Randomized Groups

Introduction

---

Educational research designs can involve randomly assigning students in a class to groups. For example, one such group can serve as a control and others can provide interventions under investigation. This courseware demonstrates forming such groups based on the order in which students in a class begin an online courseware assignment—a random variable.

Concept

---

Once an online courseware is assigned to a class, students go online and begin the assignment. The order in which they will do this is entirely unknowable in advance, assuming there is no conspiracy or instruction to establish a knowable order. eLMS courseware deliveries are stateful, meaning that, once started, a student's courseware delivery record forever retains information about the student's use of the courseware. The eLMS platform is itself stateful, retaining information that can be shared among sets of courseware deliveries called profiles. These features and the randomness of the time at which a student begins an assignment can be used to place the student into a group when her or his delivery record is created, based on the order in which a set of such requests arrive at an eLMS server.

The animation below shows this concept. As students begin the assignment, an n-state shared variable in the Class profile is used to assign each student to a group and then incremented. This is visualized in the animation as a turnstile that gates the student to a particular group.

A significant advantage of this approach is that the groups are automatically balanced. Regardless of the number of groups, the maximum difference in size between the groups is exactly one. Another advantage is that the number of groups can be defined simply by changing the control variable n. The design can thus be used to investigate multiple alternative interventions simultaneously.

Try It Yourself

---

A three group demonstration of the courseware is accessible via a QuotaLink. You can create new student identities in trying out the courseware by providing (potentially fictitious) information into the Launcher, as shown below. Each identity created must bear a unique email address.

Design

---

The straightforward courseware design appears in the figure below. The model window on the left provides the module design. The window on the right shows the contents of the Conds condition set that provides services that control the assignment to groups.

Entries in the Class profile (Class_Profile) are keyed by the courseware name. The assignment of leaners to groups is also recorded in the Class profile. The identity of the courseware and learner are obtained using services (Courseware and Learner, respectively) provided by the eLMS Platform Services mini-project from the CAPE Repository.

The module design (left window) uses three granules to simply represent the "treatment" provided to each of the groups. Adaptation is used to provide the appropriate content using the element Select with branching conditions based on the student's group assignment contained in Conds.Group. Note that such adaptation points can occur multiple times within the courseware design to gate between common and group-specific elements.

The Conds condition set (right window) provides a set of definitions used in the group assignment. We have already mentioned the Group member that holds the learner's group assignment. The function assign peforms the actual assignment. Authors reusing the design need only provide an Action, such as the one named Assign, with the definition:

Conds.Group = Conds.assign()

This design is available in the CAPE Repository in the author area of Larry Howard in a folder named Research Designs.

Discussion

---

We are strong advocates of using disciplined inquiry to study courseware design innovations created with CAPE and delivered via eLMS. One of the primary emphases of our work has been to enable information gleaned from student use of online courseware to inform reflection on designs and their incremental improvement over time. The capabilities described in this courseware demonstration are applicable to such an improvement process.

Research designs that involve splitting a class into groups are useful when a control is unavailable from another section of the class or from an earlier semester. A down side of the approach is a consequential reduction of sample sizes.

It should be noted that the usefulness of capabilities of this demonstration courseware are not limited to research designs. There are many legitimate pedagogical uses of dividing up the students in a class. Some of these can make use of random assignment. One that we have mentioned in the past is forming discussion sub-groups, where materials presented earlier in the courseware design are discussed among learners in small groups online.

Feedback

---

We welcome your feedback on any aspect of the design or delivery of this courseware. Please send your remarks to the author. Learn more about CAPE and eLMS at the Adaptive Learning Technologies web site.