A study published Jan. 27 in Mental Health Research found an AI model predicted substance use disorder defining behaviors with up to 83% accuracy and addiction severity with up to 84%.
University of Cincinnati researchers Sumra Bari, PhD, the study’s lead author and senior research associate at the university, and Hans Brieter, MD, professor in the college of engineering and applied science, said in a Feb. 5 news release the tool could help clinicians identify substance use disorder, particularly in cases where patients underreport symptoms because of stigma.
The study analyzed data from 3,476 participants ages 18-70 who completed questionnaires and rated how much they liked or disliked 48 “mildly emotional stimuli,” such as a photo of a beach. Researchers used the image ratings to quantify mathematical features of human judgment, including variables commonly linked to behavioral economics, and combined those data with a small set of demographics to train AI algorithms.
The system also identified the type of substance used — stimulants, opioids or cannabis — with up to 82% accuracy. Statistical analysis showed participants with higher disorder severity were more risk-seeking, less resilient to losses, showed more “approach behavior” and had less variance in preferences.
“Anyone with a smartphone or computer can do the picture rating task,” Dr. Bari said in the release. “It’s low cost, scalable and resilient to manipulation.”
The method could eventually extend to other additions, including behavioral addiction such as excessive social media use or gaming, she said.
Read the full study here.
