We all know that the GMAT is a computer adaptive test, and computer adaptive tests give us questions based on the difficulty level that we "earn" as we take the test. How do the test writers at ACT (the organization that writes the GMAT; it used to be ETS, but ETS lost the contract to ACT 4-5 years ago; GMAC manages the algorithm and "owns" the test) determine which questions are harder than others?

First, ACT engages in a process called "normalization," wherein all freshly written questions are tested by actual test takers to determine what percentage answer the questions correctly (we know these questions as "experimental" questions). If too many people answer correctly, the question may need to be toughened up. If too few people answer correctly, the question may need to be dumbed down. ACT is looking to assemble a pool of questions that covers a range of difficulty, from cakewalk to mind-bending, and the test takers help them do so.