Using AI in Mental Health + Healthcare

TL;DR: 

Mental health conditions can be identified through techniques like Natural Language Processing and Phenotyping. The use of chatbots can help to target individuals with mental health challenges (like anxiety or depression), and evidence has shown that chatbots can be very effective in their treatment. In the future, we expect to see AI and digital technology continue to positively impact mental health care. 

Key terms:

• Chatbot 

• Digital Phenotyping 

• Natural Language Processing (NLP) 

• Artificial Intelligence (AI) 

• Digital Therapeutic Alliance (DTA) 

Introduction 

This article simplifies a research paper that talks about how AI is being used in health care to predict, detect, and treat mental health conditions. The content will focus on AI being incorporated into digital interventions, such as websites and smartphone apps. AI can enhance user interactions and help make mental health care personalized for individuals. 

The research paper focuses on 4 main ways that AI is being used in mental health: 

1. Personal sensing & digital phenotyping 

2. Natural language processing of clinical texts and social media context

3. Chatbots 

4. Digital therapeutic alliances

Digital Phenotyping & Personal Sensing	Digital phenotyping & personal sensing is the use of devices such as smartphones to collect and analyze data to conclude behaviors that can predict mental health conditions. AI is applied in digital phenotyping & personal sensing by gathering and reviewing data from devices like smartphones, as said above, and using it to interpret human behavior/health (look below for an example of the interpretation of human behavior/health). For example, when digital phenotyping & personal sensing is integrated into computerized programs such as chatbots mentioned below, the programs can deduct the signs of mental health struggles and can later notify users when it’s time to reach out for help.
Natural Language Processing (NLP)	NLP is the examination of how one talks or expresses emotion for mental health analysis. AI is applied in NLP by using digital technology to interpret human language patterns, such as sadness, loneliness, and hostility. It is even being worked on to predict the risk of psychosis and suicidal ideation. NLP is used in chatbots, as defined in the following column, and helps them understand human moods and behaviors.
Chatbots	Chatbots are programs that stimulate human interactions with a user through text or voice, and can perform tasks without an actual person on the other end of a conversation. AI is applied in chatbots via the use of apps or programs which generate fake human interactions, just like the ones people have on a daily basis. Chatbots such as Wysa offer therapy-based exercises for those struggling with mental health conditions.
Digital Therapeutic Alliance (DTA)	The term “digital therapeutic alliance” describes the general relationship between a therapist and their client over the course of online therapy sessions, such as videoconferencing. AI is applied in DTAs by computerized web apps such as chatbots mentioned above. The exploration of DTAs between human users and conversational agents is limited by considerations such as the ability to measure these alliances.

Why is AI being used in personal sensing or digital phenotyping?

AI is being used in personal sensing or digital phenotyping because it can use our data from our phones, tablets, computers, etc. AI can use clues about what we do in our daily lives. It also knows the signs of mental health struggles like anxiety, depression, and high stress. So when someone is in a crisis, AI can help and give reminders and alert doctors so people can get support sooner. 

Why is AI being used in natural language processing of clinical texts and social media context? 

AI is being used in natural language processing of clinical texts and social media contexts because we can reveal the way that we are really feeling. For example, the changes in our tone, speech speed, or the kind of words that we can post online can point to stress, depression, or other types of struggles. AI can pick up on these issues faster than people can. Some ChatBots like Woebot or Wysa can give a ton of support. They don’t replace real therapists, but they can make it easier to reach and it is less intimidating. 

Why is AI being used in Chatbots and Virtual agents? 

AI is being used in Chatbots and Virtual agents because Chatbots are AI programs that can talk with people through text or voice. Some Chatbots like Woebot, Wysa, and Tess, help with mental health by checking in with people, tracking their mood, and offering therapy-based exercises. These Chatbots and Virtual agents are available anytime and they can make support easier to access, but they don’t replace human therapists and they can’t handle emergencies. Some advanced versions can pick up on nonverbal cues to better understand users. 

Why can't AI replace human therapists yet and why does it need to be kept in check? 

Between AI chatbots and users, there is no way to show how strong the DTA (Digital Therapeutic Alliance) is. On the other hand, interactions between human therapists and clients are proven to result in deep/powerful bonds. These relationships can be a very big factor in how successful therapy is. Moreover, the increased use of AI in mental health care makes it important to have it be extremely reliable, especially since it could lead to ethical issues. It’s also important to note that there is no representation of actual mental health condition patients in AI, since it is a computer program. 

Conclusion 

AI and digital tools are making mental health care more accessible and effective, and they can help make it easier to spot individuals’ struggles and provide assistance for one’s outreach for support. Ultimately, AI can help mental health condition patients whenever they need moral support, even if it’s something as small as encouragement. Lastly, As technology advances, it can work with humans to create better care for everyone. 

The original article can be found at: https://doi.org/10.1016/j.copsyc.2020.04.005

This simplified article was written by Nicholas Smith and Noam Mistry