Chatbot success signals new dawn for mental health services

A new study has shown that an AI chatbot has led to an increase in mental health referrals among minority communities in England. 

Research examined data from 129,400 people visiting websites to refer themselves to NHS Talking Therapies services across England. Half of the observants used the chatbot, made by Limbic AI, and half used web forms and other data-collection methods. 

The chatbot increased referrals by 179% for those identifying as nonbinary, and 40% and 39% for Black and Asian patients respectively. 

The chatbot would ask about a patient’s medical history and their current symptoms, refining its questions to the symptoms flagged most relevant to the patient’s problems. It would then create a detailed referral for an NHS professional to assess. 

This process uses two distinct AI models: natural language processing tools to analyse a patient’s responses and to provide relevant answers, and probabilistic models which match the patient’s data and chatbot’s responses with the patient’s most likely mental-health problem. The chatbot is currently capable of identifying 8 mental-health issues with 93% accuracy, according to the article’s authors.

Limbic AI’s success comes as NHS mental health services are facing increasing demand; referrals increased by 44% between 2016-2017 to 2021-2022. With another young person’s mental health service, Shropshire Beam, announcing its closure earlier this week, could chatbots offer an alternative pathway to care? Could AI technology be the cost-effective solution to NHS cuts increasingly targeting low-level mental health interventions?

An NHS psychiatrist, specialising in child and adolescent treatment, shared their scepticism about this with The Monitor. 

While recognising that chatbots like Limbic serve as a gateway to help to refer those from communities with ‘less awareness, more stigma and different cultural experiences of mental ill health’, the doctor contested that chatbots are ‘not a viable alternative’ for diagnosis or therapeutic intervention.

‘Most important to diagnose’, unlike to refer, ‘is to consider symptoms in context of the whole person (the biopsychosocial model) and in the treatment relationship with the professional’. The therapeutic relationship, the doctor continued, depends on a ‘real relationship’; that is, a relationship between humans. 

While the extent that AI technology could lower the barrier to care is contested, it is clear that chatbots like Limbic are contributing to making the UK’s mental health system more accessible. And amidst increasing cuts, that can only be a good thing.