Personally loving words and language, I was recently sidetracked by a news item highlighting the use of AI to analyse speech patterns to determine the mental health of teens and young adults. The news headline by Medical Today read: ‘AI model may predict Alzheimer’s by analysing speech patterns.’
The actual article explained that groundbreaking research had been done at the University of Edinburgh using artificial intelligence to analyse speech, to detect risk of psychosis in adolescents and young adults.
While speech analysis itself is nothing new, the leverage to this field of science through AI has ‘groundbreaking’ implications, according to researchers quoted in the news.
A wealth of insights into mental health can be gained through speech analysis. Speech analysis is used in psychology, psychiatry, and neuroscience to help identify patterns linked to mental disorders.
A person’s tone when they speak, the speed of their words, the pauses they use, their word choice, their chosen sentence structure, and the rhythm and emotional expression can prove invaluable when analysing their mental and emotional state. To give an example, people with depression usually speak slower and with a flatter tone than others. Their language might contain more negative or self-focused words. Those with anxiety disorders tend to speak with rushed, shaky, or hesitant words. Someone experiencing mania in bipolar disorder may speak very quickly and jump between ideas, and struggle to stay on one topic.
Schizophrenia is also identifiable through speech patterns. Here, disorganised speech, unusual word associations, or fragmented sentences can indicate an underlying mental condition. If these subtle nuances in speech are noted, they can help doctors to detect psychosis early.
More intriguingly so, the use of AI, such as through the studies done by the University of Edinburgh, has leveraged this speech analysis to new levels of accuracy. This ‘groundbreaking’ AI-driven approach could, in the words of professionals interviewed, transform the way clinicians identify at-risk individuals before full-blown psychotic episodes develop.
What is Psychosis?
Psychosis is characterised by symptoms such as hallucinations, delusions, and disorganised thinking. It often emerges during late adolescence or early adulthood.
Early detection is critical. Studies show that interventions at the prodromal stage; that is, the period before psychosis fully manifests; can dramatically improve outcomes, reduce hospitalisation rates, and increase long-term recovery prospects.
Traditional diagnostic methods have typically relied on clinical interviews, behavioural observations, and self-reported experiences. However, these assessments are often subjective and can miss subtle early warning signs.
The University of Edinburgh study, published in Nature Digital Medicine, employs AI algorithms to examine minute speech features in real-time.
For the study, researchers collected thousands of audio samples from participants aged 15 to 25, analysing variables such as speech coherence, pitch variation, response latency, and semantic consistency. The AI models detected patterns which correlated strongly with established markers of early psychosis, with accuracy rates surpassing 80 percent in preliminary trials.
Dr. Fiona MacLeod, lead researcher on the project, explained: “Young people often experience symptoms internally and may not recognise them as early signs of psychosis. By using AI to identify subtle shifts in speech patterns, we can flag at-risk individuals earlier than ever before, allowing for timely clinical support and intervention.”
This technology could revolutionise mental health services by providing an objective, scalable, and non-invasive screening tool.
Practically speaking, teens could undergo brief speech assessments during routine appointments, in schools, or even through mobile applications, with results automatically analysed by AI algorithms trained to detect early warning signs. Unlike traditional diagnostic approaches which depend heavily on patient self-reporting, this method uses quantifiable markers, potentially reducing the risk of misdiagnosis and delayed care.
A Potential For Wider Application
Experts have said that the benefits don’t end there. Professor Andrew Thompson, a consultant psychiatrist at King’s College London, commented: “Early detection is the holy grail of psychosis prevention. If AI can reliably identify youth at risk based on objective speech markers, it could significantly shorten the time to intervention and improve outcomes across the population.”
According to the UK Mental Health Foundation, approximately one in four young people experience a diagnosable mental health disorder before the age of 25. Still, many remain undiagnosed or receive delayed care.
Tools which enable early, objective screening could both improve individual outcomes and also reduce the strain on healthcare systems.
Importantly, the study underscores the potential for AI-driven diagnostics to combat stigma associated with mental health. Adolescents may be hesitant to seek help due to fear of judgment or misunderstanding. However, unobtrusive, data-driven screening could normalise early monitoring as part of routine care. “We want to create tools that empower young people,” Dr. MacLeod emphasises. “Early detection is about providing timely support and fostering resilience before a crisis occurs.”
It goes to show that words really do have power. We already know that they can be used as an avenue to bring healing, however, this research suggests even greater possibilities than could be imagined.