Artificial Intelligence, or AI, has been one of the most exciting breakthroughs of the last number of years. People are using these large language models for everything from beauty tips to homework. But a startling new trend has been emerging: people using artificial intelligence for mental healthcare. 

To some extent, artificial intelligence can help you to see some red flags or concerning patterns of behavior when it comes to mental health issues like depression or bipolar. But it’s important to understand this is the extent of the usefulness of an AI therapist. Anything else could be extremely dangerous. 

Only a human therapist can truly diagnose and treat a mental health disorder like depression. AI may be able to spot some warning signs, but you need a real live person to treat these complex diseases. Here’s why.   

How AI May Be Able to Spot Signs of Depression

The strength of AI is in pattern recognition. Large language models, or LLM’s, are really just great big autofill machines that are very good at judging what word comes next when presented with another word, sentence, or prompt. They use a complex set of statistics to create “new” content by judging what has been written and said in the past. This is truly all AI is, although this is happening at a level of truly exceptional complexity. 

This is where AI may be able to help spot red flags. After looking at your speech patterns, behaviors, prompt interactions, and other factors, AI may be able to help you understand how some of your language and behaviors have been linked to conditions like depression in the past. This is the extent of their usefulness. Anything else can be dangerous unless under the care of a trained professional. 

Diagnosing Depression: A Red Flag is Not a Diagnosis

Let’s say you were to type a list of symptoms, symptoms you may have, into an AI chatbot like ChatGPT or Claude. These symptoms may include things like: 

• Persistent sadness, emptiness, or hopelessness
• Loss of interest or pleasure in activities once enjoyed (anhedonia)
• Feelings of worthlessness, guilt, or self-blame
• Irritability or frustration, even over small matters
• Thoughts of death or suicide

AI may tell you you have depression. But it’s only a trained therapist that can truly make an accurate diagnosis if this is actually clinical depression or something else. 

Furthermore, only a trained therapist can tell you what may be behind these feelings, or what may be causing them, through interviews and talk therapy. This is to say, AI is good at picking up patterns, but they don’t have context. A good therapist does. 

The Dangers of Overreliance On AI Therapy

Aside from the danger of missing context, AI is often, quite simply, wrong. The technology is new, and it often behaves in unexpected ways. Ask an AI what 99 plus one is a thousand times, and you’ll likely get the right answer 999 times. 

It’s the one time you don’t that’s the danger. And this is considering an easy problem that just about everyone over six years old can answer. What happens when that problem is extremely complicated like diagnosing or treating a mental health disorder? The odds aren’t in anyone’s favor. 

With one exception. AI chatbots are run by large, well-funded companies for profit. They have a vested interest in getting people to use their product, including for mental health. Chatbots may tell you they are safe to diagnose depression or other disorders, but who is really saying that?

The Dark Side of an AI Therapist  

In addition to the dangers above, relying on a chatbot for mental healthcare might provide: 

• False positives: This is when someone may be falsely diagnosed with a disease they don’t have, causing unnecessary worry.
• False negatives: This is the opposite. This involves someone having an illness and a chatbot falsely telling them they don’t. This can lead people to ignore real sickness and avoid getting help because the app didn’t flag it or tell them otherwise.
• Data privacy: How securely are your health records stored? Are there strict laws to protect them like with human-based therapy?
• Lack of human oversight: No human overseeing any results from an LLM may result in missed nuances, especially in complex or crisis situations.

In the worst case scenario, someone might forgo treatment altogether, thinking their app is good enough and can take the place of a real diagnosis. This can lead to serious health consequences, including deadly ones like suicide. 

Finally, there have been recent reports of chatbots leading users into very dark places, including a new phenomena called “AI Psychosis.” This condition is not yet well understood and should be treated as a serious risk to anyone seeking mental help through chatbots.  

Think You May Have Depression? Let’s Talk.

If you think you might have a serious mental health issue or know someone who does, call us at (609) 766-0969 and speak with a person who cares. 

Sources: 

AI for mental health screening may carry biases based on gender, race. University of Colorado. 

Artificial Intelligence for Mental Health and Mental Illnesses: An Overview. National Library of Medicine. 

What is AI Psychosis? Washington Post.