Lifeline, a not-for-profit organisation which started in 1963, have reached and benefited the mental health of hundreds of thousands of Australians.
About 1 million calls each year are made to its national telephone hotline, wherein, according to research, an estimated 3% – 5% of the Australian population has called at some point.
However, few studies have been undertaken to identify the outcomes achieved for callers to the service in the half a century of its operations.
Was the people who used the service able to derive help appropriate to their needs?
More than A$ 1.1 million has been set aside for a 5-year study that will enable a multi-institutional, multi-disciplined analysis of the organisation’s crisis support services, both telephone and online, and the impact they make for those who use them.
A team from the University of New South Wales Faculty of Engineering will be looking at how machine learning can be used to analyse the vocal tone of people calling in to identify the kind of help they need.
Artificial intelligence (AI) methods will be used to automatically identify different types of help-seekers based on their vocal qualities.
Moreover, AI will also be able to examine written communication in cases of online chat and SMS text.
The analysis of acoustic and linguistic information from the speech of crisis callers is a highly novel research area, where there is significant potential for new technology to contribute.
While this is still somewhat speculative, machine learning could be ever-present, listening in to calls to triage them while contributing to the organisation’s long-term strategies for supporting help-seekers.
Most probably, the crisis supporter will be the first responder, but the machine learning will contribute by cueing other specialists to get involved part-way through the call.
It will also help assess how many distressing calls a crisis supporter has had to handle.
Analysis was made on voice samples of people who are depressed and compared them with those who are not, focusing exclusively on the acoustic qualities of the voice.
As a person speaks, the qualities being observed include the way it sounded, the timbre, and the prosody, which is the intonation and changes in energy.
The results have been quite striking. There are tell-tale characteristics of a depressed person’s speech, such as flatness in tone, low energy and lack of expressiveness.
These characteristics can be automatically extracted using signal processing methods.
The tests so far were conducted mainly in controlled laboratory conditions with high-end audio equipment.
The research challenge is whether the same patterns can be detected using the audio conditions of a telephone call.
Researchers are confident that machine learning in the call centre conditions will be up to the task based on experience working with a smartphone app start-up, doing something similar to this.
Academics with backgrounds in psychiatry, psychology, sociology and engineering from eight Australian universities and one in the US will be contributing to the study.
Lifeline has moved into the digital age and offers crisis support via online chat and soon via SMS text messaging.
This research is expected to enable crisis supports to be able to better meet the needs of community members in crisis, by using advanced technology and research methods.
The project will directly impact help-seekers by ensuring they are provided the most appropriate crisis support at the time they need it most.
In all ways, the focus on outcomes and the monitoring service performance is a way of putting the person first, something that is deep within the tradition and culture of the organisation.