Artificial Intelligence for mental health

Whichever way you look at it, mental health is a colossal public health issue, and the current pandemic has made things worse. The Centre for Mental Health estimates that up to 10 million people will need either new or additional mental health support as a direct consequence of the Covid-19 crisis.

So how do we build scalable and cost-effective mental health interventions for the UK population? Not just to help us now, but also in the uncertain future that we face.

This is where I think artificially intelligent ‘chatbots’ and mobile technology can help. Chatbots apply a range of technologies and approaches, including natural language processing to simulate human conversation. You might have seen them on websites providing text-based assistance. I believe that combining this technology with the near ubiquitous ownership of mobile phones offers an exciting new way to enable individuals to access mental health support.

The combined use of these tools offers some significant advantages to the status quo. These include:

• 24/7 availability: Using mobile phones as a platform to provide mental health interventions means that we can ensure service users always have access to support, even in times of crisis. For example, if someone needs help dealing with a panic attack in the middle of the night, they can access a breathing exercise on the one thing they almost certainly have with them - their mobile phone.
• The opportunity to reach more people: Once a digital service is up and running, it can scale rapidly. This means that more people can access care quickly and with minimal extra cost.
• The ability to replicate effective interventions: The advantage of conversations created with AI chatbots is replicability. You can be sure that each individual is receiving a similar level of care, and strong quality assurance can take place on the chatbot exchanges.
• Continuous learning: Every conversation that a chatbot has can improve its efficacy. In this way it’s like a human. The more it does, the better it gets. That means it can also adapt to new situations and audiences.
• The ability to tackle stigma and increase accessibility: Making something available to individuals digitally can help people overcome many of the barriers that they face in accessing care, whether these are logistical (such as work or childcare) or psychological (like the stigma of asking for help).

To be clear, I don’t believe that digital interventions can ever replace the empathy and skills of a trained, human professional. But thinking about AI and humans as mutually exclusive is not helpful. Instead we need to find ways for these new digital services to work alongside existing provision. For example, digital interventions could be used to provide rapid support to individuals with less severe conditions, or could be used in combination with a human therapist.

This is why I’m so grateful to the Churchill Fellowship and its Covid-19 Action Fund for supporting me to develop an AI-based mental health coach. The aim of the funding is to prototype an initial version of the tool, so that I can understand what technology is required and then test it with service users to ensure that it meets their needs.

I’m midway through the project and getting really positive results. There’s been a lot of research and trialing of different technology and methodologies. Technically I’ve ended up using the Open Source platform RASA. It’s very powerful and is enabling me to rapidly prototype and test the conversations. I’m largely basing the therapeutic intervention on third-wave therapies.

The early results from my research and testing are very interesting and encouraging. It seems that people are so used to speaking in text-based conversations that receiving support in this format feels very natural. On top of this, the interactive nature of the conversation really helps people to stay engaged with the therapeutic content. Lastly, we’re finding that accessing support in this way is easy, convenient and fits the pace of modern life.

That said, there’s still a lot of work to do. The next steps are to develop more conversational content, and validate the clinical effectiveness of the AI coach with the support of independent psychologists. But after the initial success of this early prototyping, I am hugely excited about the future possibilities.