This article is part of our Young Leaders in Behavioral Health series, where we’re highlighting the six impressive young people chosen as our Youth Advisors for the 2024 Behavioral Health Tech Conference. In this article, we spoke with Alex Muir, MBA candidate at Kellogg School of Management, and Zoe Tait, Strategy and Business Operations Specialist at Jimini Health/Chief Research Assistant at the Stanford Computational Psychology & Well-Being Lab, about how artificial intelligence could help treat mental health disorders.

When you think of artificial intelligence, what image first comes into your mind? Many people wince at the idea of driverless passenger vehicles or roll their eyes at the thought of college students turning in papers written by ChatGPT. We don’t necessarily think of artificial intelligence as something that could have positive applications in a mental health care setting. But what if it did?

When it comes to artificial intelligence and mental health disorders, Zoe Tait is well-versed: Her undergraduate honors thesis, “The Naturalistic Uses of Large Language Models (LLMs) for Mental Health Purposes,” studied how individuals use LLMs for mental health support. 

“The ongoing mental health crisis and loneliness epidemic has decreased many individuals’ quality of life and ability to find quality mental health care,” she says. “The recent technological innovation and the recent rise in AI applications create the perfect opportunity for increased innovation in the digital mental health space. It is critical that we continue to ramp up the investigation of how to best integrate AI into settings where mental health care is a key focus.”

Zoe and Alex shared their insight on whether AI can help treat mental health disorders, and how to appropriately deploy AI solutions.

Can Artificial Intelligence Diagnose or Treat Mental Health Disorders? 

Alex Muir, MBA candidate at Kellogg School of Management, believes that artificial intelligence can do both.

“Fundamentally, AI is fantastic at organizing and learning from large swaths of data to produce an output,” she says. Our job, she explains, is to provide AI models with high-fidelity data of prior diagnoses or what treatments work. From there, “the learnings could be applied to more effectively diagnose and treat mental health disorders.” 

Zoe Tait, Strategy and Business Operations Specialist at Jimini Health/Chief Research Assistant at the Stanford Computational Psychology & Well-Being Lab, agrees. Artificial intelligence has potential for both diagnosis and treatment of mental health disorders. But that doesn’t mean the human element can be taken out of the picture.

“I believe that for AI to be successfully integrated into the current care landscape, there must be a collaboration between mental health practitioners and those developing AI-based tools.”

AI and Diagnostics 

Mental health disorders are typically diagnosed through a psychological evaluation, lab tests, and a physical exam to rule out physical problems. The Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, Text Revision (DSM-5-TR) helps clinicians and providers correctly diagnose a specific mental illness. How could artificial intelligence step in to make these diagnostic processes more accurate and efficient? 

“Recent work has shown that individuals with higher levels of depression will show significantly greater use of first-person singular pronouns (e.g., ‘I,’ ‘me,’ ‘my’) (Nook et al., 2022),” says Zoe. “Understanding how AI and LLMs can be integrated into the assessment or diagnosis of mental health disorders by identifying language patterns is critical.” 

Alex also cites an additional study published in 2017 that used deep learning to search for biomarkers of psychiatric and neurologic disease. “This study used a deep learning model to accurately classify patients diagnosed with schizophrenia based on their brain scans. Traditionally, the mental health field has struggled to have biological markers of mental illness/disorders. If we can truly find a way to develop models to classify appropriately mental disorders, this could be a huge leap forward in our ability to catch and treat mental illness.” 

AI and Treatment

Some experts think that artificial intelligence should step in before symptoms get severe, offering cognitive behavioral therapy (CBT) to people who are dealing with mild anxiety, depression, or burnout. Zoe agrees that while AI may not be able to completely replace therapy, it has the potential to increase “support, access, and overall accuracy in information-gathering, leading to a potentially significant improvement in patient outcomes.” 

And the benefits of AI aren’t just about the patients — they apply to clinicians, too. LLMs can augment clinical training, assist clinicians with their tasks, and lift the cognitive burden that often comes hand-in-hand with delivering care.

Next Steps to Implementing AI in Mental Health Care Settings

“Just like how humans are not perfect, AI is also not perfect,” Alex warns. “We should be wary of becoming overly dependent on a model, especially when we are not confident the underlying data is of a sufficient quality.”

Alex and Zoe discuss some of the considerations and risks around using artificial intelligence in the mental health field, and the next steps and advice they would propose.

Collaborate With Stakeholders

“When implementing AI in the clinical setting, it’s critical to collaborate with stakeholders across multiple areas to ensure patients receive the best care possible,” Zoe says. 

This includes people who play a key role in clinical or technological oversight, such as:

• Anyone involved in developing the technology
• Clinicians and physicians involved in patient treatment
• Health plans
• Policymakers
• Legal 

Maintain the Human Connection

A crucial part of mental health care hinges on human connection. There’s even a term for this — the “therapeutic alliance,” which refers to the collaborative alignment and bond between a clinician and patient.

“This alliance allows the patient to feel supported throughout their care journey,” Zoe explains. “Understanding if it is possible for an individual to form a similar therapeutic relationship with an AI agent will provide insight into how to best integrate patient-facing AI into care.”

AI can remember more than a human can. But humans carry certain qualities of empathy and validation that can’t quite be replicated in a bot. Researchers are still learning how to build AI that retains its memory capacity, yet also knows how to best support a human. Patient trust is too precious to lose — and the human element that fosters that trust may be irreplaceable. 

Utilize Frameworks

To successfully and safely integrate AI into a confidential care setting, use a framework for the deployment process. For example, the READI framework proposes a framework for readiness by outlining six important criteria:

• Safety
• Privacy/confidentiality
• Equity
• Engagement
• Effectiveness
• Implementation

“Realistically, we need to start collecting, organizing, and cleaning the appropriate data so that we can deploy the proper AI models,” Alex says, adding, “Additionally, all AI deployments should be used in tandem with a licensed provider.”

Where Else Can AI Help?

The areas where AI can help with mental health aren’t limited to diagnostics and treatment. AI-based solutions could also help people by:

• Providing resources and education around mental health disorders and treatment. “AI can help people navigate our complicated health care system, match them to the appropriate providers, prompt them with the appropriate next step in their treatment, or provide them with education around their particular disorder or symptoms they may be feeling,” Alex says. A simple example: When asked, ChatGPT will direct you to resources or helplines that you can use in time of need for your mental health. 
• Increasing access to care. With the current provider shortage, many people are waiting months for an appointment with a psychiatrist or therapist. “AI offers a scalable solution to ensure more people can get help faster,” Zoe says. She also points out that AI-based tools such as LLMs and chatbots are available 24/7 for individuals who need someone to talk to right away and can’t get ahold of a friend. 
• Reducing stigma around seeking care. “Oftentimes, people are afraid to seek care due to perceived judgment and fear of how others will view them,” Zoe points out. “Having something widely available that can provide clinically guided advice and resources could be extremely helpful in this area.” 

AI in Mental Health Care: A Timely Solution

Our society is at the cusp of a major opportunity. Technology and innovation are at the forefront of the news cycle. Gen Z uses tools like ChatGPT sometimes every day. Finding ways to integrate this new familiarity with AI into mental health care settings is a smart way to ensure more people receive the care they need. And young leaders like Alex Muir and Zoe Tait are poised to help make this happen. 

“AI has the potential to assist in ensuring increased access to quality mental health care around the world, therefore making mental health care more equitable across communities,” Zoe says. She adds: “I believe that we are further along than we think in terms of making valuable tools and technologies available to clinicians and patients to improve the current care landscape.”