We define learning as the increase in knowledge. Since RELATE’s goal is “…to study and improve learning…” many factors related to learning have been studied. Generally, two assessments of knowledge are required to measure learning. Given these, we primarily use two measures of learning, both involving a pre- and a post- assessment of knowledge.
- Effect size: Given the class’ pre-instruction bell-shaped curve of knowledge, the effect size is the improvement in knowledge on the post-assessment measured in standard deviations of the pre-test.
- Normalized Gain: The improvement of score divided by the maximum possible improvement from the pretest. More intuitively, normalized gain is the fraction of what is unknown on the pre-test that is learned.
We first developed Models of Learning, including IRT (Item Response Theory) and its adaption to analyze continuous learning.
To improve instruction, we start by studying Learning in Entire Courses, generally to find which course elements impart the greatest amount of learning per unit of student time and whether this depends on a student’s initial knowledge.
The “gold standard” of education research is randomly assigned Control group – Experimental Group experiments. RELATE’s research using this protocol is summarized in Learning in Controlled Environments.