Knowing ground truth
There’s a weird thing about intelligence services and political reporting that I notice the other day. When Nate Silver’s data journalism dominated the election coverage by being right, it was contrasted against the other political coverage by journalists. The split that was revealed was in the theory of how one learns truth, the ground reality.
The 538 approach was to look at numbers, fiddle with them for accuracy, and then derive ground truth from what the number revealed. In contrast, the other approach was to cultivate sources inside a political campaign and ask them about ground truth. That is, the difference was basically “measure values and trust those” vs. “ask knowledge sources and trust those.”
In some ways these two approaches are the differences between the Soviet and the Western approach to intelligence collection and analysis. (The soviets did almost no analysis, they relied almost exclusively on assembling reports directly from collected documents and assessments.) This difference is kinda interesting in that measuring values and assessing them can give very reliable insight into ground truth, when human sources are blinded or self deluded into false beliefs.
On the other hand, the values can be just as erroneously interpreted as the human sources closer to the action. Some examples include 538 getting the trump nomination wrong, and the US IC completely missing the Indian nuclear tests despite having extensive satellite coverage of the test sites.
Neither approach is correct or wrong. It’s just interesting to me how the mainstream press is very Soviet style intelligence apparatus in their approach. “Find a well informed source and get documents or statements on the issue then assemble a report using this data.” Maybe there’s a good argument to be made for better analysis on news stories?