Social media platforms provide a unique indicator of activity in the rapidly changing market for new psychoactive substances (NPS). On the popular Reddit website, drug-related discussion occurs in user-generated subreddits dedicated to a drug, a class of drugs, or more general topics related to drug use or experimentation. The contents of drug subreddit discussions can be useful for estimating temporal trends in NPS use, including early and real-time identification of emerging drugs.
In collaboration with the Machine Perception and Cognitive Robotics (MPCR) Lab, under the direction of Dr. Elan Barenholtz and PhD candidate Paul Morris, the NDEWS Coordinating Center has developed a web monitoring platform for early detection of NPS in drug subreddits. Trends in drug discussion are quantified through anonymized, aggregate keyword counts derived from algorithmic monitoring of ~80 drug-oriented subreddits. Keyword metrics count mentions over time of keywords that refer to a drug. Machine learning models are employed for automated detection and aggregation of drug keywords.
In combination, these methods detect drug-related activity that is anomalous and potentially indicative of an emerging trend in the development or use of novel substances. Validation on historical trends reveals that detection of an NPS in drug subreddit discussion is predictive of its subsequent emergence in toxicology data and other real-world signals. Early detection of NPS trends by web monitoring serves as a source for further investigation and collaboration with NDEWS partners.