4 ms·
You know what happened "shortly before" Snowden's leaks, two terrorists killed 3 people and injured hundreds others in a terrorist attack in Boston. Is that a
by slg 11y ago
You know what happened "shortly before" Snowden's leaks, two terrorists killed 3 people and injured hundreds others in a terrorist attack in Boston. Is that a good time to establish a baseline for number of page views?
I'm not saying there isn't a chilling effect, just that I am highly skeptical of this measurement of it. Wikipedia pages on these subjects are practically news articles. They are going to be pretty highly linked with whatever is going on in the world at the time. Considering that, I would fully expect terrorism searches to be high after a recent terrorist attack and to drop as terms like "Al Queda" become less newsworthy on a global scale.
- garrettgrimsley 11y agoCTRL + F "shortly before" 0 results. Putting quotes around shortly before makes it seem like you are drawing from the linked story, but you aren't. A quote that is actually from the article: >In the 16 months prior to the first major Snowden stories in June 2013, the articles drew a variable but an increasing audience, with a low point of about 2.2 million per month rising to 3.0 million just before disclosures of the NSA's Internet spying programs. Views of the sensitive pages rapidly fell back to 2.2 million a month in the next two months and later dipped under 2.0 million before stabilizing below 2.5 million 14 months later, Penney found. They did not use the day or week of the Boston Bombing as their baseline. Don't misrepresent the research.
- slg 11y agoYou are right on the quoting thing. I could have sworn it said "shortly before" but it said "just before". My mistake. Although those are almost semantically identical so I am a little confused on why you are calling it out in such specificity. >They did not use the day or week of the Boston Bombing as their baseline. Don't misrepresent the research. Snowden's leaks came out less than two months after the largest terrorist attack on US soil since 9/11. Since they are measuring monthly stats, they are using either the month of or the month after the Boston bombings as part of their "just before" time frame. If you look at all those number, the 3 million is really the biggest outlier. Isn't it possible that views for these terrorist related pages generally falls between 2 and 2.5 million with 20-50% jumps in page views in the aftermath of large terrorist attacks?
- garrettgrimsley 11y agoBecause the "just before" is referring to the overall trend, not the single month before the Snowden leaks. The actual highest single month view count is found in November 2012 with around 4 million views. The months around the Boston Bombing are not outliers. November 2012, as well as July 2014, are outliers. Eliminating these outliers from the analysis strengthens the case for the existence of a chilling effect. See pages 28-39 of the paper [0] for their methodology and findings. If I'm being unduly short it is because the sort of low-effort skepticism [1] that you offered is all too common with any article involving a statistical analysis. [0] http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2769645 http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2769645 [1] "Did the researchers account for {{obvious factor}}?" Yes, they did. Check before commenting.
- notahacker 11y agoI'd be interested to see the rest of their results, because I think you've just demolished that headline figure. Reading the article I noted that an audience varying between 2.2 million to 3 million in the months before and 2.2 million to 2.5 million in the months after didn't look like a step change in user behaviour; the claimed "fall" was no bigger than the variance over the previous 12 months. And I think you've just explained why the monthly results peaked at 3 million in one of the months before. I don't think the hypothesis is altogether implausible, but the evidence presented in the news article summary is less than compelling, particularly when this is the sort of research where researchers are at liberty to cherry pick their results. In the absence of the actual paper, here's the underlying Google trends https://www.google.co.uk/trends/explore#q=al%20qaeda https://www.google.co.uk/trends/explore#q=al%20qaeda https://www.google.co.uk/trends/explore#q=terrorism https://www.google.co.uk/trends/explore#q=terrorism https://www.google.co.uk/trends/explore#q=bombing https://www.google.co.uk/trends/explore#q=bombing Clearly (i) the hypothesised step change in user behaviour post-Snowden doesn't exist for global web searches (ii) there are a lot of spikes attributable to individual events, not least one in early 2013 and (iii) there's a trend decline in searches for terrorism starting over a decade ago
- garrettgrimsley 11y ago>I'd be interested to see the rest of their results http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2769645 http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2769645 I can email a copy if that one goes down.