Alcoholism can be a difficult condition to diagnose, especially in cases where individuals’ drinking habits are not observed and physical symptoms have not yet appeared. In a new study published in Alcohol addiction: clinical and experimental researchCo-author Andrew Schwartz, of the Department of Computer Science at Stony Brook University, and colleagues determined that the language people use in Facebook posts could identify those at risk for risky drinking habits and alcohol use disorders.
Collaborating with Schwartz on the Data Science for Unhealthy Drinking project are Stony Brook University doctoral candidate Matthew Matero, and Rupa Jose, PhD, lead author and postdoctoral researcher at the University of Pennsylvania.
Key to the research was the use of Facebook content analyzed with Contextual Decorations, a new artificial intelligence application that interprets language in context. The contextual inclusion model, Schwartz, Goose and colleagues say, has a 75 percent chance of correctly identifying individuals as high or low risk drinkers from their Facebook posts. This rate of identifying people at risk for excessive alcohol intake is higher than other traditional models identifying high-risk drinkers of alcohol and those at risk of alcoholism.
“What people write on social media and on the Internet offers a window into psychological mechanisms that would otherwise be difficult to spot in research or medicine,” says Schwartz, commenting on the study’s unique aspect.
Our findings suggest that drinking is not only an individual driven behavior, but a contextual one; With social activities and group membership helping set the tone when it comes to encouraging or discouraging drinking.”
Rupa Jose, Ph.D., lead author and postdoctoral researcher, University of Pennsylvania
Investigators used data from more than 3,600 adults who were recruited online -; average age 43, mostly white -; who have agreed to share their Facebook data. Participants filled out surveys about demographics, their drinking behaviors, and their perceived stress—a risk factor for problematic alcohol use. The researchers then used a diagnostic scale to organize participants -; Based on self-reported alcohol use -; to high-risk drinkers (27 percent) and to low-risk drinkers (73 percent).
Facebook language and high-risk drink-related topics included more frequent references to going out and/or drinking (eg, “party”, “beer”), more expletives, more informality and slang (“lmao”), and more references Negative emotions (“miss,” “hate,” “lost,” and “hell.” These may reflect factors associated with high-risk drinking, including access to nearby bars, and personality traits such as impulsivity.
Low-risk drinking status was associated with religious language (“pray”, “Jesus”), references to relationships (“family”, “those who”), and future-oriented verbs (“will”, “hope”). These may reflect purposeful support networks that encourage moderation in drinking and the presence of future goals, both of which protect against dangerous drinking.
Overall, the authors conclude, “Social media data serves as a readily available, rich, and untapped resource for understanding important public health problems, including excessive alcohol use… (study findings support) the use of Facebook language to help identify potential alcohol-exposed populations. which need follow-up assessments or interventions, and note the multiple language tags describing individuals in the high/low alcohol exposure groups.”
Jose, b. et al. (2022) Using Facebook’s language to predict and describe excessive alcohol use. Alcohol addiction: clinical and experimental research. doi.org/10.1111/acer.14807.