The data-hypothesis relationship

Furthermore, seemingly hypothesis-free statistical tools and computational techniques also contain latent hypotheses and theories about what is important—what might be interesting, worth measuring or paying attention to. [...]while Y&L argue that a “hypothesis is a liability,” we argue that hypothesis-free observation is not possible (nor desirable) and that hypotheses in fact are the primary engine of scientific creativity and discovery. Hypotheses are needed. [...]there is a mismatch between the experiment and what Y&L are claiming, on a number of levels. [...]if radiologists or experimental subjects were asked to, say, “look for something unusual” or to “see if you can find the animal,” they would presumably find the gorilla. [...]visual awareness or recognition has little to do with size or surprisingness. [...]experimental subjects themselves might suspect that the study actually is not about counting basketball passes or about analyzing health data or finding cancerous nodules in lungs.