Kellen Sharp

Big Tobacco and other stakeholders, such as vape shops and smaller e-cigarette manufacturers, have adapted traditional tobacco marketing techniques to digital platforms. Warning labels are essential for informing consumers about the potential harms of tobacco use, including e-cigarettes. However, in a rapidly changing digital landscape, social media platform policies often lag behind, leaving digital marketing largely unchecked. This has allowed Big Tobacco to modernize traditional cigarette marketing in the digital sphere with e-cigarettes, a phenomenon we term ‘remediated marketing’. Without adequate warning labels, exposure to tobacco promotion may increase e-cigarette use among youth, who engage with social media at particularly high rates. This article presents a rule-based classifier developed to detect warning labels in TikTok and YouTube videos by combining computer vision technology with rule-based classification. Our classifier achieved 97.33% accuracy in detecting posts with warning labels. However, only 2.32% of YouTube video frames (240 out of 10,344 frames) and 1.32% of TikTok video frames (61 out of 4639 frames) contained warning labels, suggesting that warning messages are infrequent across e-cigarette content on platforms popular among youth, including TikTok and YouTube. Among the detected warning labels, there was notable diversity in wording and length, indicating a lack of standardization. Additionally, within YouTube and TikTok video frames, 63.7% and 30.0% of the warnings appeared in the first five seconds of the videos, respectively. These results highlight the need for improved policies and standardized warning labels to better protect young adults from e-cigarette promotion on social media.

This study investigates the potential for integrating AI and journalistic networks to create real-time, priority-driven maps of infrastructure damage during natural disasters. Focusing on Hurricane Florence in 2018, we collected over a million tweets using the REST Twitter API and extracted 11,638 images for analysis. Tweets were categorized by source, including news organizations and citizen journalists. We applied the OpenAI CLIP unsupervised machine learning model for image classification, splitting the data into 80 % for training, 10 % for validation, and 10% for testing. The model achieved an average precision of 92 %, recall of 78 %, and an F1 score of 85 %. When compared to other models such as ViT and DeiT, which achieved F1 scores of 82.9 and 81.2, respectively, CLIP performed similarly but stood out due to its accessibility and zero-shot learning capabilities, making it ideal for rapid deployment in newsrooms and crisis scenarios. The framework's success was further demonstrated by cross-referencing model predictions with geotagged metadata and journalist sources, which linked damage locations with credible information. By leveraging this AI-based framework, journalists can significantly reduce the time needed to identify disaster-response targets, helping to focus relief and recovery efforts in real time. This approach enhances disaster data collection, analysis, and dissemination, ultimately saving lives and reducing harm by providing more efficient and accurate damage assessments. The study highlights how AI and journalistic networks can collaborate to improve crisis response efforts.

Mpox (formerly monkeypox) is a zoonotic disease caused by an orthopoxvirus closely related to variola and remains a significant global public health concern. During outbreaks, social media platforms like X (formerly Twitter) can both inform and misinform the public, complicating efforts to convey accurate health information. To support local response efforts, we developed a researcher-focused dashboard for use by public health stakeholders and the public that enables searching and visualizing mpox-related tweets through an interactive interface. Following the CDC's designation of mpox as an emerging virus in August 2024, our dashboard recorded a marked increase in tweet volume compared to 2023, illustrating the rapid spread of health discourse across digital platforms. These findings underscore the continued need for real-time social media monitoring tools to support public health communication and track evolving sentiment and misinformation trends at the local level.

Social media research is confronted by the expansive and constantly evolving nature of social media data. Hashtags and keywords are frequently used to identify content related to a specific topic, but these search strategies often result in large numbers of irrelevant results. Therefore, methods are needed to quickly screen social media content based on a specific research question. The primary objective of this article is to present generative artificial intelligence (AI; eg., ChatGPT) and machine learning methods to screen content from social media platforms. As a proof of concept, we apply these methods to identify TikTok content related to e-cigarette use during pregnancy.