Its kinda within the definition of Big Data, which is that you have too much data to really inspect all of it. this means that the previous approach had to be i

Its kinda within the definition of Big Data, which is that you have too much data to really inspect all of it. this means that the previous approach had to be inverted: now its necessary to propose a theory, define queries to validate it and then evaluate the results, whereas before you'd take everything in and define a theory from the data set. the new approach is pretty much a textbook definition of a confirmation bias. But there is no other way to really approach these vast amounts of data, so its become the default.