Assesment of depression and anxiety with AI-analysed language

Assessment of mental health is currently conducted quantitatively with rating scales and clinical interviews. Rating scales have limitations as they are not the natural way for people to communicate their mental health. Clinical interviews, on the other hand, have drawbacks as they lack quantitative measures, and are fundamentally based on intuitive and subjective interpretations of trained clinicians with limited time.

New possibilities with AI

Recent progress in natural language processing, (NLP) and machine learning, (ML) opens up new possibilities. A promising alternative is to pose specific mental health-related questions to patients that they answer in their own language and analyse their responses with methods based on artificial intelligence.

We apply this question-based computational language assessment method, (QCLA) to depression and anxiety and show under what circumstances it provides good, or better validity than the conventional rating scale.

Important questions here are:

• How you can combine rating scales and open questions to improve assessment accuracy
• Which questions provide the best validity
• If it's better to have a single or a combination of response formats
• How many words do respondents need to make precise answers
• How many open-ended questions are needed for good assessments, and
• How can you visualize high and low depression in word clouds?

Finally, we investigate what attitudes patients and clinicians have towards language-based assessments of mental health.

Register to participate

Please register via #frAIdays website, and we will send you a link in good time before the event open to everyone who is interested in AI!