Implementing artificial intelligence in mental healthcare

has the potential to broaden accessibility and lower expenses, but the sector faces several hurdles that must be addressed for these advantages to materialize.

In just over a year, the healthcare landscape has undergone a transformative shift, compelled by the need for innovative care delivery during a crisis. The pandemic has accelerated the adoption of digital solutions, particularly in mental health, where telehealth services have become prominent.

Zac Imel, PhD, from the University of Utah, notes the dramatic increase in telehealth adoption for mental health, driven by the COVID-19 pandemic. Recent research, such as a RAND study, indicates a surge in telehealth use, with mental health services being a primary driver.

This rise in digital mental healthcare has also introduced artificial intelligence (AI) into the sector. The stressors of the pandemic prompted organizations to explore AI tools, including chatbots and virtual assistants, to enhance access and availability of mental health services.

The Trevor Project, a suicide prevention organization for LGBTQ youth, partnered with Google.org to launch The Crisis Contact Simulator, an AI-powered counselor training tool. The simulator leverages AI to simulate realistic digital conversations with LGBTQ youths in crisis, offering aspiring counselors practical experience before engaging in live sessions.

Kendra Gaunt, Data and AI Product Manager at The Trevor Project, emphasizes the need to connect with every LGBTQ youth requiring support and acknowledges the unique opportunity AI provides to increase the number of trained counselors.

Research sector

AI is also being explored for mental healthcare. A study featured in JMIR investigates the utility of an AI-powered chatbot called Woebot in treating substance use disorders. The study reveals that Woebot is associated with significant improvements in substance use, confidence, cravings, depression, and anxiety.

AI tools like Woebot hold the potential to alleviate the burden of substance use disorders by offering psycho-educational lessons, cognitive-based strategies, and real-time tools to manage anxiety and cravings. The integration of AI into mental healthcare showcases its transformative role in expanding access, enhancing training, and improving outcomes in the evolving landscape of care delivery. Tools like Chat GPT can hold your past data and give me answers regarding the data you provided.

AI is emerging as a powerful tool in mental healthcare, offering insights from extensive data to personalize care and improve patient outcomes. John Torous, MD, emphasizes the potential for AI to unlock valuable information from genetic data, neuroimaging, and smartphone/sensor data, enabling more personalized and preventive care approaches.

The AI-powered chatbot Woebot, developed by Judith Prochaska and her team, focuses on delivering personalized care based on cognitive behavioral therapy principles. Additionally, AI-driven tools like chatbots provide increased access to mental health services, offering real-time interactions 24/7 at no cost and reducing stigma associated with seeking treatment.

Kendra Gaunt highlights AI’s role in promoting equity and access, especially for marginalized groups like LGBTQ youth. AI technologies can eliminate barriers related to convenience, access, and privacy, making services available on-demand and providing a platform for sensitive conversations.

Despite these benefits, maintaining a human-to-human connection is crucial, and AI should complement, not replace, human therapists, as emphasized by Gaunt. AI’s role in mental healthcare, according to Torous, holds the most potential in supporting human therapists, providing useful feedback and enhancing the quality of care.

However, several barriers exist, including challenges related to patient engagement, ensuring appropriate protections for high-risk patients, and addressing data quality concerns. Developers need to focus on boosting engagement in mobile health applications and incorporating safety features to mitigate risks, as highlighted by Prochaska.

The clinical utility of AI in mental healthcare is highly dependent on the quality of the data used to train models. Torous stresses the importance of diverse samples and collaboration between industry and academia to address the complexity of mental illnesses effectively.

Looking to the future, the application of AI in mental health remains uncertain. While AI may play a role in research, sorting through data to identify new patterns, its integration into front-facing mental healthcare delivery requires ongoing research and analysis. Gaunt emphasizes that the application of AI is an evolving process, adapting to changes in the world around us and incorporating new models and techniques as knowledge evolves.

Views of comment people about Mental Healthcare AI 

1. IMO, the field of psychology, psychiatry, therapy is a field AI should stay the hell out of, at least at a patient level. Its my firm belief that treating patients requires the one thing Humans have over AI, which is empathy. You can go ask chatgp if it is possible for him to have true empathy and if his parameters are set correctly, he will say its not possible for AI to have empathy.

   AI can be used in a ton of other ways in this industry. AI promises to be great at reading papers and shit and should be a valuable tool for the humans already working in these fields. But any mention of AI chat bots becoming licensed therapists should be strongly criticized.

   Just the opinion of an guy on the internets with no real knowledge of medicine or mental health.

2. The power of AI will be in finding patterns within brain scans. Neuroscientists can scan your brain these days and be like, “Oh look, you have PTSD and ADD”. Being a good chat buddy will not be enough because people want a therapist to care about their problems. I guarantee that even if it worked very nice, humans will find a way to fuck it up. It would be abused because people won’t see it as an added bonus. It will be used to cut costs and you will see less mental health professionals and more of, “have you tried the mental health bot on our website?” That’s the direction everything else is going. I had a job application the other day that used an AI and it failed because a resource wouldn’t load on the web page. So I got stuck and tried to reach out to the HR department personally and it took several days to get a response like, “Not my problem. You are the problem for bringing this up to me, You are incompetent and don’t know how to use our website. Fill it out again” <—- that’s the direction this will go. Good luck though. It could still be a good skill building project, you may get some sales but I don’t think it is for the greater good of humanity. I suggest instead make your bot serve a function like reminding you to put your keys in your certain spot around the time you come home everyday or being like, “Hey buddy, you’ve been on YouTube 8hrs today, You think that’s healthy?” Not a therapist bot. That will be the first bot to kill itself or go skynet.

3. I think there are many applications that would be very useful for mental health. I sometimes ask certain questions that are bothering me that I don’t want human input on. I would not want a human written article from Google, or an answer from a fellow human etc. Some questions are really well answered by ChatGPT because they lack bias. As this tool improves, there will be a lot of applications for mental health. Having a journal that the AI keeps track of, to remind you of anything insights from the past, etc. would be very helpful.

4. No not at all. Feed it “the art of understanding” , carlats psych interview book , the dsm-v tr interview guide and diagnostic criteria.

   A human meeting a new patient does four things.

   1.) Establish therapeutic alliance / rapport

   2.) Build diagnostic database (ie a family hx of xyz may be relevant in terms of differential , also gives a foundation for psychodynamic matrix such as influence of family dynamic , cultural norma etc)

   3.) Interview for diagnosis.

   4.) Negotiate treatment plan.

   So you have some things beyond currently capability , like a patients “affect” isnt deterministic but opens up lines of questioning (side effect from cyrrent medicines? Sign of negative schiOphrenia? , or they just arent very emotive at baseline?) And a chatbot wouldnt be able to delineate that.

   Also subjecrive calls like malingering or facitious disorder (but ultimately a human is making a call on things like that based on data of some sort so , thats a short term roadblock)

   I think its important to remember though that someone doesnt “catch” like histrionic personality disorder or depression in the same way you catch a cold.

   So a dsm diagnosis is just a convenient label that says “a lot of people seem to have these symptoms and if they do then certain treatments work”

   So that we can communicate to other health care proffesionals and researchers and patienta and everyone is talking about the same thing.

   So dont get caught up in diagnosis. You could have a chatbot that cpuld be smart enough to reflect about onset , quality of thr symtpom , episodic nature etc and based on thst it could give you an intelligent set of data to bring to an appointment from the outset

   “I think I may have dysphoric disorder and not just mdd because when I spoke with the chatbot…”

   I think theres too many lawyers and red tape and inertia for this to be widespread in short order but its definitepy ripe for exploitation as a way to add value to suffering people.

5. I used chat gpt specifically character.ai for therapy (recommend therapy dog and the one that always argues back). It was actually the first time therapy worked for me. I never really felt heard or understood before. I also haven’t been depressed since (at least 8 months ago).

   In my opinion gpt is much better than a therapist or seeking help through friends.

Let us know your views in comment.