BACKGROUND
Posts on social media that depict vaping in positive social situations serve to shape expectations about vaping, suggesting it is socially acceptable. TikTok, a social media platform popular among adolescents, has such content, despite restrictions on uploading depictions or promotions of controlled substances. There is a need to understand strategies employed in promoting vaping on TikTok, especially among susceptible youth audiences.
OBJECTIVE
This study seeks to describe direct and indirect themes promoting vaping on TikTok and to understand how such themes may influence youth perceptions of vaping to inform public health communication and regulatory policies regarding vaping endorsements on TikTok.
METHODS
We collected 14,002 unique posts from TikTok using 50 vape-related hashtags (i.e., #vapetok, #boxmod, etc.) and used K-means, an unsupervised machine-learning algorithm, to identify clusters. Following an in-depth examination of the posts in each cluster, we categorized posts into themes. Thereafter, we retrieved images from thematically organized video datasets. We assessed the performance of three machine-learning-based model architectures (i.e., ResNet50, VGG16, and ViT) when extracting and clustering visual features. We evaluated results and selected the best-performing model for the qualitative analysis. We examined the characteristics of 25% of the samples from each image cluster to determine the accuracy of classification. Finally, we selected 50 videos randomly from each of the five themes to compare with the theme descriptor.
RESULTS
We identified five major themes. One theme ‘vaping marketing’ (8.3%) reflected direct marketing. The other four themes reflected: 'general vape’ (19.6%), ‘TikTok influencer’ (27%), ‘vape brands’ (14.6%), and ‘vaping cessation’ (9.1%). The ResNet50 model successfully classified clusters based on image features (obtaining an F1 score of 0.97 on average, the highest of the three models). Content analyses indicated that vaping was depicted as a routine part of daily life and TikTok influencers subtly integrated vaping into popular culture (e.g., games) and social practices. Moreover, analyses revealed that the ‘vaping cessation’ theme had the lowest success of classification into the appropriate cluster, underscoring the relative low frequency of cessation-focused material on TikTok.
CONCLUSIONS
Results from both computational and qualitative methods on text and visual data indicate that vaping is normalized among youth on TikTok. The identified themes emphasize that various elements, such as everyday conversations, promotional content, and the influence from popular figures, work together to create an environment on TikTok where vaping is portrayed as a normal and accepted part of daily life. Furthermore, the high accuracy of our computational models, validated through qualitative content analysis, reinforces the reliability of our findings. This comprehensive examination provides valuable insights for regulatory policy and public health initiatives designed to address challenges posed by the normalization of vaping on social media platforms.