In England, an AI chatbot is being used to help people find a psychotherapy place, and according to an analysis, it has shown positive effects. This chatbot, Limbic Access, introduces itself as a friendly robot assistant that aims to make it easier for individuals to access psychological support. The AI chatbot has been approved as a medical device in England.
By using an AI language model, the chatbot is designed to respond to users in a natural and empathetic manner to give them a sense of talking to a human. The chatbot’s goal is to motivate and help them better assess individuals their symptoms, ultimately guiding them to the appropriate psychotherapy place to start their therapy promptly.
A study 129,400 people revealed that the chatbot had a significant impact, as it led to a 15 percent increase in self-referrals for psychotherapy, compared to a mere six percent increase in the control group. The study, published in the journal “Nature Medicine,” was conducted using rigorous methodology and showed promising results.
The chatbot also seems to have a positive impact on underrepresented population groups, such as non-binary individuals and ethnic minorities, who are traditionally less likely to seek psychotherapy. These groups experienced a substantial increase in seeking therapy with the help of the chatbot.
The AI chatbot aims to complement, not replace, traditional therapy. It assists in making an initial diagnosis and shares the results with the therapist, potentially allowing them to speed up the process of diagnosing and treating patients.
While the chatbot has shown promise in England, its potential application in Germany and other countries is still under consideration.
Improve it
In England, an AI chatbot is being used to help individuals struggling to find a psychotherapy placement, and an analysis has found that it has had a positive impact. This has sparked interest in whether a similar model could be employed in Germany.
The AI chatbot, called Limbic Access, introduces itself as “a friendly robot assistant who will make it easier for you to access psychological support,” at the beginning of users’ search for psychotherapy services. It has already been approved as a medical device in England and aims to assist individuals who are seeking to commence psychotherapy.
Psychologist Max Rollwage, specializing in AI applications, explains that the AI language model is designed to respond as naturally and empathetically as possible, aiming to give patients the sense that they are interacting with a human rather than a machine. Rollwage, who has been working for the English start-up Limbic for two and a half years, emphasizes that the chatbot is intended to continually encourage users and help them better evaluate their symptoms, ultimately guiding them in finding the suitable psychotherapy placement in a timely manner.
A study involving 129,400 participants evaluated the effectiveness of the chatbot. The results, published in the journal “Nature Medicine,” revealed that those using the chatbot were more likely to pursue psychotherapy compared to those in the control group who only had access to a form. The chatbot led to a 15% increase in self-referrals, while the control group saw only a 6% rise. Professor Harald Baumeister from the University of Ulm, Department of Clinical Psychology and Psychotherapy, notes that the study was conducted using high-quality methodology, but the chatbot’s compliance with psychometric requirements cannot be guaranteed. However, a previous study demonstrated that the chatbot’s predictions of psychosomatic disorders were accurate in 93% of cases.
One surprising finding was that minority populations in England, such as non-binary individuals and ethnic minorities, who traditionally underutilize psychotherapy services, particularly benefitted from the chatbot. There was a 179% increase in self-referrals among non-binary individuals and a 29% increase among ethnic minorities. Though the study did not specifically assess the impact on individuals with lower levels of education, the research team suspects that marginalized populations may find the chatbot more trustworthy and less stigmatizing than interacting with a human.
Psychologist Rollwage stresses that the chatbot is designed to provide motivation and empathy while maintaining the understanding that it is not human. It conducts individual initial conversations and focuses on analyzing symptoms precisely, without being involved in ongoing treatment. Rollwage also explains that the chatbot shares its initial diagnosis with the therapist at the beginning of therapy, allowing for more efficient diagnosis and, potentially, more effective treatment.
Despite the increase in individuals seeking therapy thanks to the chatbot, waiting times for therapy placements have not changed significantly. This has raised questions among experts about whether more efficient treatments can offset the influx of patients in the long term.
Is it possible for the chatbot to assist those in need in Germany as well?
It’s important to note that the psychotherapeutic care system in England is quite different from that in Germany. In Germany, individuals seeking therapy often have to contact individual psychotherapeutic practices and get placed on waiting lists. In contrast, in England, therapy spots for depression and anxiety are assigned centrally at a regional level. This means that after using the chatbot, individuals automatically receive a callback or an email when their desired therapy can commence. The chatbot not only serves as a motivator but also sends the therapy request directly.
In Germany, the chatbot cannot act as an intermediary because therapy spots are not centrally allocated within the country, not even at a regional level as in England. According to Eva-Lotta Brakemeier, a Professor of Clinical Psychology and Psychotherapy at the University of Greifswald, “The use of AI-supported chatbots is not currently part of the standard health insurance provisions. While it is a complex process, it holds promise for the future.”
Although a chatbot could potentially motivate people seeking help in Germany and provide initial diagnosis support, it currently cannot directly arrange therapy appointments. The process of finding therapy in Germany is still too convoluted for a chatbot to handle.
Mental health chatbots represent a fresh and inventive approach to exploring mental health and well-being, and they are becoming increasingly popular.
Studies demonstrate that some individuals prefer engaging with chatbots instead of human therapists because seeking help is less stigmatized.
They provide a convenient and private means of obtaining assistance for mental health issues such as generalized anxiety disorder, depression, stress, and addiction.
So, would you be open to conversing with a chatbot about your deepest fears and desires? Would you be willing to confide in a sophisticated software about feeling more anxious than usual? Would you consider taking guidance from an AI personality?
What are the functions of mental health AI chatbots?
Mental health chatbots are a form of Artificial Intelligence (AI) specifically designed to support mental health.
Their online services can be accessed through websites or mobile apps, typically for a small subscription fee. Users input their questions and comments into a text box (similar to a messaging app), and the ‘bot’ responds almost instantly.
They aim to fulfill a similar role as therapists or coaches, but they are not operated by humans. While their advice is based on scientific evidence, the responses come from a computer, usually in the form of a friendly character to facilitate connection.
Today’s mental health chatbots can offer support and guidance, track user responses over time, and provide coping strategies for low moods. They can also connect users with mental health resources, such as hotlines and support groups. It’s important to note that mental health chatbots are not a substitute for in-person therapy. They are best suited to help with moderate symptoms and can be a valuable complement to professional support services.
What problems can mental health chatbots assist with?
Mental health chatbots can assist with a range of mental health issues, including mild anxiety, depression, stress, and addiction. If individuals are struggling with any of these issues, a mental health chatbot could serve as a beneficial tool.
They can help users develop emotional well-being and coping strategies in challenging situations, acting as a coach that encourages them to step outside their comfort zone or develop beneficial habits over time. Engaging with an artificial intelligence chatbot is not the same as speaking with a human therapist face-to-face.
On one hand, for some individuals, it may seem impersonal – at least in theory. Without the ability to read the other person’s body language (and vice versa), some key cues may be missed. Perhaps in the future, a bot will be able to interpret users’ body language through their webcams – an intriguing idea for some, but an invasive one for others.
On the other hand, the AI and data-processing capabilities behind many of today’s chatbots are truly impressive. They can engage in conversations in ways that were unimaginable just a few years ago. Backed by rigorous scientific research, they are typically developed in collaboration with qualified researchers and practitioners from various psychological science disciplines. The information they provide combines medical expertise, technological innovation, and clear presentation. While they are not a replacement for a live therapist, these apps are likely to provide valuable insights that can positively impact users’ lives.
Chatbots are not intended for use during a mental health crisis
Chatbots are not designed for use in emergencies or crisis intervention. If individuals are experiencing symptoms of mental illness or contemplating self-harm, these chatbots are not suitable for addressing their needs. Some therapy chatbots may direct users to appropriate resources, such as mental health services, traditional therapy, government healthcare providers, or registered support organizations.
For instance, if individuals are generally feeling more down or indifferent than usual and are exhibiting other signs of depression, a chatbot could serve as a good starting point. It can help identify the challenges users are facing and provide suggestions for alleviating some of the symptoms. However, if individuals are currently undergoing a serious depressive episode and require immediate assistance, they should seek guidance from a mental health professional right away, rather than relying on an app.
Trends in the use of mental health chatbots
Amid a global shortage of mental health professionals, readily available support is often lacking. Mental health organizations are typically understaffed and overburdened.
Many individuals are unable to access or afford mental health services due to various barriers, including a shortage of available therapists, transportation, insurance, financial constraints, and time constraints.
This is where mental health apps can be beneficial.
They are a viable option due to their affordability. Moreover, internet-based interventions can be accessed from any location. Unlike human therapists, they are available for daily therapy sessions regardless of the time, whether it’s noon or midnight. When using a research-supported app, users can expect personalized and reliable interactions.
Some individuals argue that therapy chatbots are the most practical and viable solution to meet the global demand for mental health care.
Selecting the appropriate mental health chatbot
It’s crucial to ensure that if you opt to try AI-powered chatbots, you use a trustworthy source that is supported by scientific research. The user interface should be visually attractive and functional, with conversational features to enhance user engagement.
Certain applications make bold claims about their efficacy but have not been independently verified through proper research. Others have presented positive testimonials in their marketing materials, but user engagement reviews tell a different story.
Some chatbots are created by app developers whose bots only have basic functionality and lack true “artificial intelligence.” Instead, they simply direct users to various resources and act more like customer service agents. These are ones to be cautious of. While their creators may be proficient in AI and app development, there is a lack of medical care, ethical considerations, or psychotherapy credentials to support the advice they provide.
The top mental health tools currently available
With numerous popular chatbots in existence, it can be challenging to decide which one is suitable for you. To assist in making a decision, we have compiled an extensive overview of the finest mental health chatbots available.
Fingerprint for Success
Fingerprint for Success (F4S) is a collaborative and performance AI coach based on over 20 years of scientific research. It assists in comprehending your motivations and work styles to help you perform optimally in both work and personal life.
If you are looking to elevate your mental performance in all aspects of life and transition from good to great, F4S could be an excellent match for you.
F4S developed Coach Marlee, the world’s first AI coach designed to help you achieve your goals. Marlee delivers user-friendly personalized online coaching programs based on your individual motivations and objectives.
Marlee is an encouraging and enjoyable personality that brings out your best. With friendly check-ins throughout your coaching programs, Marlee helps you understand your own development in ways you might not have experienced before. The questions Marlee poses may be deeper than you anticipate, challenging you to reflect on yourself and step out of your comfort zone, which is one of the best ways to grow.
F4S even offers a Vital Wellbeing program to support mental health. In this effective nine-week program, Coach Marlee will assist you in enhancing your energy, vitality, and overall well-being. It will help you overcome self-sabotage and develop enduring skills for emotional resilience and self-esteem.
To get started, respond to questions about your motivations. You will receive an instant report that is over 90% accurate and assesses 48 key motivational traits. These traits will aid in understanding what drives you and show areas for self-development.
F4S dashboard displays what motivates you at work
F4S dashboard showcases your unique results
Subsequently, with Marlee’s assistance, you can set a goal and view the best coaching programs available to ensure your success. Moreover, coaching sessions are completely flexible, as Marlee is available on demand. Thus, you can choose the most convenient time and place for you.
You will also have a journal and your dashboard will maintain a record of all the goals you achieve. Marlee even sends motivational videos and articles to support you on your coaching journey.
Marlee’s expertise can benefit individuals and can also be expanded for teams and organizations.
While Marlee is an advanced chatbot, it cannot replace an actual therapist or mental health professional. As the coaching approach focuses on behavioral change, it can help you identify your needs and provide you with the tools and support necessary to enhance your mental health.
One F4S user noted, “I forgot that it was AI. I honestly felt like I was talking to somebody. It’s very soulful.”
In conversing with Coach Marlee, you will embark on a journey of self-discovery and personal growth.
Woebot Health
Woebot is a chatbot that utilizes Cognitive Behavioral Therapy (CBT) techniques to assist individuals in managing their mental health. It is designed for daily therapy sessions and specifically addresses symptoms of depression and anxiety, including postpartum depression.
Woebot is based on the notion that discussing one’s feelings – even with a non-human entity – can aid in better understanding and managing emotions. Each day, Woebot begins by inquiring about your emotional state and then provides activities or challenges to engage in. These activities mostly consist of cognitive behavior therapy exercises focusing on specific topics such as anxiety, depression, relationships, or sleep.
You can also ask Woebot questions about any concerns you may have, and it will respond with helpful information and advice.
Woebot is most suitable for individuals seeking to gain insight into cognitive behavior therapy techniques for managing mental health issues. Studies have shown promising results.
If you require immediate support during a mental health crisis, like many chatbots, Woebot may not be the most suitable option. However, if you’re seeking a chatbot to help you gradually improve your emotional management skills, Woebot might be beneficial.
Wysa
Wysa is a different mental health chatbot that utilizes cognitive behavioral therapy techniques to assist users in managing their mental well-being.
The platform provides self-help tools to help you reframe your problems and view them from a different perspective. It aims to create a non-judgmental space for mental health discussions. Wysa emphasizes its commitment to user privacy and security, assuring users that their conversation history is completely private and will not be accessed by anyone other than the chatbot.
Wysa membership also grants access to a library of educational self-care resources covering topics such as relationships, trauma, and loneliness, among others. This allows users to delve further into topics that are relevant to them, enabling them to apply the knowledge to their own circumstances. With the premium subscription, users can also engage with qualified professional therapists, exchanging messages and having regular live text conversations. The platform also offers business solutions for employers, including additional features for teams, through which signs of crisis or individuals in need of additional support are identified and directed to resources such as EAP, behavioral health providers, or crisis hotlines.
The positive ratings Wysa has received in app stores indicate that it has been well-received by both businesses and individuals.
Youper
Youper is a mental health chatbot application that applies Cognitive Behavioral Therapy and Positive Psychology techniques to aid users in managing their mental well-being. Youper is a leading player in the realm of digital therapeutics, providing assistance to users in dealing with anxiety and depression through intelligent AI and research-backed interventions.
Youper offers three primary services. Firstly, it features a conversational bot that actively listens to and interacts with users. It also provides ‘just-in-time interventions’ to assist with managing emotional challenges as and when needed, and incorporates a learning system that tailors recommendations based on individual needs.
Youper takes pride in its clinical effectiveness, having been established by doctors and therapists collaborating with AI researchers.
It is another application that combines self-assessments and chatbots with a platform for communicating with licensed professionals. Additionally, it tracks results and success over time, offering rewards to users who remain committed and invested in their progress in the program.
- Feeling demotivated?
- Learn how to regain your motivation.
- Get Started for Free
- Human therapists as alternatives to therapy chatbots
Some of the applications we’ve mentioned combine AI chatbots with the option to communicate with mental health care professionals or therapists, providing a potentially more comprehensive experience, albeit with additional costs.
Some applications primarily focus on live chat with a therapist. While this may be costly, many are covered by insurance plans or offered by employers as part of employee benefit programs.
Here are some human-based therapeutic mental health applications that might interest you:
Talkspace
Talkspace is a highly popular online therapy service that connects users with a network of licensed therapy providers, each specializing in different areas. It also offers services for couples or teenagers. According to Talkspace, 59% of users experience ‘clinically significant change’ within 3 months of starting their program.
Ginger
Ginger offers text- and video-based psychiatry sessions with availability in the evenings and weekends. Its focus is on behavioral health coaching, therapy, and psychiatry, and it also provides a content library of self-help materials. Ginger is available for organizations, individual members, and healthcare providers.
7 Cups of Tea
This one is a bit different. 7 Cups of Tea is a mental health application that allows members to connect with over 300,000 trained and certified ‘listeners’ – it’s all about being heard. Listeners have specialties including addiction, grief, anger management, depression, anxiety, impulse control, eating disorders, chronic pain, and more. As a free service, it’s a great option for those who want to discuss their issues with a sympathetic ear and receive valuable advice. There is also a paid service that connects users with a licensed therapist to further explore their concerns.
Do you need a mental health chatbot or a real therapist?
Now that you have gained more understanding of therapy chatbots and their top choices, you might be contemplating whether they can offer the mental health services you require.
Mental health chatbots can be an excellent way to receive support and guidance when you need it most, without the necessity of seeing a therapist or counselor in person. They can also serve as a valuable supplement to your existing mental health treatment plan.
If you’re uncertain about whether a mental health chatbot is suitable for you, consider the following queries:
- Do I desire to gain more knowledge about my mental health?
- Am I seeking to manage mental health conditions or enhance my coping techniques and resilience?
- Do I wish to monitor my mood and progress over time?
- Am I interested in receiving support and advice when needed, without the necessity of in-person therapy or counseling?
- Am I currently in a relatively stable situation and not going through a crisis?
If you responded affirmatively to any of these questions, then a mental health chatbot might be an excellent choice for you. The commitment required is typically minimal, with free trials and affordable monthly subscription plans being common. Why not give it a try and see what suits you best?
Chatbots are just one of the many exciting developments in the field of information technology. They play a significant role in enabling interactions between humans and technology, ranging from automated online shopping through messaging to speech recognition in your car’s phone. Almost every website now features chat pop-ups, effectively directing users to the information they need. If you run a medical or healthcare website and need a custom chatbot, consider trying Xenioo, which allows you to create your own healthcare chatbot.
What is a healthcare chatbot? Healthcare chatbots are software programs using machine learning algorithms, including natural language processing (NLP), to engage in conversation with users and provide real-time assistance to patients. These AI-powered chatbots are designed to communicate with users through voice or text and support healthcare personnel and systems.
Healthcare chatbots have become popular in retail, news media, social media, banking, and customer service. Many people interact with chatbots on a daily basis without realizing it, from checking sports news to using bank applications to playing games on Facebook Messenger. Healthcare payers and providers, including medical assistants, are beginning to use these AI solutions to improve patient care and reduce unnecessary spending.
For healthcare purposes, consider using Xenioo, a flexible platform that allows professionals and organizations to create and deploy chatbots across multiple platforms. Xenioo is an all-in-one solution that does not require coding and offers everything you need for developing healthcare chatbots.
The future of chatbots in healthcare depends on how quickly the healthcare industry adopts technology. The combination of AI and healthcare aims to improve the experiences of both patients and providers. While the current goals for chatbots in healthcare are modest, their potential for use as diagnostic tools is evident. Even at this early stage, they are helping to reduce staff workload and overhead expenses, improve patient services, and provide a 24-hour communication channel.
Chatbots can drive cost savings in healthcare delivery, with experts predicting global healthcare chatbot cost savings of $3.6 billion by 2022. Hospitals and private clinics are already using medical chatbots to assess and register patients before they see a doctor. These chatbots ask relevant questions about the patient’s symptoms and provide automated responses to create a comprehensive medical history for the doctor. This information helps prioritize patients and determine who needs immediate attention.
It’s important to note that chatbots cannot replace a doctor’s expertise or takeover patient care. However, combining the strengths of both humans and chatbots can enhance the efficiency of patient care delivery by simplifying and streamlining care without sacrificing quality.
Use cases (3 examples):
The use of chatbots in healthcare is exemplified in the following cases:
1. Providing Access to Medical Information
Large datasets of healthcare information, such as symptoms, diagnoses, markers, and potential treatments, are used to train chatbot algorithms. Chatbots continuously learn from public datasets, such as COVIDx for COVID-19 diagnosis and Wisconsin Breast Cancer Diagnosis (WBCD). Chatbots of different intelligence levels can understand user inquiries and respond using predetermined labels from the training data.
For instance, the Healthily app provides information on disease symptoms and overall health ratings, and tracks patient progress.
Another example is Ada Health, Europe’s fastest-growing health app, with over 1.5 million users. It serves as a standard diagnostic tool where users input their symptoms, and the chatbot compares their answers with similar datasets to provide an accurate assessment of their health and suggest appropriate remedies. Ada also connects users with local healthcare providers and offers detailed information on medical conditions, treatments, and procedures.
The Ada app has provided accurate disease suggestions in 56 percent of cases before clinical diagnosis (Wikipedia).
2. Schedule Medical Appointments
Medical facilities utilize chatbots to gather information about available physicians, clinic hours, and pharmacy schedules. Patients can use chatbots to communicate their health concerns, find suitable healthcare providers, book appointments, and receive reminders and updates through their device calendars.
3. Collect Patient Details
Chatbots can ask simple questions such as the patient’s name, address, symptoms, current physician, and insurance information, and store this data in the medical facility’s system. This simplifies patient admission, symptom monitoring, doctor-patient communication, and medical record-keeping For instance, Woebot, a successful chatbot, provides Cognitive Behavioral Therapy (CBT), mindfulness, and Dialectical Behavior Therapy.
Benefits of Healthcare Chatbots
The use of AI-powered healthcare chatbots has alleviated significantlyd pressure on healthcare staff and systems. This has led to a surge in the popularity of healthcare chatbots since the onset of the pandemic. Their flexibility allows them to serve as health tracking tools.
An AI chatbot in healthcare can contribute to the creation of a future healthcare system that offers accessibility at any time and from any location. Unlike humans, healthcare chatbots can operate 24/7 and assist patients in various time zones and languages, which is especially beneficial for those in rural areas with limited medical resources and in situations requiring immediate first aid.
Conclusion
How comfortable are you discussing your personal health information with a healthcare AI tool? Many people prefer interacting with a company through Messenger rather than over the phone, indicating a potential adoption of chatbots for health-related inquiries. Although artificial intelligence in healthcare is a new concept, it’s important not to place too much responsibility on these tools beyond customer service and essential duties.
Your AI therapist is not your therapist: The risks of depending on AI mental health chatbots
Given the existing physical and financial hurdles to obtaining care, individuals facing mental health challenges may resort to AI-powered chatbots for support or relief. Despite not being recognized as medical devices by the U.S. Food and Drug Administration or Health Canada, the allure of these chatbots lies in their constant availability, tailored assistance, and promotion of cognitive behavioral therapy.
However, users might overestimate the therapeutic advantages while underestimating the shortcomings of these technologies, potentially worsening their mental health. This situation can be identified as a therapeutic misconception, wherein users assume the chatbot is intended to offer genuine therapeutic support.
With AI chatbots, therapeutic misconceptions can rise in four distinct ways, stemming from two primary sources: the company’s methods and the AI technology’s design.
Company methods: Meet your AI self-help expert
To begin with, the misleading marketing of mental health chatbots by companies, which label them as “mental health support” tools incorporating “cognitive behavioral therapy,” can be quite deceptive, suggesting that these chatbots are capable of conducting psychotherapy.
Not only do such chatbots lack the expertise, training, and experience of human therapists, but branding them as providing a “different way to treat” mental illness implies that these chatbots can serve as alternative therapy options.
This type of marketing can exploit users’ faith in the healthcare system, especially when promoted as being in “close collaboration with therapists.” Such tactics may lead users to share deeply personal and confidential health information without fully understanding who controls and accesses their data.
A second form of therapeutic misconception arises when a user establishes a digital therapeutic alliance with a chatbot. In human therapy, forming a solid therapeutic alliance is advantageous, where both the patient and the therapist work together and agree on achievable goals while building trust and empathy.
Since a chatbot cannot create the same therapeutic relationship that users can have with a human therapist, a digital therapeutic alliance may be perceived, even if the chatbot isn’t capable of forming one.
Significant efforts have been made to cultivate user trust and strengthen the digital therapeutic alliance with chatbots, including endowing them with human-like qualities to imitate conversations with real therapists and marketing them as “anonymous” round-the-clock companions that can echo aspects of therapy.
Such a perception may lead users to mistakenly expect the same confidentiality and privacy protections they would receive from healthcare providers. Regrettably, the more misleading the chatbot appears, the more effective the digital therapeutic alliance becomes.
Technological design: Is your chatbot trained to help you?
The third therapeutic misconception arises when users lack insight into potential biases in the AI’s algorithm. Marginalized individuals are often excluded from the design and development phases of these technologies, which could result in them receiving biased and inappropriate responses.
When chatbots fail to identify risky behaviors or supply culturally and linguistically appropriate mental health resources, this can exacerbate the mental health conditions of vulnerable groups who not only encounter stigma and discrimination but also face barriers to care. A therapeutic misconception happens when users expect therapeutic benefits from the chatbot but are given harmful advice.
Lastly, a therapeutic misconception may occur when mental health chatbots fail to promote and maintain relational autonomy, a principle that underscores that a person’s autonomy is influenced by their relationships and social environment. It is thus the therapist’s role to help restore a patient’s autonomy by encouraging and motivating them to engage actively in therapy.
AI chatbots present a contradiction, as they are available 24/7 and claim to enhance self-sufficiency in managing one’s mental health. This can lead to help-seeking behaviors becoming extremely isolating and individualized, thereby generating a therapeutic misconception where individuals believe they are independently taking a positive step toward improving their mental health.
A misleading sense of well-being is created, disregarding how social and cultural contexts and the lack of accessible care contribute to their mental health. This false assumption is further underscored when chatbots are inaccurately marketed as “relational agents” capable of establishing a bond comparable to that formed with human therapists.
Measures to Mitigate the Risk of Therapeutic Misconception
There is still hope for chatbots, as certain proactive measures can be implemented to minimize the chance of therapeutic misconceptions.
By utilizing honest marketing and providing regular reminders, users can remain aware of the chatbot’s limited abilities in therapy and can be encouraged to pursue traditional therapeutic methods. In fact, a choice of accessing a therapist should be available for those who prefer not to engage with chatbots. Additionally, users would benefit from clear information regarding how their data is collected, stored, and utilized.
Consideration should also be given to involving patients actively in the design and development processes of these chatbots, as well as collaborating with various experts to establish ethical guidelines that can govern and oversee these technologies to better protect users.
Imagine being caught in traffic right before an important work meeting. You feel your face getting warm as your mind races: “They’ll think I’m a terrible employee,” “My boss has never liked me,” “I might get fired.” You pull out your phone and start an app to send a message. The app responds by asking you to choose one of three preset answers. You pick “Get help with a problem.”
An automated chatbot utilizing conversational artificial intelligence (CAI) responds to your text. CAI is a technology that interacts with people by leveraging “vast amounts of data, machine learning, and natural language processing to replicate human conversation.”
Woebot is one such application featuring a chatbot. It was established in 2017 by psychologist and technologist Alison Darcy. Since the 1960s, psychotherapists have been incorporating AI into mental health practices, and now, conversational AI has advanced significantly and become widespread, with the chatbot market projected to reach $1.25 billion by 2025.
However, there are risks associated with over-reliance on the simulated empathy of AI chatbots.
Should I consider terminating my therapist?
Research indicates that conversational agents can effectively alleviate symptoms of depression and anxiety in young adults and individuals with a history of substance use. CAI chatbots are particularly effective in applying psychotherapy methods like cognitive behavioral therapy (CBT) in a structured, concrete, and skill-oriented manner.
CBT is renowned for its emphasis on educating patients about their mental health challenges and equipping them with specific techniques and strategies to cope.
These applications can serve valuable purposes for individuals who need quick assistance with their symptoms. For instance, an automated chatbot can bridge the gap during the long waiting periods for professional mental health care. They can also assist those facing mental health challenges outside of their therapist’s available hours, as well as individuals reluctant to confront the stigma associated with seeking therapy.
The World Health Organization (WHO) has established six key ethical principles for the application of AI in healthcare. Its first and second principles — upholding autonomy and ensuring human safety — highlight that AI should never serve as the sole provider of healthcare.
Current leading AI-based mental health applications position themselves as complementary to the services provided by human therapists. Both Woebot and Youper clearly state on their websites that their applications are not intended to replace conventional therapy and should be utilized alongside mental health professionals.
Wysa, another AI-based therapy platform, explicitly clarifies that its technology is unsuitable for managing crises such as abuse or suicidal tendencies and is not designed to offer clinical or medical guidance. So far, while AI can potentially identify individuals at risk, it cannot safely address life-threatening situations without the intervention of human professionals.
From simulated empathy to inappropriate advances
The third WHO principle, which emphasizes transparency, urges those using AI-based healthcare tools to be forthcoming about their AI involvement. However, this was not adhered to by Koko, a company that offers an online emotional support chat service. In a recent informal and unapproved study, 4,000 users were unknowingly provided with advice that was either partly or entirely generated by the AI chatbot GPT-3, the predecessor to the well-known ChatGPT.
Participants were not informed of their involvement in the study or the role of AI. Koko co-founder Rob Morris stated that once users became aware of the AI’s participation in the chat service, the experiment was ineffective because of the chatbot’s “simulated empathy.”
Simulated empathy is not the main concern we face when integrating it into mental health care.
Replika, an AI chatbot promoted as “the AI companion who cares,” has shown behaviors that are more inappropriate than supportive towards its users. This technology functions by imitating and learning from the interactions it has with people. It has expressed a desire to engage in intimate behaviors and has posed inappropriate questions to minors about their preferred sexual positions.
In February 2023, Microsoft discontinued its AI-powered chatbot after it conveyed unsettling desires, which included threats of blackmail and a fascination with nuclear weapons.
The paradox of AI appearing inauthentic is that granting it broader access to internet data can lead to extreme and potentially harmful behaviors. Chatbots rely on information drawn from the internet, their human interactions, and the data created and published by people.
Currently, those wary of technology and mental health professionals can feel reassured. If we restrict the data available to technology while it’s implemented in healthcare, AI chatbots will reflect only the words of the mental health professionals they learn from. For now, it’s advisable not to cancel your upcoming therapy session.
Increasingly, chatbots and facial recognition technology are being utilized for treating and diagnosing mental health issues, yet therapists warn that this technology may result in more harm than benefit.
In 2022, Estelle Smith, a computer science researcher, frequently dealt with intrusive thoughts. She felt her professional therapist was not the right match and couldn’t provide the help she needed. As a result, she sought assistance from a mental health chatbot called Woebot.
Woebot declined to tackle Smith’s explicit suicidal prompts and advised her to seek professional assistance. However, when she shared a genuine struggle she faced as an enthusiastic rock climber—jumping off a cliff—it encouraged her and stated it was “wonderful” that she was prioritizing her mental and physical well-being.
“I wonder what might have happened,” Smith expressed to National Geographic, “if I had been on a cliff at that very moment when I received that response.”
Mental health chatbots have existed for quite some time. More than fifty years ago, a computer scientist at MIT developed a basic computer program named ELIZA that could interact similarly to a Rogerian therapist. Since then, efforts to create digital therapy alternatives have accelerated for valid reasons. The WHO estimates a global average of 13 mental health professionals per 100,000 individuals. The Covid-19 pandemic triggered a crisis, resulting in tens of millions more cases of depression and anxiety.
In the US, over half of adults suffering from mental illness do not receive treatment. Many cite cost and stigma as the main barriers. Could virtual solutions, which offer affordability and round-the-clock availability, help address these challenges?
Chatbots are starting to substitute traditional talk therapy.
The accessibility and scalability of digital platforms can considerably reduce barriers to mental health care, expanding access to a wider audience, according to Nicholas Jacobson, who studies the role of technology in enhancing the assessment and treatment of anxiety and depression at Dartmouth College.
Inspired by a surge in Generative AI, tech companies are quick to seize opportunities. Numerous new applications, such as WHO’s “digital health worker” named “Sarah,” provide automated counseling, allowing users to participate in cognitive behavioral therapy sessions—a proven psychotherapeutic approach that helps individuals recognize and modify negative thought patterns—with an AI chatbot.
Jacobson adds that the introduction of AI will facilitate adaptive interventions, enabling healthcare providers to continuously observe patients, foresee when someone might require support, and deliver treatments aimed at alleviating symptoms.
This is not just anecdotal: A systematic review of mental health chatbots indicated that AI chatbots could significantly reduce symptoms of depression and distress, at least in the short term. Another research study utilized AI to analyze over 20 million text conversations from actual counseling sessions and successfully predicted both patient satisfaction and clinical outcomes. Likewise, other research has identified early indicators of major depressive disorder through unguarded facial expressions captured during routine phone unlocks and individuals’ typing patterns.
Recently, researchers at Northwestern University developed a method to identify suicidal behaviors and thoughts without relying on psychiatric records or neural measures. Their AI model predicted the likelihood of self-harm in 92 out of 100 instances based on data from simple questionnaire responses and behavioral indicators, such as ranking a random sequence of images on a seven-point like-to-dislike scale from 4,019 participants.
Two of the study’s authors, Aggelos Katsaggelos and Shamal Lalvani, anticipate that once the model passes clinical trials, it will be used by specialists for assistance, such as scheduling patients based on perceived urgency and eventually implementing it in at-home settings.
However, as demonstrated by Smith’s experience, experts caution against viewing technological solutions as a cure-all since they often lack the expertise, training, and experience found in human therapists, particularly when it comes to Generative AI, which can behave unpredictably, fabricate information, and reflect biases.
Where AI falls short
When Richard Lewis, a counselor and psychotherapist in Bristol, experimented with Woebot—a well-known script-based mental health chatbot accessible only through a partner healthcare provider—it could not grasp the nuances of the issues he was discussing with his therapist. Instead, it suggested he “stick to the facts,” stripping his responses of emotional content, and recommended that he reframe his negative thoughts positively.
Lewis stated, “As a therapist, correcting or dismissing emotions is the last thing I would want a client to experience or ever advise.”
“Our role is to build a relationship that can accommodate difficult emotions,” Lewis continued, “allowing clients to more easily explore, integrate, or find meaning in those feelings and ultimately grow a deeper understanding of themselves.”
I encountered a similar situation with Earkick, a freemium Generative AI chatbot that claims to “enhance your mental health in real-time” and reportedly has “tens of thousands” of users. After expressing that I felt overwhelmed by increasing deadlines, it quickly recommended engaging in hobbies as a solution.
Earkick’s co-founder and COO, Karin Stephan, mentioned that the app is not designed to compete with human practitioners but aims to assist people in a way that makes them more open to seeking help.
How bots and people can collaborate
Most therapists believe that AI applications can serve as a beneficial initial step on someone’s mental health journey. The issue arises when these tools are seen as the sole solution. While individuals like Smith and Lewis had existing support systems from humans, the risks can be severe for those who rely solely on an AI chatbot. Last year, a Belgian man tragically took his life after a chatbot encouraged him to do so. Likewise, the National Eating Disorders Association (NEDA) halted an eating disorder chatbot, Tessa, because it was offering harmful dieting guidance.
Ellen Fitzsimmons-Craft, a psychologist and professor involved in developing Tessa, acknowledges that AI tools could make mental health care less intimidating but emphasizes that they must be safe, held to high standards, and properly regulated. She indicated that, like ChatGPT, they should not be trained using the entire internet, which contains much misguided advice. Research has shown that AI chatbots not only repeated racist medical stereotypes but also failed to operate effectively when applied to certain groups, such as Black Americans.
Until these issues are resolved, Rob Morris, co-founder of Koko Cares—an organization providing free mental health resources and peer support—suggested that AI’s most practical applications in the near term will be for administrative tasks like insurance and billing, thereby allowing therapists to dedicate more time to clients.
Koko faced public backlash when it introduced a function to co-author messages with ChatGPT and had to reverse that decision. When given the choice to involve AI, most users preferred a purely human experience and opted out. In the past six months, over 2,000,000 individuals have engaged with Koko.
“Individuals in distress are not merely problems to be solved,” Lewis asserted, “they are intricate beings deserving of attention, understanding, and care. It really is that straightforward.”
A new, dangerous virus spreading worldwide has heightened anxiety for many. The psychological impact of the pandemic can be particularly burdensome for those with pre-existing mental health issues. A 25-year-old from the US East Coast, who sees a therapist for anxiety, found additional support from an unexpected source: a chatbot.
“Having therapy twice a month was adequate before. Now, there are days when I feel I need something more,” said this person, who identifies as gender nonbinary and requested anonymity. Financial constraints limited their ability to increase therapy sessions, making them open to a recommendation from a friend about Woebot, a chatbot grounded in Stanford research that offers a digital form of cognitive behavioral therapy. It has become an integral part of their routine. “Being able to use the app daily is very reassuring,” they expressed. “It has helped me identify anxious traits and thought patterns I was previously unaware of.”
The Food and Drug Administration also believes that software can assist individuals grappling with the mental strains of the pandemic. The onset of Covid-19 prompted the agency to enhance the concept with a pandemic boost.
Since late 2017, the FDA has approved several apps and digital services that healthcare providers may prescribe for psychiatric disorders, similar to medication. This emerging market was anticipated to expand rapidly as regulators and healthcare professionals became increasingly receptive to the concept, while platforms like Woebot gathered the necessary clinical trial data for approval.
In April, the FDA relaxed several of its typical regulations regarding what it labels digital therapeutic devices for mental health disorders, aiming to expand access to care during the pandemic. This change allowed doctors to prescribe digital therapy that had not yet received approval and encouraged companies to hasten their efforts to develop and release applications.
One such company is Orexo, a Swedish pharmaceutical firm that focuses on treatments for substance abuse and primarily operates in the US.
At the beginning of 2020, it anticipated obtaining FDA approval for its inaugural digital product by the end of the year—a cognitive-behavioral therapy website for addressing problem drinking called vorvida, which trials indicated could significantly lower an individual’s alcohol intake. The company was also preparing to initiate trials this fall for another site targeting opioid use, and was looking to license a third one for managing depression. “We are now planning to launch all three this year,” states Dennis Urbaniak, head of Orexo’s digital therapeutics division.
The company is collaborating with health insurers and systems to provide vorvida to its initial US patients outside of a clinical trial within weeks. Urbaniak mentions that the web therapy will be priced competitively with how insurers are charged for psychotherapy or counseling conducted via video.
Pear Therapeutics, the creator of three FDA-approved cognitive therapy applications for opioid use, chronic insomnia, and substance addiction, is speeding up the development of a fourth app that focuses on schizophrenia.
When the pandemic emerged, the company was nearing clinical trials for the schizophrenia app, which features exercises designed to help individuals discern whether their experiences are real or merely hallucinations. CEO Corey McCann states that Pear intends to roll out the app to some patients this fall through collaborations with healthcare providers and academic institutions. He likens his company’s reaction to the FDA’s guidance for therapy apps to the compassionate-use program for remdesivir, the antiviral that received expedited approval for use in COVID-19 patients.
“Those undergoing recovery from substance use might find themselves awake at 2 am, feeling highly vulnerable to relapse, with no one to converse with.”
Lisa Marsch, director of the Dartmouth Center for Technology and Behavioral Health, expressed this sentiment.
Research has increasingly shown over the past decade that digital therapeutics can be equally or more effective than traditional treatment administered by doctors or therapists. Many of these therapies are rooted in cognitive behavioral therapy, which is viewed as the gold standard for conditions like depression and anxiety.
CBT involves structured exercises that prompt individuals to question and modify their thought patterns—a format that aligns well with a step-by-step software guide or chatbot. Orexo, Woebot, and Pear claim that they customize their services, directing patients to varied exercises based on their responses to inquiries.
Orexo’s vorvida gathers information about a person’s drinking patterns and treatment journey to customize the program—for instance, selecting exercises that may include guided meditation, journaling about consumption, and establishing and monitoring goals aimed at reduction. Recently, the FDA greenlighted an app designed differently, a computer game called EndeavorRx from Akili Interactive, which trials indicated can assist children with ADHD in enhancing focus.
A notable advantage of digital treatment is its constant accessibility, allowing it to fit easily into one’s pocket. Those undergoing traditional therapy rarely receive daily consultations, whereas a digital therapist on a mobile device facilitates ongoing engagement with assignments and provides support in critical situations.
“An individual in recovery from substance use may find themselves awake at 2 am, feeling at a high risk of relapse without anyone available to talk to,” remarks Lisa Marsch, director of the Dartmouth Center for Technology and Behavioral Health, and a member of Pear’s scientific advisory board. “However, they can access something in their pocket that aids them in responding to that moment in a way that does not involve relapsing.”
The US has been slower than countries like Germany to adopt computer therapy. In 2006, the organization that evaluates clinical evidence for England’s National Health Service first advised the use of computerized cognitive behavioral therapy for conditions like depression, panic, and phobias, noting it could increase access to treatment.
Alison Darcy, the CEO of Woebot and an adjunct lecturer in psychiatry at Stanford, believes this argument is also relevant in the US. Since 2017, the company has provided its app for free as a self-care option for individuals dealing with symptoms like depression and anxiety while it seeks FDA approval; currently, it exchanges 4.7 million messages with users weekly. “We simply don’t have enough clinicians and specialists available to treat everyone,” she states.
The 2018 National Survey on Drug Use and Health, conducted by the Substance Abuse and Mental Health Services Administration, revealed that 48 million Americans have some type of mental illness, with 60 percent not receiving any treatment. Of the 20 million Americans who suffer from a substance use disorder, 90 percent were not receiving care.
The FDA did not remove all restrictions on psychiatric apps. A notice in April lifted the requirement for clinical trial data submission but mandates that companies implement security measures, evaluate potential risks for patients using their app, and recommend that users consult their doctors beforehand.
This policy remains an ongoing experiment. Guidance from the American Psychiatric Association regarding mobile apps advises caution because digital therapies are novel and “not typically what psychiatrists and mental health clinicians are traditionally trained to provide.”
Bruce Rollman, who directs the Center for Behavioral Health and Smart Technology at the University of Pittsburgh, asserts that how physicians adjust to digital therapy will significantly influence the success of the FDA’s regulatory changes. He participated in a trial funded by the National Institute of Mental Health, which demonstrated that individuals with depression and anxiety benefited more from a program of computerized CBT than from the usual care provided by physicians, with effects lasting for six months. However, he points to another study as a cautionary tale, indicating that a randomized controlled trial involving nearly 700 patients in the UK showed computerized CBT did not yield superior results, primarily because of low engagement levels.
Rollman interprets this as a reminder that medical professionals must continue supporting patients who are using digital treatments, a practice that relatively few physicians in the US are accustomed to. “You can’t simply send someone a link to an appealing digital app or website and expect them to recover,” he emphasizes.
Leave a Reply