AI Can Predict Anxiety Levels Using Picture Tasks, New Study Shows

Researchers at the University of Cincinnati have developed an artificial intelligence (AI) tool that uses picture tasks to predict anxiety levels.

This unique approach involves a brief task where individuals rate pictures and answer a few contextual questions. The AI then analyzes this data to determine if the person is experiencing anxiety.

'Comp Cog AI'

Sumra Bari, the first author and a senior research associate at UC's College of Engineering and Applied Science, emphasized that anxiety is a common experience at various levels and stages of life.

Bari noted that the new method employs minimal computational resources and a small set of variables, focusing on processes crucial to judgment. This approach, named "Comp Cog AI," integrates computational cognition with AI.

The team's approach involved a brief task where participants rated pictures as positive or negative and answered a few contextual questions, such as age and loneliness.

Hans Breiter, co-author and contact principal investigator of the study, highlighted the challenges of interpreting predictions from big data. Breiter pointed out that having a small number of variables rooted in mathematical psychology helps to address these challenges.

He added that this method supports what other AI scientists call a "standard model of the mind," providing a foundation for artificial general intelligence. The researchers concluded that judgment measures combined with some demographics play a crucial role in predicting anxiety levels.

The team hoped that the system could be a prototype for tools like apps that medical professionals, hospitals, or the military could use to identify individuals at urgent risk of anxiety.

The study involved 3,476 participants whose demographics mirrored those of the United States based on Census Bureau data. Participants answered questions about their demographic characteristics and perceived loneliness. They also rated their liking or disliking of 48 pictures with mildly emotional content.

According to the study, this picture rating data was leveraged to quantify mathematical features of people's judgments, which, combined with machine learning algorithms, predicted anxiety levels with up to 81% accuracy.

AI for Anxiety

Aggelos Katsaggelos, co-senior author and the Joseph Cummings Professor of Electrical and Computer Engineering at Northwestern University explained that understanding patterns in picture preference helps uncover critical components for various behaviors.

Bari added that the picture-rating task, combined with contextual variables affecting judgment, provides unbiased snapshots of a person's mental health without direct questions that might trigger negative feelings.

Anxiety affects about 12% of the U.S. population, manifesting as intense fear and persistent worry without a clear threat. It impacts mental and physical health, relationships, and overall quality of life.

The new tool aims to provide a simple yet effective method for identifying this condition. The study's findings were published in the journal npj Mental Health Research.