CONCEPT
Exploratory Data Analysis
John Tukey's discipline of looking at data before modeling it—the open-ended, surprise-ready examination that catches what formulas miss—and the practice AI most systematically discards.
Exploratory data analysis is the discipline
John Tukey founded on a single convicting observation: before you model your data, you should look at it. In the 1960s and 70s, mainstream statistics was dominated by confirmatory analysis—formulate a hypothesis, fit a model, compute a test, accept or reject. Tukey did not reject this, but he insisted it was the less important half of the job. The other half, the half everyone was skipping, was EDA: an open-ended, almost playful examination of data to discover what is in it before committing to any model at all. His 1977 book
Exploratory Data Analysis made the case, introducing the stem-and-leaf display, the box plot, and a constellation of resistant summaries designed to show structure to the human eye without letting a handful of extreme values dictate the picture. The philosophy was detective work: approach the evidence without knowing what you will find, let surprises surface, follow anomalies, form hunches that later analysis can test. Set this beside the dominant paradigm in modern machine learning—take a