My therapist is a chatbot: Using AI in mental health 

The use of artificial intelligence to transform health and deliver mental healthcare is continuing to expand. The sophistication of the programming used in AI therapy tools has particularly come a long way in recent years. While we have not seen wide use of chatbot therapy yet, the next few years of use will provide a roadmap for their optimal application.

The most common application of AI in mental healthcare is using logic to determine triage where patients are routed to the correct treatment and the right provider based on self-identified symptoms, behaviors and specific demographic characteristics. Another use of AI not only identifies symptoms, behaviors and demographics but will also take into consideration medical claims data to look for synergies for a unified treatment plan.

The idea of a therapist AI tool is not new. In fact, it was one test in 1950 of the Turing Machine, invested in by AI pioneer Alan Turing to see if humans could be convinced that a computer was actually a human. The tool was rather simple, and it could be tricked, but many subjects reported feeling understood by the computer program thinking it was in fact a human.

In recent years, there are a number of companies that are using AI to help perform therapy. In these scenarios, unlike the Turing Machine, the goal is not to trick the subjects into believing the program is human, but rather to offer "human-like" support for their concerns.

However, given the limitations of the service model, chatbots are not expected to take the place of more traditional therapy, at least not for the near future. Many of us, for example, have experienced trying to use AI chatbot online help support. While some of these services do offer personalized support, the most common responses have not lived up to their promise, with most users eventually entering "Speak to a representative” in the chat.

Moreover, the unique relationship between therapist and patient provides the opportunity to redefine past maladaptive relationships and build trust to help the patient move on and create more positive relationships. The therapeutic relationship works to heal many past relationship scars. This can currently most effectively be accomplished through a human relationship in the interaction of a therapist and a patient.

There remains a great deal of potential for the use of chatbot tools in mental healthcare treatment. Recently, we have seen examples of self-help tools (without AI) creating clinical improvement in some populations. The person seeking help will be given access to tools – procedures they can read through or watch and then take some of those examples and use them in their own lives.

Also, a number of organizations are now employing "coaches" who are specifically trained in these self-help tools to help guide the individual on their therapeutic journey. It should be noted, the online support and coaching tends to be most effective with sub-clinical or less profound mental health issues.

While there could yet be a more impactful place in the mental healthcare ecosystem for chatbot therapy, at this point, we have only just seen the advent of AI therapy. Research needs to be conducted and repeated to see if these tools do in fact produce sustainable improvement and for what issues and presentations. For now, chatbot therapy is perhaps best understood as being a step up from self-help and just below coaching, essentially a next level of self-help with some extra support programmed in.