How can AI be used to improve mental health care?

Therify is a mental health startup using artificial intelligence to help match patients and mental healthcare providers. The company was founded by James Edward Murray with the mission to make finding a therapist easier for underrepresented communities.

Therify’s website doesn't go into much detail as to how they use artificial intelligence but by understanding the goal of Therify’s system — to match patients to providers — we can start to imagine different ways AI and machine learning could be useful.

While it is not immediately obvious that you would need AI for a problem of matching patients to providers, it could still be useful. For example, if you give patients the option to choose from five types of issues like clinical depression, anxiety, body dysmorphia… you could then build a rules-based system to match them to a provider who specializes in the given issue. However, it is not always clear what category of mental health illness you fall into - which might be why you are going to the therapist in them first place - in that case, instead of telling a platform, “this is the clinical issue I’m facing” you might write a short description of the challenges you have been facing. In that case, you could use machine learning to detect different patterns in the text that map on to different issues. Then once you have categorized someone's issue, the system can match them with the most appropriate provider. 

Another way in which you could use machine learning to more accurately match patients and providers, is to use historical data on patient-therapist matches to predict the likely success of a new match. For example, if you had data on 1,000 different therapist-patient matches and a basis to track success, you could use this information to train a model with which you could input a patient’s information and output a patient-therapist match with a given likelihood of success. However, that leads us into the various questions that are outstanding for a company like Therify, among them, how do you measure success? 

For a company like Therify, in order to use machine learning, they would need to use data and potentially lots of it. But what data, or better yet, whose data are they using and how? In order to bring value to customers on day one using machine learning, they will likely need to use historical data and not just user data. Another question is whether they will or are using machine learning. While articles have cited the company as “using AI”, this is such a general term that it could simply mean that they are automating some tasks that are generally regarded as relying on human intelligence. Automation doesn’t immediately imply machine learning (training a model to learn and improve the performance of a task given data), it could be a system that relies on hard-coded rules. The last question and likely most challenging is, what does success look like and how is it measured? Is success just finding two people who are compatible and get along? Or does it mean a match that leads to many therapy sessions? Or is success only when someone is “cured”? 

Therify offers an innovative solution to the problem of connecting patients in underserved communities with mental health providers but there is still a shortage of desired providers. Hopefully, along with connecting more people to the resources they need, Therify will help raise awareness of the shortage of mental health care providers in minority communities.