. Openai CEO Sam Altman said there is no legal confidentiality to those using ChatGpt.
“Guardrails are not safe for the kind of confidential information that is revealed,” Hoffmann says of people who use AI as therapists. “People need to be aware of where their information is heading and what happens with that information. When I think about training psychologists in Northeastern, I'm very aware, like a really big problem in psychology practice.”
professional
While artificial intelligence poses risks when used by patients, experts in the northeast say certain models are trained in the right way and may provide appropriate privacy safeguards to help clinicians.
Curtiss, a member of the Northeastern's Institute for Cognitive and Brain Health, says he has done a lot of work in artificial intelligence, particularly machine learning. He has conducted studies using these types of models to find that they can help predict treatment outcomes for specific mental health disorders.
“I use a lot of machine learning in predictive modeling where users say more about what's going on, as opposed to larger language models like the common language models we use,” Curtiss says.
Northeastern's Institute for Cognitive and Brain Health is partnering with experiential AI partners to see if they can develop therapeutic tools.
Hoffman says he also sees the possibility that clinicians will use artificial intelligence if necessary to improve their practice.
“It could help with evaluations,” says Hoffman. “It could be a useful tool used by clinicians to assist in intake and assist in assessments to guide a more personalized plan for treatment. However, it is not automatic. Surveillance must be provided to trained clinicians, and it must be done on a safe and secure platform.”
For patients, Northeastern experts say there are positive uses for chatbots that don't need to be used as therapists. For example, the AU says these tools can help people come up with ways to summarise their ideas and continue with the specific practices that clinicians propose for their health. Hoffman suggests that it could also be a way for people to connect with providers.
But overall, experts say it's better to find a therapist than to resort to chatbots that aren't designed to function as a therapeutic tool.
“There's a lot of concern, but there's a lot of hope,” says Au. “The key agents in the commercialization and monetization of mental health care tools are primarily technology people, venture capitalists, and researchers who lack clinical experience, clinicians and patients who understand what psychotherapy is, and people who are not practitioners for patients. Some users argue that these tools are in order to reduce the sense of isolation and loneliness (some users argue that this tool is really useful.”
