Document text
Principal Investigator: Dongmei Li
Organization: UNIVERSITY OF ROCHESTER
Fiscal Year: 2024
Award: $565,181
Funding agency: National Cancer Institute
PROJECT SUMMARY
The National Youth Tobacco Survey showed the prevalence of e-cigarette use among high school students has
skyrocketed from 12% in 2017 to 28% in 2019 and remains high at 14.1% in 2022. While different regulatory
policies (such as the FDA e-cigarette flavor enforcement and Tobacco 21) are currently in place trying to alleviate
this vaping epidemic among youth and young adults, there is an urgent need to effectively communicate with the
public about the risks of e-cigarette use and nicotine addiction. However, how to engage the public with the e-
cigarette prevention messages is particularly challenging. Social media platforms such as Twitter/X, Instagram,
TikTok, and YouTube are very popular in the United States, especially among youth and young adults. Previous
studies have found that social media platforms are widely used to promote e-cigarette products by vape shops
and companies but are under-used by public health authorities for educating the community about the health
risks of e-cigarette use. Social media marketing of e-cigarettes as healthier alternatives to conventional
cigarettes resulted in the common perception among youth that vaping is a harmless activity. Social media
exposure to e-cigarette-related content could affect e-cigarette use (vaping) behavior, with vaping promotion
content leading to more vaping and vaping prevention content leading to reduced vaping. Therefore, social media
provides a rich and natural data resource about vaping-related messages in different formats (text, image, and
video). The purpose of the proposed study is to identify key features of e-cigarette-related social media
(Twitter/X, Instagram, TikTok, and YouTube) posts (especially vaping prevention posts) associated with high
social media user engagement (such as the number of likes) by applying advanced artificial intelligence and
statistical modeling techniques, and further validate them using the combination of an online survey study and a
semi-structured interview study. To achieve our goal, we will characterize key features of Twitter/X posts (tweets)
related to e-cigarettes associated with high social media user engagement, such as the number of retweets and
favorites, using natural language processing, machine learning, and statistical modeling techniques (Aim 1). We
will characterize key features of e-cigarette-related Instagram posts associated with high social media user
engagement through deep-learning algorithms and statistical models (Aim 2). We will characterize key features
of e-cigarette-related TikTok and YouTube videos with high social media user engagement using artificial
intelligence techniques (Aim 3). We will validate the key features selected from social media posts through a
cross-sectional anonymous online survey study and a semi-structured interview study (Aim 4). Results from the
proposed study will provide valuable guidance in designing effective vaping prevention messages for future
public health campaigns to help with the effective communication of risks associated with e-cigarette use, which
will help prevent the initiation and counter-uptake of e-cigarettes by youth and young adults.
Terms: <AI language models><AI system><Affect><Algorithms><Artificial Intelligence><Attitude><Behavior><Belief><Breathing><Categories><Comment><Commentary><Communication><Communities><Computer Reasoning><Data><Data Sources><Disparity population><Editorial Comment><Electronic cigarette><Epidemic><Ethnic Origin><Ethnicity><Exposure to><Food and Drug Administration><Future><Geography><Goals><Health><Health Campaign><High School Student><Image><Industry><Instagram><Interview><Junior High School Student><Junior High Student><Language><Machine Intelligence><Machine Learning><Marketing><Measures><Methodology><Methods><Middle School Student><Modeling><Multimedia><Multimedium><Natural Language Processing><Nicotine Dependence><Participant><Perception><Persuasion><Persuasive Communication><Policies><Population><Population Heterogeneity><Prevalence><Prevention><Probabilistic Models><Probability Models><Public Health><Published Comment><Race><Races><Reporting><Research><Respiratory Aspiration><Respiratory Inspiration><Risk><Sales><Secondary School Student><Secondary Student><Smoking><Statistical Models><Structure><Survey Instrument><Surveys><Techniques><Text><TikTok><Time><Tobacco><Tobacco 21 bills><Tobacco 21 laws><Tobacco 21 legislation><Tobacco 21 policies><Tobacco Consumption><Tobacco use><Training><Transformer language model><Twitter><USFDA><United States><United States Food and Drug Administration><Validation><Variant><Variation><Viewpoint><Youth><Youth 10-21><addiction><addictive disorder><adult youth><aged><artificial intelligence language models><authority><biopsychosocial><cease smoking><combustible cigarette><community advisory board><community advisory committee><community advisory panel><conventional cigarette><data resource><deep learning algorithm><deep learning based model><deep learning model><design><designing><disadvantaged group><disadvantaged individual><disadvantaged people><disadvantaged population><disadvantaged subgroup><disparities across groups><disparity across subgroups><disparity among groups><disparity among subgroups><disparity between groups><disparity between subgroups><diverse populations><e-cig><e-cig use><e-cig user><e-cigarette><e-cigarette use><e-cigarette user><ecig><ecig use><ecig user><ecigarette><ecigarette use><ecigarette user><electronic cigarette use><electronic cigarette user><experience><feature selection><group disparity><group inequality><group inequity><heterogeneous population><high schoolers><imaging><inequalities among populations><inequalities between populations><inequalities in populations><inequality across populations><inequality among groups><inequality between groups><inequality in groups><inequities among populations><inequities between populations><inequities in populations><inequity across groups><inequity across populations><inequity between groups><inequity in groups><innovate><innovation><innovative><inspiration><large language model><large scale language model><machine based learning><massive scale language models><natural language understanding><nicotine addiction><nicotine dependent><novel><population diversity><population inequality><population inequity><prevent><preventing><quit smoking><racial><racial background><racial origin><smoking cessation><social media><statistical linear mixed models><statistical linear models><stop smoking><subgroup disparity><tobacco product use><tobacco products><traditional cigarette><unequal group><unequal population><uptake><validations><vaping><young adult><young adulthood>