OG Article: here
View our Fair Use Policy: here
An artificial intelligence model can detect if someone has recently used cannabis. This could one day help to identify intoxicated people who may require medical care, but some experts have stressed that further research is required with a larger group of people.
Urine, saliva and hair strand tests can reveal if someone has used cannabis, however, these can take a fair amount of time to analyse …
and cannabis use does not immediately show up in some bodily fluids. Sang Won Bae at Stevens Institute of Technology in New Jersey and her colleagues wanted to create a way of quickly assessing whether someone may be dangerously intoxicated.
They therefore had 33 adults, who used cannabis at least twice a week, report their use every day for up to 30 days.
During the study period, the participants wore an activity tracker to collect information on their heart rates, step counts and sleep quality. Sensors on the participants’ phones also provided information on their micromovements, such as how they held their phone, to gauge their stability and coordination.
Data from some of these participants was used to train the AI to detect if someone may have used cannabis. The researchers then tested the trained AI on the remaining participants’ data. When taking into account the number of false positives and negatives that the AI flagged, it was 85 per cent accurate at detecting someone who had been moderately intoxicated within the past 5 minutes.
The AI can continuously interpret data collected from the activity tracker and sensors. Medical professionals could look back at historical data stored within the AI and see when any unusual activity started, which may suggest when someone became intoxicated. “It’s about preventing harm from happening and keeping people safe,” says Tammy Chung, one of the researchers, at Rutgers University in New Jersey.
The researchers also didn’t report how much tetrahydrocannabinol (THC), the psychoactive component in cannabis that makes users feel “high”, the participants consumed or their route of cannabis administration, such as smoking versus oral ingestion, says Chandy. Both of these can affect a person’s degree of cannabis intoxication, he says.
Chung says that the accuracy of the AI depended on how the participants self-reported their cannabis use. “If they’re not reliably reporting on that, that could be a problem for the [AI] model,” she says. Wu says that the Drug Effects Questionnaire, which asks people to report how they feel after taking a substance, may be a more quantitative way of assessing cannabis intoxication.