Monitoring the drinking habits of teenagers is tricky - kids tend not to be totally honest when surveyed about their alcohol use, and those that are aren't normally a representative sample.
So researchers from the University of Rochester in the United States have turned instead to Instagram. They say it can expose data more cheaply and faster than conventional surveys, and also discover what alcohol brands and types are favoured by different demographic groups.
Jiebo Luo and his team wrote in a paper that underage drinkers "are willing to share their alcohol consumption experience" more readily on social media, and that monitoring Instagram can also allow researchers to observe it in an "undisturbed state".
While Instagram doesn't allow users to be selected by age, the team was able to use computer vision technology to obtain "sufficiently accurate" guesses for age, gender and race. Then a constructed slang dictionary and the names of alcohol brands were used to collect the necessary data.
Weekends and Holidays
The results showed that underage alcohol consumption tends to happen more on weekends and holidays, and at the end of the day. There was no gender bias in alcohol consumption - it matched the ratio of Instagram users. The researchers did, however, identify brands that tend to attract younger audiences - information that could be useful to people working with teenage drinkers.
"We can keep government agencies or schools better informed and help them design interventions," said Luo. "We could also use social media to incorporate targeted intervention and to measure the effect of any intervention. And perhaps other things we haven't thought about."
Elizabeth Handley, clinical psychologist and research associate at the University's Mount Hope Family Center, added: "This new method could be a useful complement to more traditional methods of measuring youth drinking. It could provide important new insights into the contexts of youth drinking and be a valuable tool for evaluating the effectiveness of school or community-based preventive interventions."
The team's work will be presented at the 2015 IEEE International Conference on Big Data in Santa Clara this week.
