The rapid development of natural language processing and conversational interfaces has enabled a more progressive way of dealing with mental health problems, stress management, and psychological relief.
It's a big deal. According to the World Health Organization research, more than 300 million people are suffering from depression alone, not to mention other types of mental issues. Just a couple of months ago burnout was officially recognized as a medical condition. And this kind of thing is almost inevitable among professionals of any field.
According to a Scientific American study, the economic cost of depression in the United States accounts for hundreds of billions of losses per year.
And while the need for mental health services is getting higher, its availability is unable to catch up. In this environment, the traditional means of psychological relief and stress management aren’t efficient enough.
If you want to make Mental Health Chatbot, this article is right for you. In this article, we will look at:
- What mental health chatbots are?
- How do mental health chatbots work?
- Four major mental health chatbot applications and their business models;
A mental health chatbot is a type of a conversational interface application designed to:
- Have a conversation with the patient regarding his mental wellbeing;
- Provide instant 24/7 available chat;
- Deliver detached statistics for the patient to self-regulate his mental state;
- Give users basic recommendations on how to improve a patient’s mental wellbeing;
The primary goal of a mental health chatbot is:
- to help patients to manage and understand their mental states on their own as much as possible;
- connect with mental health professionals upon necessity.
Mental health chatbots originate from the very beginnings of natural language processing conversational interfaces - ELIZA.
Joseph Weizenbaum developed this chatbot in 1964-1966. Originally, ELIZA was proof of the concept “to demonstrate the superficiality of communication between humans and machines.” However, ELIZA quickly proved itself to be more than that.
As it turned out, ELIZA was good at talking with people. At its core, ELIZA had mere pattern matching and substitution scripts that gave the illusion of machine understanding of the user’s input message.
One of its scripts, titled DOCTOR, imitated a classic person-centered psychotherapy session blueprint. It wasn’t anything sophisticated - just a couple of template phrases, but it worked incredibly well.
Despite its simplistic design, ELIZA was engaging enough to let people speak out about their problems (which is the simplest model of providing psychological relief). This laid the groundwork for future healthcare chatbots.
These days, mental health chatbots are not just a couple of template phrases that imitate language understanding.
Modern mental health chatbots integrate into the healthcare system and involve certified medical professionals. These chatbots automate specific processes and also streamline the interaction between the patient/user and mental health professionals.
Mental health chatbots are designed to maintain a conversation, not to lead it. In a way, using a mental health chatbot resembles practicing tennis against a wall.
The general functional framework of mental health chatbots is based on Cognitive Behavioral Therapy methodology. CBT is a form of talking therapy designed to manage mental health states by rearranging the way the patient perceives it, i.e., making negative thoughts positive.
The list of chatbot features revolves around:
- Kickstarting a topic for conversation,
- asking directional questions,
- providing follow-ups to expand responses.
The NLP algorithm, with Sentiment Analysis features, handles the flow of the conversation. It recognizes keywords and terms. Each trigger word has a decision tree. It is designed to gather as much information as possible and provide a viable resolution - either a simple conclusion, a piece of advice or contact professional help.
The critical design component in mental health chatbots is the so-called empathetic engagement.
In the context of a conversational interface, empathetic engagement means:
- making the impression of a credible and trustworthy conversation partner that can hear you out and offer a detached point of view on things.
Since the user is already aware of the artificial nature of the conversational interface, there is no need to bend over backward to imitate a full-blown human-human conversation. Instead, the chatbot needs to provide the necessary minimum credibility to enable the user’s suspension of disbelief.
One of the biggest challenges of mental health chatbots is privacy and confidentiality. Since the entirety of user activity is related to personal matters and thus is sensitive information - it is necessary to address this issue.
The most effective solutions for this are:
- End-to-end encryption of the user-bot interaction;
- Making the user profile in the application database anonymous.
Now let’s look at several major mental health chatbot applications and explain their inner workings.
Both Moodkit and Moodnotes are traditional paid apps that charge $4.99 for an install, after which you get the full scope of features and customer support.
Moodkit was the first mental health app of the company. Launched in 2011, it is less of a chatbot and more of a mental health CRM. Moodkit was one of the trailblazers of mental health applications in the early 2010s. From a modern point of view, it is more of a proof of concept that was later perfected in Moodnotes.
Moodkit's framework is specifically designed to engage users as much as possible to have control over their mental state and substantially improve it. In a way, Moodkit motivates the user to improve his mood and mental state with its handy tools.
The application itself consists of four tools:
- Moodkit Activities - a tool that suggests what things to do to improve the user's mood (for example, take a walk or talk to somebody);
- Thought checker to manage negative feelings caused by a specific situation. It is used to identify and subsequently reiterate the thoughts into more positive ones;
- Mood Tracker - designed to chart the permutations of the mood over the day, week, month, and so on;
- Moodkit Journal - a kind of a notebook for a user to keep thoughts and comments on his mood and related events.
Moodnotes is a kind of more sophisticated and elaborate version of Moodkit. The app was released in 2015, its framework is similar to Moodkit, but there is more thorough automation involved. Moodnotes applies Natural Language Processing and Sentiment Analysis to provide more profound and helpful insights into the user's mental state.
- The user needs to describe his mood of the moment, rate it, and add some comments
- Next, the app runs an NLP Sentiment Analysis algorithm that recognizes the polarity of the mood, and also its pattern, based on the available user activity.
- After that, the application's chatbot engages with the user in a conversation to determine the current thought pattern of the user
- The questions are designed to reflect emotions and iterate them into more positive patterns
As a result, the bot provides some words of encouragement in a manner of "keep on keeping on." It is very similar to how Grammarly engages with users to methodically fix texts.
At the time of writing this article, Woebot is one of the most prominent players in the field of mental health chatbots with over 100k downloads in GooglePlay alone. Founded back in 2017, it managed to take over the game by simply making the most efficient version of the product. In 2018, Woebot had managed to raise $8 million in series A funding.
Woebot is a distillation of the best conversational interface solutions of the decade, streamlined and adapted for healthcare purposes. The other reason for its rising popularity is the fact that it is a free application.
At its core, Woebot is a chatbot that keeps an eye on the mood of the user. It provides a platform for the user to speak out, contemplate, and reflect through meticulously designed conversation trees.
Every once in a while the bot asks the user how he feels and how he's doing. In response, the bot provides some handy advice or drops relevant contacts with mental health professionals if the situation is that dire.
- The NLP component provides a necessary level of personalization and spices up the conversation with a bit of humor
- Sentiment Analysis is used to identify critical patterns in the user's input and drive the conversation in a more positive direction
The continuous use of application accumulates its efficiency. There is daily check-in that provides the bot with the necessary information for analysis and, that in turn contributes to better healthcare advice in the future.
According to 2017 research by Stanford School of Medicine, throughout a two week test period, mental health chatbot users reported a decrease in anxiety and depression compared to a control group.
Wysa is a kind of more ambitious spin on a mental health chatbot concept. Launched in 2015, Wysa is a result of a collaboration between Columbia and Cambridge universities (and also Touchkin, who did the development).
Wysa markets itself as an "emotionally intelligent" chatbot to manage emotions and thoughts. The bot itself is built on a Facebook Messenger framework, compatible with IOs and Android so that it is available to as many people as possible.
One of Wysa's key competitive advantages over other mental health chatbots is its scope.
- In addition to the standard Cognitive Behavioral Therapy method, which handles more casuals moods, Wysa applies Dialectical Behavior Therapy (something Woebot is currently working on)
- It is a big deal because DBT aims at unhealthy, suicidal, and self-destructive behavior, and finds ways of getting out of such situations
The other Wysa innovation is the addition of meditation and yoga advice. While it is not everyone's cup of tea, it might be a solution for some.
The rest of the framework is similar to Woebot. The user interacts with the bot, provides some information, and, based on that, the bot gives a custom response. There are daily check-ins and detailed stats.
The other features that differentiate Wysa from the rest are in-app purchases. There is a subscription fee ($29.99 monthly) that enables a more personalized approach with a human operator.
Sanvello is a more glamorous variation of mental health chatbots. Developed by PacificaLabs in 2014, Sanvello (known initially as Pacifica) is more inclined to so-called "mindfulness" than straightforward "psychological relief" that is the core of CBT-based chatbots.
As such, Sanvellois a much broader tool that adds to mood tracking and management meditation and relaxation features.
Unlike the other bots on the list, Sanvello adds audio and video recognition tools to traditional text input. The use of image and voice machine learning algorithms helps to determine the state of the user and provide some relevant and helpful advice.
The app's features include:
- General mood tracking for everyday use
- Human help feature for emergency cases
- Coping tools to handle stressful situations
- Progress assessment feature to see things from a long-term perspective
The other significant feature of Sanvello is the social element. Unlike other mental health apps, Sanvello connects users to share experiences and provide mutual support. While the quality of expertise of such help might be questionable, the concept itself is worth exploring.
Sanvello's business model is more in line with traditional mobile applications. There is a free version with a basic set of features, and a premium version with more tools and in-depth customization of service for a monthly ($8.99), yearly ($53.99), or lifetime ($199.99) fee.
Mental health is one of the healthcare fields that require cutting edge technologies to provide an effective and equally available service for anybody who needs it.
The conversational interface seems to be a viable solution capable of handling basic needs for anxiety and depression management. Even if the ultimate benefit is encouraging people to speak out about their worries and relieve stress - that’s already a giant step forward.
In this article, we have looked at four different approaches to mental health chatbots.
Such chatbots can significantly contribute to the refinement of natural language processing and sentiment analysis algorithms. Interaction with humans and different case studies may drastically expand the scope of language models and their capabilities.
What is even more important, in the long term perspective, this will help in further psychological research.
Drop us a line