Thursday, November 8, 2012

The role of intuition in data analysis

Most of you readers probably followed the US elections with varying degrees of interest and passion, but as scientists are also aware of the "role" of Nate Silver and his blog FiveThirtyEight.com. Lots has been written about his success, which is "just" a nice example of the strengths of the scientific process, and thus not that surprising. What does surprise me, is the reluctance of journalists to come to terms with what he does. See this very insightful blog post by Mark Coddington.
"Silver’s process — his epistemology — is almost exactly the opposite of this: Where political journalists’ information is privileged, his is public, coming from poll results that all the rest of us see, too. Where political journalists’ information is evaluated through a subjective and nebulous professional/cultural sense of judgment, his evaluation is systematic and scientifically based. It involves judgment, too, but because it’s based in a scientific process, we can trace how he applied that judgment to reach his conclusions."
This reluctance reflects, amongst other aspects, a classic case of a an either/or role for intuition in science/data analysis. We already had a lengthy exchange about the scientific process on this blog (see for instance this and this), and Stefan's main critique against the scientific process as a method is maybe summarized in this statement:
"I am claiming here that in order to do good science, one must know which auxiliary assumptions are reasonable to question in the face of evidence that conflicts with one’s predictions."
One way to look at this is that intuition plays that role of knowing "which auxiliary assumptions are reasonable". So it is not an either/or relationship, but really a feedback between intuition and hypothesis testing.

Or my last commandment: "Thou Shalt Listen to thy Intuition, but Follow the Data."

No comments:

Post a Comment