AI Learns to Identify MDMA Intoxication Through Speech
A team of neuroscientists, artificial intelligence experts, and psychologists from IBM has developed a new method to determine whether someone is under the influence of MDMA—simply by analyzing their speech patterns. In a study published in the journal Neuropsychopharmacology, researchers were able to identify MDMA intoxication with nearly 90% accuracy based on the types of words and emotions expressed in short speech segments. This detection method can distinguish MDMA use from natural oxytocin effects, such as those experienced when someone is in love. While MDMA produces similar effects, they are still significantly different from those caused by oxytocin.
The study involved 31 participants: 12 women and 19 men. Each participant completed two separate speech tasks four times each, for a total of eight tests per person. All subjects received either a placebo, two different doses of MDMA (0.75 mg/kg and 1.5 mg/kg), or oxytocin (20 IU), allowing researchers to identify differences in speech patterns among the groups.
The experimental procedure was conducted in a double-blind, randomized manner, meaning neither the participants nor the researchers knew which substance was administered during each test.
How the Speech Analysis Worked
The first speech task required participants to describe someone very important to them for five minutes. In the second task, participants were asked to speak freely about anything they wanted, for as long or as little as they wished, while alone for five minutes. All speech was recorded for machine learning analysis. Subjects who took MDMA displayed noticeably different speech patterns compared to those who received only oxytocin or a placebo, with the differences being more pronounced at higher MDMA doses.
Generally, people under the influence of ecstasy used more words related to intimacy, understanding, and emotions. Their speech also showed more signs of nervousness, different vowel pronunciations, and a broader use of adjectives and nouns.
Implications for the Future
This research suggests that, in the foreseeable future, doctors may be able to determine if you have taken MDMA—or possibly other drugs such as marijuana, alcohol, meth, cocaine, or heroin—simply by recording your conversation and analyzing it with a specialized device.
If doctors can do this, law enforcement could as well, which raises concerns about personal freedom. Currently, Australian authorities are testing infrared cameras at festivals to detect people on MDMA based solely on body heat levels—no breathalyzers, blood draws, or urine or saliva samples required.