The Wild Wooly Internet

Grapevine of Worry

Lucy Kellaway wrote in the Financial Times a couple of days ago that her own "mild fearfulness" about the economy had ballooned to hyperventilating paranoia after she spent time surfing the web and opening e-mail.

"Through blogs, websites and e-mails the world's economic ills are fed to us on a drip all day long. It is not just that we hear about bad things faster, we hear about more of them and in a more immediate way. My worries become yours, and yours become mine."

Since I don't "sit over my computer all day and feed my anxiety", I disagree. I don't succumb to bad news, rather I cheer myself when Obama talks about limiting publicly financed executive pay, or when the head of the Bureau of Land Management puts a hold on the drilling leases near national parks auctioned off by the Bush administration. In dire moments, I distract myself with unicorn chasers and happy news. Don't you? I walk away from the computer at will. I turn it off.

Back To Math Class You Go

But lets move beyond my anecdotal evidence. Lucy Kellaway speaks, as always, slightly tongue in cheek, but other news stories might convince you that the internet truly does harbor inescapable and vile corruption that needs to be caged. Take for instance, the New York Times piece yesterday about MySpace and their campaign to purge registered sex predator names from their site. According to the NYT, MySpace turned over 90,000 names to Attorney General Richard Blumenthal of Connecticut and Attorney General Roy Cooper of North Carolina.

Officials are pressuring social networking sites to adopt more stringent safety standards to assure children's safety. This is a welcome but confusingly priority since a report by the Internet Safety Technical Task Force, commissioned by 49 state attorneys general found that bullying online was a far more serious problem than sexual solicitation. Nevertheless, Attorney General Blumenthal said in a recent statement:

"Almost 100,000 convicted sex offenders mixing with children on MySpace -- shown by our subpoena -- is absolutely appalling and totally unacceptable...for every one of them, there may be hundreds of others using false names and ages."

I'm all for blocking names. But lets sort through his math. 90,000 names, times "hundreds" of "others". We'll interpret his "hundreds" conservatively, let's say 300-- although perhaps the Attorney General meant 900. So 90,000 * 300 = 27,000,000 sex offenders on MySpace? Maybe up to 81,000,000? The population of the US is ~303,824,640. So on the conservative side, Blumenthal tells us that 1 in 10 US sex offender citizens trolls MySpace. YIKES!

The Times reports later in the story that there are 700,000 sex offenders in the US. The paper doesn't worry with the math discrepancy. Instead they quote John A. Phillips, "chief executive of Aristotle, a company that supplies identity and age verification technologies for companies like the New York State Lottery, breweries and film studios", who is trying to sell his software to Myspace, and so piles on: '"this is just the tip of the iceberg on MySpace".

So, fear for the little children. Fear for the investor class, homeowners, and retirement fund enrollees. Who else?

Fear For the Suggestible, the Unvaccinated

If 1/1000 to 4/1000 registered YouTube users rate vaccination videos with 1 to 5 stars, adding comments like, "your video is stupid, and your a dumbass that's what my mom thinks", should we use this "data" to propagate concern that YouTube feeds the public irrational and dangerous opinions about vaccinations? A year ago the Journal of the American Medical Association (JAMA) published a study that did exactly that.

In December 2007, University of Toronto researchers announced a "first-ever study of its kind". The investigators selected YouTube videos relating to immunizations or vaccinations, and concluded that much of the video content "contradicts the best scientific evidence". The public health community should find this "very concerning", they wrote. The press pounced on this announcement like a starved puppy tossed a Porterhouse steak. Articles titled "YouTube Full Of One-Sided Anti-Vaccination Videos", littered the news.

The authors selected and watched 153 videos. 73 (48%) had so called "positive" messages (in favor of immunization) , 49 videos (32%) had so called "negative" messages, and 31 videos (20%) had so called "ambiguous" messages. The study concluded:

  • 1) "negative videos were more likely to be rated by viewers"
  • 2) negative videos were more likely to "receive more views"
  • 3) negative videos "received a higher mean star rating".

The authors then generated their dire warnings.

Garbage In...

When JAMA published the study I spent some time looking at their data, which I'll briefly highlight here. Unfortunately, it was impossible to repeat the study. Obviously, time had elapsed between authors' video viewing and publication, but also the authors' described their methodology fleetingly: "On February 20, 2007, we searched YouTube ( using the keywords vaccination and immunization." Straight-forward and repeatable? Hardly. Different permutations of the keywords and Boolean operators yielded anywhere from 63 to 1300 videos, when we copied their methodology. This result may not seem important, but such unreliability prompted us to look at the validity of the study's conclusions.

The authors found that "negative videos were more likely to be rated by viewers." Of 73 "positive" videos, 46 had a rating. Out of 49 "negative" videos, 42 had a rating. But you have to wonder how meaningful a metric like "number of ratings" or "likely to rate" is. Looking at the raw data a different way, you would also learn that more people rate "positive" videos.

We multiplyied 73 "positive" videos by the study's "mean number of positive" video views - 181, which gives 13,213 views. Yet there were only 37 viewer ratings. Multiplying 49 "negative" videos by the mean, 520 views per video, gives 25,480 viewers -- but only only 36 ratings. So yes, there were "more views" of "negative" than "positive" videos, and more negative videos were rated. But also the data showed that an individual is more likely to rate a "positive" video (.28% of viewers rated), than the "negative" video (.14% of viewers rated). Why? Do more people watch "positive" videos to the end? Who knows.

There were other confounding questions unanswered by the study. How long had the videos been posted? Does rating a video actually signal a change in attitude? Behavior? Anything? How many people rate videos -- only registered users can rate videos, so do registered YouTube users represent the vaccinating public? Is a "negative" Gardasil video a "bad" public health message, given the uncertainty about the pros and cons of that vaccine? Moreover, can tabulating viewer ratings translate to anything meaningful? Especially when only ~1-2 in 1000 viewers rates a video?

...Garbage Out

The authors also tallied the YouTube star ratings and concluded that "negative" videos received higher ratings. But in a 1-5 star rating system such as YouTube's, what do we learn from reports that the mean "positive" video rating was 3.5, with 1.5 standard deviations (SD), whereas the mean "negative" video rating was 4.4 with .9 SD?

Does running statistics on shaky data make it more meaningful? According to analyses of 5-Star Rating Systems there are plenty of other problems with drawing many conclusions from ratings. Individual ratings tend to be either very low, or very high (1 or 5), in a bimodal distribution. Problematically, an average score of 3 or 4 might only describe "conflicting opinions". As it turns out, averaging most 5 star ratings gives a mean 3-4 star rating.

Another bias of 5 star rating systems is upward-bias from "fans". For instance, when we looked at available videos in December, 2007, in a video from "House MD", the TV program, a doctor very sarcastically scorns a woman for not vaccinating her child. This ("positive") video got rated very highly (4+ stars). But tans will rate a show highly no matter what the public health message. Problematically, then, the JAMA study uses these crude ratings to make some serious public health claims about the dangers of YouTube.

In December 2007 we did our own little mini-study on YouTube to confirm the JAMA data showing that only 1-5 of every 1000 viewers rated these videos (true). Ironically, at the time, the most popular YouTube video about vaccination was a "positive" one put out by a pharmaceutical company, which only showed up in some searches. This corporate video got almost 800,000 views, more than 10 times the 69,000 total views of the 153 videos the authors studied, far surpassing all the "negative" videos.

The pharmaceutical company was advertising a video contest for homemade videos about getting a flu shot with a $500 prize. The video got negative reviews, but some comments reflected people's annoyance that the contest had ended or they hadn't won. The House MD video was the second most popular video.

Barbarians on the Net

This idea that the internet will tear down society one way or another by undermining civility, by cultivating irrational fear, spreading disease, crime, or irrational behavior is not new, and in fact reflects various bricks and mortar versions of the same fear-mongering. See for instance, The Coming Anarchy, by Robert D. Kaplan and similar titles. In reality, nation-states quite adeptly control the internet, as they do their roadways, waterways, and airspace.

Despite the constant threat of unreasonableness and anarchy, it is reason that often trumps unreasonable cacophony on the internet, the opposite of what people predict. Would Obama have been elected without the internet? Would the Palin candidacy have met the same fate without TV's internet availability to the hordes who watched Couric and Fey?

The internet has its problems, but I suspect its vagaries offend most people when the internet disrupts the power assumptions they hold dear. One can find all the nastiness, the worry, the fear, and the bizarre opinions of the internet on the streets. In reality, predators pass kids everyday on the street, as anonymously as on the the net.

The internet provides only an illusion of anonymity for ne're-do-wells and oafs, just as your house with its fence and well surveyed lot and planted trees provides an illusion of safety to you. Do those in privileged positions avoid the awfulness of the cement ghetto more easily than they elude the unsettling and unwanted spam in their AM inbox, and thus be more offended by the internet?

Power brokers of course become threatened by the internet. Record companies, the networks, and politicians, and pharmaceutical companies -- they've all had run-ins with the internet. Professors object to "RateMyProfessor", as it mucks up the power structure. But it's certainly helps the public forum.

One needs to exam the data behind assertions that the internet is dangerous. Corporations have far more power on the internet than so called fringe groups -- to advertise, to astroturf, to datamine, and to collect personal information, although they may claim that's not so. The pro-vaccination video put out by the pharmaceutical company, even by the very dubious standards put forth in the JAMA study, was more "influential" then all the rest of the videos on YouTube combined.

Authors, consultants, the media, have always tried to pin down and characterize internet communication trends, but their calculations and predictions often miss because they are only a static snapshot of the evolving internet at a point in time. John Perry Barlow predicted the World Wide Web without a government (1996); consultants predicted internet "content was king" (1997); Cass Sunstein dreamed up regulatory schemes so that the polarizing internet wouldn't destroy democracy (2001); and print journalists talked about how doomed blogging was (2004). They misunderstood the adaptability of the internet as a communication tool and underestimated how individuals, corporations, and governments would continue to shape it to further their own personal wish lists. One day anti-vaccination videos seem prevalent, the next, pharmaceuticals have usurped YouTube just fine thank-you.

The film critic Robert Ebert is right, newspapers can be great to read, (all five of them) they also tend towards banal, narrow-minded, wrong, and biased, so we better get used to the excellent, disparate, positive as well as very negative flux of the internet.

follow us on twitter!