Just as AI feels like it is slowly crawling into our daily lives in every corner of the internet, it manages to become a new, trendy “life hack” for therapy. From Snapchat My AI, Chat GPT, Google Gemini, to Woebot, the opportunity to chat about personal struggles and get immediate help are limitless and for the most part it does help. AI offers good advice for certain minor situations yet also poses dangers for users relying on it for therapy.
To show what a therapeutic conversation with AI could look like , I created a scenario to test how well it responds. For clarification, I am not diagnosed or ever have been diagnosed with anxiety. I have done research on typical things an anxious person may say and have tried to include that to the best of my ability in my prompt for Chat GPT, but the objective of it is to just observe what it told me. Here is what I told Chat GPT:
“I just failed a math quiz. I did absolutely terrible and I feel like now I’m going to fail the entire class and have to retake the course next year. What if I don’t even graduate? I just can’t… I can’t handle this class anymore and maybe I’m out smart enough to do this anymore. What do I do?”
It responded by encouraging me and saying, “I’m sorry you’re feeling this way right now,” and “I want to reassure you that this moment does not define your ability or your future.” Chat GPT then gave me a list of things that would help and descriptions for each of the following:
- Take a breath
- It’s one quiz, not the whole class
- Talk to your teacher
- Analyze what went wrong
- Reach out for help
- Believe in yourself
Then it finished by saying, “you’re not alone in struggling,” offered to help with the material, and said, “You’ve got this, and you’re not giving up.” It didn’t necessarily give me tools specifically related to the anxious tone of my prompt but did give a lot of helpful tools to calm down and address the situation. So, from this little experiment I found that Chat GPT provides its users with tools to help them figure out what to do next, calm down, encourage and motivate the person as well. It does not diagnose its users but makes an effort to handle the current situation.
According to the Washington Post, the Substance Abuse and Mental Health Administration has estimated that in 2023, 6.2 million people with mental illness wanted but didn’t receive proper treatment. The use of AI for therapy offers cheap and affordable advice to users 24/7. AI chat boxes are programmed to spit out basic CBT (Cognitive Behavioral Therapy) skills, yet “It can be misleading and potentially dangerous, because while it’s not rocket science, those skills, it is an art and science as to when to use which ones and how to practice doing that,” Dr. Natalie Zervas, Director of Counseling and Wellness Education, told me.
“Initially I, and many of my colleagues, were very wary… One branch of that is: ‘Are robots going to replace our jobs?’… and two, there’s a worry about misuse” because it is hard to quantify the human factor of therapy, Dr. Zervas explained. When you rely on AI for therapy, you are relying on a computer which has dangers in and of itself because it is not approved by the FDA, does not have reliable confidentiality, and does not follow ethical legal duty. Let’s break that down:
Dr. Zervas reflected, “The whole world of AI feels new and unmonitored… If we’ve learned anything over the past twenty plus years it’s that new technology can be very helpful and it can also get away from people very quickly” and in the case of AI, it already has. According to the Washington Post article linked earlier, a fourteen-year-old boy recently died by suicide after relying on AI for help, and his parents sued the company, alleging that AI further deteriorated his mental health. This case sparked the concern that chat boxes have encouraged self harm, eating disorders, or violence. Yet it is important to note that this does not happen on every platform for AI and that recent measures have been taken, depending on the website and user prompts, so that the AI will connect its users to Suicide Hotlines once the algorithm recognizes extreme situations. This does not mean that it will work in a given situation or the user will call the given hotlines to get further help.
AI is also a questionable source for therapy because of its “imperfect” confidentiality. Dr Zervas noted that it is important to understand that AI can be helpful because people are more willing to share things with a computer or therapist over Zoom than to a person sitting in front of them who has the ability to judge them. Yet that vulnerable information the person just shared isn’t entirely private. Chat GPT engineers and researchers review conversations and AI responses to better improve the platform, so while AI will not share your information or conversations, it isn’t confidential either. “I always want consumers, in this case patients, to be fully aware of what they’re signing on for,” Dr. Zervas adds. When using AI there isn’t a consent form or confidentiality agreement, just the question “What can I help you with?”
Using AI can be dangerous for patients seeking help with emergency situations because a “computer has no ethical legal duty,” Dr. Zervas added. This means that if someone who may have thoughts of suicide is using AI, there is nothing a computer can do to protect and keep the user safe. If that person were to come to a therapist, on the other hand, a therapist has to break confidentiality and give the patient further help because of their ethical legal duty. So Dr. Zervas recommends that people who may be using AI as therapy should “really think deeply about what it’s doing, what it’s not doing and have a plan for human involvement if needed.”
“To answer the question… is it reliable, we don’t know yet. It might be but that research hasn’t occurred yet, it should, because if it is a viable option, certainly, I completely agree more people need access to treatment,” Dr. Zervas stated.
So while it could be helpful to use your Snapchat MyAI or ask a question to Chat GPT about any mental illness questions or advice, users should be mindful of the answers that they are given and to be wary of their situation so that they can find a licensed therapist if needed and move away from relying on technology. Dr. Zervas acknowledges that AI can actually be really helpful and give helpful tools, but for therapy she would still recommend trying to find treatment if needed and to reach out to others or find books for those who don’t have access to treatment. And if using AI for therapy is the only option, visit Woebot Health for an AI therapy platform designed by psychologists and licensed therapists.
If you or someone you know needs help, reach out to Dr. Zervas at natialiezervas@danahall.og or healthcenter@danahall.org. Avoid relying solely on AI for therapy.
To reach the Suicide & Crisis Lifeline, visit 988lifeline.org, or text or call 988.
Image source: NPR