When teens and young adults see posts promoting tobacco on social media, it puts them at higher risk. They’re more likely to start using tobacco products like e-cigarettes, and they might not see them as harmful. Even though platforms like TikTok and Meta ban tobacco promotion, not all of it gets removed.
Julia Vassey, a health behavior researcher, and Chris J. Kennedy from Harvard Medical School conducted a study. They used computer vision, a type of AI, to track tobacco-related items on social media. The study found a significant increase in such content—up to 100% from 2019 to 2022.Â
Their findings, published in the journal Nicotine & Tobacco Research, focused on influencer posts on TikTok, making it the first study of its kind. The research is part of the Keck School of Medicine’s Tobacco Center on Regulatory Science (TCORS), contributing to federal policy on tobacco.
Vassey and her colleagues from USC, Harvard Medical School, and Dartmouth College created a computer vision model using AI. This model identifies specific features in visual data like photos or videos. They trained the algorithm with a dataset of 6,999 images from Instagram, teaching it to recognize objects related to e-cigarette use.
The model can distinguish between eight categories of objects:
- mod or pod devices
- e-juice containers
- packaging boxes
- nicotine warning labels
- e-juice flavors
- e-cigarette brand names
- smoke clouds
By training the model this way, the researchers can pinpoint the specific types of e-cigarette products being promoted.
After training the model to spot tobacco-related items, Vassey and her team used it to study 14,072 TikTok videos from “micro-influencers.” These users have 1,000 to 100,000 followers who get many likes and comments.
The researchers discovered that pod device appearances grew by 33% from 2021 to 2022. The mentions of e-juice flavors and e-cigarette brand names doubled from 2019 to 2022. Nicotine warning labels also increased, appearing in 3% of videos in 2019 and 9% in 2022.
The study found important information and suggested a better way to study tobacco content on social media. While humans are better at spotting tobacco-related stuff, AI can handle more data. Even if AI is less accurate, a more extensive dataset or an improved model can compensate for it. This makes computer vision valuable for going through the ever-growing social media content.
The researchers have shared their model, code, and training data with students and researchers. Others can adjust the model to recognize different tobacco products like cigarettes or cigars.
The researchers from TCORS got more federal funding for a third round. They’ll keep studying how e-cigarettes affect young people, including the role of social media content.
Vassey said, “We know this content is growing, not disappearing. That tells us there’s a need for continued research on the presence and the effects of this marketing.”
In conclusion, the study highlights a worrying increase in tobacco promotion on social media, especially on platforms like TikTok, despite regulatory efforts. It emphasizes the importance of advanced AI, precisely computer vision, in monitoring and understanding social media dynamics. These findings have implications for public health and youth interventions, calling for more research and innovative approaches to tackle the growing impact of tobacco promotion online.
Journal reference:
- Julia Vassey, Chris J Kennedy et al., Scalable Surveillance of E-Cigarette Products on Instagram and TikTok Using Computer Vision. Nicotine & Tobacco Research. DOI: 10.1093/ntr/ntad224.