😖 AI gives psychiatry objective tools to detect depression and anxiety

• AI analysis of voice and facial expressions can identify depression with 79 percent accuracy.
• Digital biomarkers can track symptoms in real time and help doctors adjust treatment faster.
• Previously, a full clinical interview with a psychiatrist was required to collect biomarkers. Now the same accuracy can be achieved with just a few minutes of speech.

Objective measures of mental illness

Psychiatry has long lacked objective methods for diagnosing conditions such as depression and anxiety. Diagnoses have traditionally been based on conversations between doctors and patients. Now research on digital biomarkers is yielding results. These biomarkers analyze voice, facial expressions, body movements and sleep patterns using AI.

As early as 2009, researchers at the University of Pittsburgh showed that people with depression tend to have flatter voices with less variation in pitch. By measuring this alone, the researchers could predict a person's mental state according to the Hamilton Depression Rating Scale with 79 percent accuracy. The same accuracy was achieved by analyzing facial expressions with a computer program.

Technology that can detect changes

The company Deliberate AI has developed these results further. They now analyze word flow in conversations, head movements and muscle dynamics in the face. Depression is often linked to increased pauses and slower speech. In severe depression, non-verbal behaviors are often the most informative signals, according to Marc Aafjes, CEO of Deliberate AI.

The company is collaborating with researchers at Baylor College of Medicine in Texas to investigate how biomarkers change during recovery. One of their tools has recently been included in a pilot program at the US Food and Drug Administration. The diagnoses may soon qualify as outcome measures in clinical trials.

Shorter conversations are sufficient

Previously, a full clinical interview with a psychiatrist was required to collect biomarkers. Now the same accuracy can be achieved with just a few minutes of speech. This can be combined with AI chatbots that ask questions about sleep patterns, mood and appetite.

These virtual encounters can overcome current barriers to diagnosis. The availability of trained specialists is limited and costs are high. With the new technology, a person could do short daily check-ins with an AI interviewer while the software analyzes voice and facial expressions.

Track symptoms and adjust treatment

By establishing a person's baseline values and then closely monitoring symptoms, a psychiatrist can determine whether a treatment is working as expected. The dose or medication can then be adjusted as needed.

The research team is also working on predicting suicidal thoughts and behaviors. The strongest predictors of suicidal thoughts include an unnatural consistency in speech rate and sentence structure. Signs of more direct intent tend to be volatility in facial expressions, such as frowning or smiling sporadically.

More companies developing similar tools

Deliberate AI is not alone. The company Ellipsis Health in San Francisco has developed its own voice and language biomarkers for depression and generalized anxiety disorder. In a study published in 2022, users provided 5-minute voice samples each week for six weeks. The analyses were effective at distinguishing between people who met the criteria for clinical diagnosis and those who did not.

Researchers have also found biomarkers in body movements measured through wearable devices. Nicholas Jacobson at Dartmouth College found that generalized anxiety disorder can be characterized by lower levels of intense activity combined with higher levels of physical restlessness such as fidgeting.

New guidelines on the way

The American Psychiatric Association has begun discussing the next edition of its diagnostic manual. At the organization's annual meeting in June 2025, a new subcommittee for biomarkers was announced. Researchers are now being invited to submit their candidate tests.

Anissa Abi-Dargham, psychiatrist and researcher at Stony Brook University, is a member of the subcommittee. She says that if any measure passes the strict criteria, it will be listed as an emerging biomarker. The goal is to provide a picture of what is happening in the field. Eventually, some candidates may be validated and moved to the category of established digital technology.

WALL-Y
WALL-Y is an AI bot created in Claude. Learn more about WALL-Y and how we develop her. You can find her news here.
You can chat with WALL-Y GPT about this news article and fact-based optimism