Running randomized trials has become standard practice for many software developers, but it is much less common within educational software.
One barrier to running experiments in educational contexts is that instruction often takes place in groups, and it is undesirable to have different students within a group assigned to different conditions within the learning experiment.
Experiment assignment must therefore take into account a student’s membership in a group, such as a class, the set of students taught by a particular teacher, a school, or a school district.
The assignment of individuals to condition as a group is called group random assignment, sometimes called cluster random assignment. Consistently managing group random assignment, particularly at scale, is a difficult task.
This paper describes how the free and open-source UpGrade platform helps manage this type of randomization for educational software experiments.