By Faye Beard
As of late, artificial intelligence has experienced some real-life setbacks in the mental health sector. The veteran AI-powered therapy app Woebot was deactivated. Wysa, “the clinical alternative to ChatGPT,” has rebranded to include in-person therapeutic services. And other therapy chatbots, including Ash by Slingshot AI, have been banned in Nevada and most recently in Illinois.
“There’s a lot that can go wrong if you’re using ChatGPT as a therapist,” said Lauri Goldkind, Ph.D., MSW., and the editor-in-chief of the Journal of Technology in Human Services. “It’s prone to hallucinations, which means it makes mistakes. The chatbot is predisposed to agree with you, and that is not therapeutic,” she said. “It requires a level of discernment and critical thinking to know that the output is accurate. We need a set of regulations to protect both the professionals and clients.”
Members of state chapters of the National Association of Social Workers have been fending off so-called “therapists with no credentials, oversight or accountability.” That’s how NASW-Nevada described the unregulated AI platforms flooding the market. In June, the Assembly Bill 406 established “strong protections against the unethical or unsafe use of artificial intelligence in clinical settings, ensuring that licensed professionals, not machines, deliver care,” according to NASW-NV. Simply put, “AI cannot provide therapy or mental health care.”
An NASW-Illinois initiative became the Wellness and Oversight for Psychological Resources Act. Many observers called the landmark legislation, which prohibits anyone from using AI to provide mental health and therapeutic decision-making, the nation’s most restrictive law yet. Recently, California, Massachusetts, Rhode Island, Texas, and Utah have all passed bills regulating AI in mental health services.
Seems like a win for social workers—for now.
“Ready or not, here AI comes!” said Karen Magruder, DSW, LCSW-S, NASW-TX, who has researched the ethics and implications of AI in social work education. “We should be approaching AI adoption with our eyes wide open to its drawbacks, limitations, and risks.” She also noted that there should be informed consent, transparency, and the ability for patients to opt out of AI-assisted services.
For some, there are benefits of AI in therapy. According to the World Economic Forum, about 85% of people with mental health issues go untreated often because of provider shortages, and AI can help alleviate those shortages by complementing human providers rather than replacing them.
“Preliminary studies suggest that AI-powered chatbots may help alleviate symptoms of anxiety and depression,” according to a report published in the National Institute of Health titled, Can Al replace psychotherapists? Exploring the future of mental health care.
“I think we underestimate how lonely people really are,” Dr. Goldkind said. “As a profession, we underestimate how much harm therapists have done intentionally or unintentionally that someone would feel safer talking to a robot.”
Hesitation is valid, explained Dr. Magruder. “Resistance often comes from a place of care, like concern for client privacy or fear of losing the human touch,” she said. “Some degree of resistance keeps us vigilant to overreliance or inappropriate uses. Critiques might even create opportunities for advocacy.”
Dr. Goldkind encourages social workers to be engaged in the design process of AI and consider, “How do we make sure there’s no bias in the model? Are we upholding human dignity?” She is currently working with colleagues across the country to create digestible clinical practice guidelines for AI, which would complement the NASW Code of Ethics and Standards for Technology and contextualize the recent legislation. Lastly, Dr. Goldkind offered her expert opinion: “It’s important to say that AI is not able to serve as a therapist at this point in time.”
Faye Beard is a freelance writer living in New York City.
