The present disclosure features systems and methods for analyzing student learning and calibrating the difficulty of questions on a test or examination. In one embodiment, a method for analyzing the learning of a student includes administering, by an assessment agent, a task to a student, the task comprising a question having an associated difficulty. The assessment agent receives a response to the question from the student and evaluates the response to generate an observable, the observable comprising information related to the response. A posterior determination of the student's ability is then calculated by incorporating the observable into an ability model associated with the student, and the posterior determination of ability may be compared with the difficulty of the question, a skill acquisition probability, or other measure. The student's response, or a plurality of responses from students within a cohort, may be used to determine the difficulty of each question.