Questions and answer choices were developed through a two step process.
Step 1: Crowd workers were presented with a set of textbook passages and asked to develop a question and answer for one of the passages of their choosing.
Step 2: A second crowd worker was asked to take the question and answer and generate three distractors with the help of a model trained to predict good distractors.