Developing chatbots to promote mental health
Increases in life expectancy have been accompanied by an increase in non-fatal chronic conditions, including mental illness. As stated by the World Health Organization, mental health disorders account for 26 % of disease impact in EU countries. “Despite the large number of people suffering from depression, anxiety or other mental health disorders, treatment coverage is extremely low,” notes MENHIR project coordinator Zoraida Callejas from the University of Granada in Spain.
Support, personalised care and resource delivery
The MENHIR project, funded by the Marie Skłodowska-Curie Actions programme, sought to address this challenge through developing chatbot technology to assist with around-the-clock support, personalised care and resource delivery. “To achieve this, we wanted to establish the strengths, limitations and requirements of mental health chatbots, and then co-create chatbot technology together with individuals experiencing anxiety or mild depression,” says Callejas. First, the project team was able to verify the importance of linguistic information (the words we choose to express ourselves), acoustic information (how we say them) and conversational behaviour (e.g. our silences during conversation or how we take turns with our interlocutors) to identify patterns indicative of anxiety and depression. The project then worked in multidisciplinary teams, which included end users, to develop conversational technology. “This co-creation approach enabled us to develop technology suitable for several scenarios,” explains Callejas. “For example, one solution was to implement a smart reminder chatbot that complements the support offered by non-profits, to help users to stay on top of the activity plans tailored for them through written and spoken interactions.”
Advancing digital mental health services
Callejas believes that the project was successful in demonstrating the usefulness of chatbots in promoting mental health. “They can provide a space for users to express their thoughts and feelings without feeling judged, reducing the stigma often associated with mental health,” she says. “They are also accessible at any time.” Other key findings included the importance of following an evidence-based, co-creation approach. The project team was also able to confirm that speech, linguistic and conversational features can be useful markers to detect anxiety and depressive states. “Chatbots should provide mechanisms that allow users to decide which information can be collected, and what is shared with healthcare professionals,” adds Callejas. “And while the ability of chatbots to start conversations is important, this behaviour must be predictable and explainable.” The project’s results have been shared across more than 100 publications in conferences, books and journals, and communicated to the wider public through blog posts, open days and outreach events.
Best practices in AI and healthcare
MENHIR has also had a positive impact on the early-stage researchers involved. Numerous training activities, including two doctoral summer schools and numerous scientific events, were organised. “The project has generated a sustainable network of organisations from the academic and non-academic sectors, which will continue beyond the end of the project,” remarks Callejas. “We believe that our lasting legacy lies in the enduring impact left on all MENHIR members.” The project was also one of the creators of the Cancer Survivorship – AI for Well-being cluster, which brings together 12 EU-funded projects. “The cluster’s aim is to stop working in silos and generate a forum of collaboration to share best practices in AI and healthcare,” adds Callejas.
Keywords
MENHIR, chatbots, healthcare, mental health, depression, AI, cancer