Smoking Trends on Twitter: Pot or Porn?
Topic and research question
It is common sense that smoking is a serious public health problem worldwide. It is considered to be the largest single cause of preventable death and disease in the world (WHO Report on the Global Tobacco Epidemic, 2008). The more research is conducted, the more is known about the risks of smoking, hence we are certain in these days that smoking causes serious harms to peopleâs health and can even lead to death. Research has been conducted as long as scientist and public health experts have known about the risks of smoking. Moreover it seems that the interest in researching smoking risks (Catsburg et al., 2014; Pimhanam et al., 2013) and/or peopleâs smoking behavior (Noorzurani et al., 2013) is important for todayâs society. It comes as no surprise that scientists and researchers started engaging themselves with the questions as to why people start smoking in the first place. Thus many studies are concerned with the perception of smoking and tobacco in society and how the image of smoking is being shaped and influenced.
In this context the role of the media has been questioned many times. Especially when it comes to the problem of smoking among teenagers the film industry and mass media in general are under high scrutiny. Since the 1990s research provides a better understanding of the proliferation of the image of smoking in popular media. Young people indeed associated certain meanings and values to the representation of smoking in media (McCool et. al, 2005). Especially with the rise of the movie star in the 1930s and 1940s, showing images of leading male as well as female actors smoking in films, the image of tobacco has been glamorized (McCool et. al, 2003). Â With a growing number of studies examining the image of tobacco use in film and the media, more evidence has been found that these images endorse the motivation, especially among younger people, to start smoking. The media not only plays a significant role in influencing the behavior of teenagers and generally speaking the youth culture, different kinds of mass media (film, magazine, tv) are attributing favorable characteristics to the usage of tobacco (Watson et. al, 2003). Thus there is a growing concern of the media influencing young peopleâs attitude toward smoking.
With the rise of social media this concern becomes even more crucial, as many people depict themselves or others on pictures while smoking and disseminating them via the various social networks. It is argued whether those pictures have potential to influence others who see these pictures. A recent study conducted earlier this year investigates how friendship with peers who smoke and showing their substance use online could influence the teensâ own attitude towards tobacco (bitscience.com, 2014).
          However research on the representation of tobacco use and smoking in social media networks has been few and clearly shows a research gap within this field of study. Therefore this study investigates the social media platform Twitter and how smoking and the sentiment towards smoking can be identified here. The research project hereby follows the overall research question: What is the image of smoking on Twitter? We developed subquestions for the logical analysis:
What devices are used to write tweets contained #smoking?
What hashtags are connected with #smoking  on Twitter?
What is the sentiment of the tweets and retweets with the hashtag #smoking and hashtags related to it?
Methodology
In order to answer our research question, certain DRM methods were applied. First, to show the real amount of tweets connected with #smoking several online tools were used. Those allowed this research to explore findings in real time and get a more complete picture of the image of smoking. Towards the end, Gephi was used for analyzing the connections between hashtags and tweets.
Data
To have a better understanding of how smoking is being perceived on Twitter, two different datasets were used. Accordingly, the first dataset is analysed through the following tools:
For the online real time analysis made with Tweet Binder, we used 1981 tweets from October 2014, retrieved on 28 October.
For the online real time analysis made with Sentiment Viz, 419 tweets from October 2014 were retrieved on 28 October 2014.
The second dataset is analysed from the perspective of Gephi.
Hashtag_edges file, which consisted of 3676 edges, was used.
The file hashtag_edges as well as the file retweet_edges (242 edges) and smokingEnLIWC file were used for the LIWC analysis. Â
Methods and results
Tweet Binder
The report for #smoking shows significant results. Overall, 1,981 tweets that contain this hashtag have been retrieved. It is interesting to notice that 6,43% of them are original tweets (120), while 35,99% are retweets (717). Thus the number of retweets is larger and there is a certain interaction between users, since they are willing to retweet the othersâ tweets. In addition, the analysis shows that 32 of all tweets were replies. Most tweets contains additional links and pics (55,97% of all tweets, 1115 tweets). As these additional pictures and links are another type of content than text, it testifies the interest to the topic among users.
Most of the contributors (1006) tweeted once. The number of contributors who tweeted three times was smaller than the number that tweeted six times or more. It can be assumed that there may be a person or organization interested in #smoking, who posted more than 6 tweets as it can be considered as a daily (or with a certain purpose) rather than random contribution.
Among most active contributors, there are some twitter accounts connected with porn and videos for adults, for instance, @C4SUpdates, which is â#1 Downloadable Clip Site on the Netâ (https://twitter.com/@C4SUpdates, 2014). This user is also among those who has the greatest impact.
As for the top 10 sources that were used to write and retweet tweets contained #smoking, the most popular one was Twitter for Web Client. The second one was Twitter for iPhone, the third one - IFTTT. The difference between the number of tweets from each client was not significant, though: 287 for Twitter for Web Client, 235 for Twitter for iPhone and 221 for IFTTT. Interestingly we noticed Instagram among the clients (5th place with 168 tweets).
Sentiment Viz
As for the hashtags connected with #smoking, those connected to porn (#porn, #latex, #xxx) are marked as positive, while #tobacco, #cancer, #nicotine are marked as negative. Interestingly, #quit was among the positive ones. This can be interpreted as the positive emotions people feel while quitting smoking, which they share with their followers.
Hashtags connected with porn (#fetish) and pot (#420) are also connected with #smoking.
As for the sentiment of the tweets, they are mostly positive. It could be assumed that it is either because of the positive emotions people feel about smoking or because the tweets are connected with porn (and users are positive about it).Â
Gephi
Gephi was used for further examination. When examining the hashtag_edges data file in Gephi and conducting indegree and Force Atlas upon the twitter data, the following Gephi visualisation appeared.
The biggest cluster (the second one) in the Gephi visualisation represents #smoking. It can be observed that the smaller clusters are connected to the #smoking. Moreover, a lot of terms are related to this tweet. Some of the most popular ones are #nicotine, #longcancer, #LetsSmokeLoud and #ShakeTheHabit. However, most interesting of all, it is the cluster of tweets forming at the right of the cluster of #smoking. These are tweets that are related to #Smoking (with capital letter S). The findings of these tweet related messages are interesting since half of the messages are also related to #Nicotine, #Addiction and #KidsHealth, #Electroniccigs topics, which are smoking-related topics. However, the other cluster that is related to #Smoking represents porn related messages.
Examples of related tweets are #porn, #Sextweets, #Nudescene and #Slut. Therefore, an interesting perception and maybe a noticeable trend is that #Smoking is not only related to subjects concerned with smoking, but also to Porn and Sex.
We sorted the data in Excel and selected all tweets with the hashtags âsexâ,âpornâ etc. and then marked them as âpornâ. As it was done manually, it could be considered as a difficulty. Manual sorting of hashtags can lead to the error in identification, for example,#sex can appear in the non spam porn related tweet, although the probability of this error is quite small and it could be neglected; nevertheless, this can be considered a limitation. While red lines are for tweets connected with porn, green lines are for tweets which do not contain any kind of porn. It is interesting to notice, that majority of the hashtags related to the pornographic content are linked together, which means that they are used together frequently. Also #smoking with all letters in lowercase has less connection to such hashtags, meaning that porn related tweets mostly use all caps or first letter caps hashtags. It can be the case that some of the hashtag which were tagged as porn hashtags were used in regular tweets, but chances of that are rather small, and we see that such hashtags are rarely connected to the normal ones. When we looked at the tweets manually, we noticed that there could be some hashtags which were not necessarily related to porn, like #naked.
LIWC data (smokingEnLIWC) was added to hastag_edges file. Blue is for none, green is for positive, and red is for negative. It is clear, that there were a lot of tweets marked as ânoneâ.
It is interesting that tweets contained hashtags connected with porn are marked either ânoneâ or âpositiveâ in terms of emotions.
Hashtag â#420â related to pot is also noticeable and is marked as âpositiveâ.
As for analysing the retweets, Fruchterman-Reingold algorithm was applied to retweets_edges file. Red colour is for tweets connected with porn while green is for tweets about everything else. There is only one big cluster of retweets connected with porn.
The emotions for retweets were analyzed using the same algorithm: red for negative emotions, green for positive emotions and blue for neutral or none. As it can be seen, the blue retweets dominate.
Conclusion
In order to answer the main question of this research (What is the image of smoking on Twitter?), these findings show that the used database has two main points of focus when using a hashtag related to smoking: (a) health concerns and smoking in general and (b) pornographic content. Moreover, an extended range of smoking-related hashtags were used. As previously seen visualizations show, it can be assumed that #smoking is primarily related to actual smoking and health issues concerning smoking. It is interesting to note that #Smoking is not only related to actual smoking but also to porn and sexual themes. Another interesting finding concerns #SMOKING, which proves to be mostly spam or related to pornographic content. Furthermore, this research shows that when users use a smoking-related hashtag, they do not necessarily refer to cigarettes but also to cannabis and pot consumption. Tweets connected with cannabis and pot are positive, tweets connected with porn are marked either as positive, or as none in terms of emotions. Most of retweets and tweets were marked as ânoneâ or âpositiveâ in terms of emotions. These findings can be useful for future research about the image of smoking on social networks.
P.s. Visualizations can be found here.
References
Catsburg C., Kirsh V.A., Soskolne C.L., Kreiger N., & Rohan T.E. (2014). Active cigarette smoking and the risk of breast cancer: a cohort study. Cancer Epidemiol, 38(4), 376-381. Â doi: 10.1016/j.canep.2014.05.007.
McCool, J. P., Cameron, L.D. & Petrie, K. J. (2005). The influence of smoking imagery on the smoking intentions of young people: Testing a media interpretation model. Journal of Adolescence Health. 36, 475-485.
McCool, J. P., Cameron, L.D. & Petrie, K. J. (2003). Interpretations of smoking in film by older teenagers. Social Science and Medicine. 56. 1023-1032.
Noorzurani, R., Bond, A., Wolff, K. (2013). A comparison of smoking behaviour characteristics between Caucasian smokers in United Kingdom and Malay smokers in Malaysia, Preventive Medicine 2013, doi: 10.1016/j.ypmed.2013.04.010.
No author. (2014) How Party Pics on Social Media Stimulate Teens Smokinng and Drinking Behaviour, Retrieved October 24th 2014 from: http://www.bitescience.com/Knowledgedatabase/Article/tabid/81/ArticleID/251/ArticleName/HowPartyPicsonSocialMediaStimulateTeensSmokingandDrinkingBehavior/Default.aspx
Pimhanam, C., Sangrajrang, S., Ekpanyaskul, C. (2014). Tobacco Smoke Exposure and Breast Cancer Risk in Thai Urban Females. Asian Pacific Journal of Cancer Prevention, 15 (17), 7407-7411.
Sentiment Viz. Retrieved from: http://www.csc.ncsu.edu/faculty/healey/tweet_viz/tweet_app/
Tweet Binder. Retrieved from: http://www.tweetbinder.com/
Watson, N.A., Clarkson, J.P., Donovan, R.J. & Glies-Corti, B.(2003). Filthy or fashionable? Young peopleâs perception of smoking in the media. Health Education Research. Vol. 18, no. 5, 554-567.
WHO report on the global tobacco epidemic, 2008. Geneva, World Health. Organization, 2008.











