Tuesday, July 16, 2019 1:00 am
A cloud of smoke
Study of vaping-related tweets leads to some troubling realizations
Occasionally, I get a feeling I have no idea why some social media exist. I felt that way about Twitter after reading a crazy study last week.
Several scientists and a doctor from the University of Southern California studied Twitter posts (ahem, tweets) about e-cigarettes with the goal of improving public health. Although the scientists didn't make a big deal about it, they uncovered something I consider shocking.
The scientists analyzed a sample of e-cigarette-related tweets from early 2017, about 6 million in all. It is apparently common for many tweets to be automatically generated by fake accounts. These accounts are called bots and have false photos, claimed interests and everything else to imply a real person. The authors used a somewhat common program, called BotOrNot, that identifies bot accounts. The program does this by studying patterns in how each account retweets, its ratio of followers to followees and patterns of nouns and verbs in tweets.
The authors identified 4 million of the 6 million tweets as based in bot accounts.
They then analyzed all the tweets using their hashtags. For example, if someone tweeted the following (this is a real tweet, though not from the study):
“Despite overwhelming scientific consensus that #vaping is #harmless and prevents youth #smoking, a handful of anti-#nicotine zealots manipulate governmental #tobacco tax addiction into policies that guarantee our children #KeepSmokingWeNeedTheMoney”
The words after each hashtag, in this case “vaping,” “harmless,” “smoking,” “nicotine,” “tobacco” and “KeepSmokingWeNeedTheMoney,” would be associated. Based on all the tweets, they formed clusters of hashtags that commonly occurred together in posts.
By the way, the tweet above is a lie. Basically none of it is true.
The cluster analysis for real people's tweets were unsurprising. Common tags were “vaping,” “vaplife” and “eliquid.” Considering hashtags as pairs, most tweets would have a mix of hashtags, making any pair of hashtags reasonably likely.
The cluster analysis for bots' tweets, on the other hand, indicated a more simplistic set of connections. Certain pairs of hashtags were much more common in the bots' posts compared to people's posts.
From the bot accounts, one of the four clusters was dominated by new product tags, such as #starterkit and #modbox. Another cluster was dominated by health topics and other substance uses, such as #quitsmoking, #health and #marijuana.
The authors speculate that these bots are being run or funded by the e-cigarette industry. To be honest, it would be bizarre and shocking if it were anything else. Some people created a series of automatic posts to Twitter encouraging e-cigarettes by using deceptive science and promoting specific products. Of course those people are likely to be from the industry.
A quick note about smoking, e-cigarettes and health. Smoking is terrible for a person's health. One of the most important choices each of us makes about health, after diet and exercise, is whether or not to smoke regularly. E-cigarettes are not tobacco. They are typically a mix of nicotine and various flavors made into a liquid, which is heated into a gas and inhaled. We do not know the long-term health effects of e-cigarettes. The chemical mix of the liquid is more varied than traditional tobacco cigarettes, so it is harder to study. It has also not been common nearly as long as tobacco. Because of the mix of chemicals that are common in the liquids, however, many health professionals are deeply suspicious that they are unhealthy.
For those people wanting to quit smoking tobacco cigarettes, there is no reliable method. Nicotine is famously addictive. In studies where people use a mix of behavioral therapy, nicotine replacement products such as a patch or gum and other medications, the quit rates are 20% to 25% over six to 12 months. Said another way, about 75% to 80% of people are still smoking after about a year.
There was a study in the New England Journal of Medicine that found e-cigarettes led to a quit rate of 18% after one year. That is not terrible, but not clearly better than other available methods. More importantly, behavioral therapy won't kill you. E-cigarettes might.
Considering all this information, then, are e-cigarettes a good method to quit smoking? Probably not. Describing the idea as terrible science, however, is probably excessive.
Reading this study, I felt like I had entered another world. Who in their right mind would ever look to random posts on Twitter for health advice? Why would someone trust it? Well, apparently enough do to encourage the industry to create fake accounts to push science deceptively.
More widely, what is going on here? A new industry is sprouting up and basically lying about a public health issue. I thought we, as a community, had decided that was not acceptable. Pharmaceutical companies cannot lie about their products. Tobacco companies cannot lie about their products (anymore).
These are American-based companies, pushing false advertising on an American-based media platform that could easily lead to poor public health. This is not OK.
Christer Watson, of Fort Wayne, is a professor of physics at Manchester University. Opinions expressed are his own. He wrote this column for The Journal Gazette, where his columns normally appear the first and third Tuesday of each month.