I'm sitting here at Bobby's NSA conference at the University of Texas. I just listened to former OLC head Steve Bradbury discuss metadata with Jennifer Daskal, and now I'm listening to Tim Edgar and Jennifer Grannick debate content collection. So I have Big Data on the brain, and I'm also sitting in front of a computer. So I started thinking: Is there any way to use Big Data to figure out how much people really care about all of this NSA stuff? The answer, at least in a crude way, is that there is: I can mine your search histories.
Google Trends is a fascinating service provided by Google that let's you, in a highly aggregated and anonymized fashion, look at the level of "search interest" in terms and subjects. So I ran a report on the search interest associated with the NSA over the past year.
Here is the result.
A few things stand out about the report. The first is the unsurprising fact that the Snowden revelations created a massive spike in interest in the subject of surveillance and NSA, which relatively quickly began tailing off. Today, the search interest measure stands at 19 (on a scale of 0-100), roughly twice what it was prior to the Snowden disclosures but far below the peak (defined as 100) that it hit in the few weeks following the initial Guardian story.
The second is that interest in NSA seems to vary dramatically regionally. Who would have thought that interest in the subject in Paraguay would wildly exceed interest either in United States or in a Germany scandalized by surveillance of its chancellor? (Explains Google, "Numbers represent search volume relative to the highest point on the map which is always 100.")
Third, if you doubt that this is largely an inside-the-beltway story, consider this table, which shows intensity in Washington dramatically higher than any other city:
Google searches are not a measure of public opinion; we don't, after all, know why people are searching for the basket of terms associated with NSA. They are, however, a fascinating measure of the intensity of public interest in a subject. Some of this data is entirely intuitive, confirming what one largely would have guessed. Some of it, however, is not.