Authors: Elli J. Theobald et al
First author’s institution: University of Washington
Journal: Proceedings of the National Academies of Science, vol 117, no 12 (2020) [open access]
In our previous article, we discussed that racially underrepresented students were the most disadvantaged when it came to grades, with even the least advantaged non-underrepresented students outperforming the most advantaged underrepresented students. Poor performance in introductory STEM courses are a key reason for students leaving STEM or even dropping out of college. Thus, while racially underrepresented students enter college with the same interest in STEM as their overrepresented peers, attrition in part due to lower grades creates an underrepresentation in degree recipients.
Traditionally, approaches to combat disparities in performance have tended to occur outside of the classroom. These include supplemental instruction, often with a teaching assistant or advanced undergraduate students, bridge programs and comprehensive support programs, and psychological interventions to provide emotional support. While these approaches are successful to a degree, a valid question is why interventions to close performance gaps occur outside of the classroom (often through an opt-in model) rather than inside the classroom where all students could participate. Given that active learning has been shown to increase students’ grades and reduce failure rates, the authors of today’s paper wondered if active learning could also reduce the performance gap between racially underrepresented and overrepresented students. With some caveats, the answer seems to be yes.
To conduct this study, the authors analyzed the results of many studies which had previously tried to answer this question (a meta-analysis) and looked for a general trend. For a study to be included, it needed to compare either the exam scores or passing rates of an active learning STEM course to an otherwise identical lecture STEM course. In addition, the results needed to be reported for both underrpresented and overrepresented students or by student’s socioeconomic status. When searching for studies that met all of these criteria, the authors found 15 studies that compared exam scores and 26 studies that compared passing rates, together representing 225 separate STEM classrooms and over 50,000 students.
To analyze their results, the authors combined students from racially or ethnically underrepresented groups or low socioeconomic status into a single group called minoritized groups in STEM (MGS) as it is common in the literature and this group is often of highest interest to science policy experts.
When comparing exam scores, the authors found that the gap between students from minoritized groups in STEM and students not from minoritized groups in STEM shrunk by 33% in active learning courses compared to lecture courses (a 0.62 standard deviation gap in lecture courses and 0.42 standard deviation gap in active learning course; see figure 1).
In addition, the gap in non-passing rates shrunk by 45% in active learning courses compared to lecture courses (a gap of 7.1% in lecture courses and 3.9% in active learning courses).
To further examine any underlying patterns, the researchers investigated the individual results from each of the studies. Of the 15 studies looking at exam scores, 10 found that students from minoritized groups in STEM had greater gains than students from non-minoritized groups in STEM. However, in 8 of those 10 studies, students from non-minoritized groups in STEM still outperformed students from minoritized groups in STEM. Similarly, in 15 of the 26 studies examining passing rates, students in minoritized groups in STEM saw greater decreases in failure rates than students in non-minoritized groups in STEM, even though in most studies, students from non-minoritized groups in STEM passed at higher rates (figure 2).
Additionally, the authors found that the intensity of active learning, that is, the amount of class time spent on active learning, correlated with the amount the gaps decrease. Courses in which active learning occupied more of the class time showed the greatest reductions in the performance and passing rate gaps (figure 3).
When reading the results, you may have noticed that the reduction in the exam performance gap is relatively small, corresponding to only around 0.2 standard deviations or a small effect size. However, average grades for students from minoritized groups in STEM in introductory STEM courses tend to be between 2.0-2.4 on a 4.0 scale while minimum passing grades are often between 1.5-1.7. Thus, even a small boost in grades can mean the difference between passing a course and continuing onto the next course in the sequence or failing the course and possibly failing out of the program.
Overall, the results of this study of studies suggests that active learning can decrease gaps on exams and passing rates between students from minoritized groups in STEM and students from non-minoritized groups in STEM. However, the implementation matters and the greatest decreases in the gaps come from courses with high-intensity active learning.
So what does high intensity active learning look like? The authors lay out two broad elements, which they call the heads and hearts hypothesis.
First, the authors suggest the course needs to include deliberate practice, which emphasizes highly focused efforts geared toward improving performance, scaffolded exercises in areas students may not yet have the relevant skills, immediate feedback, and repetition. These are all represented in the best practices for active learning.
Second, the authors emphasize inclusive teaching, which includes treating students with dignity and respect, communicating confidence in students’ abilities to meet high standards, and demonstrating a genuine interest in students’ intellectual and personal growth. Together, the authors claim that these two elements are required for any active learning class that seeks to reduce the performance and passing rate gaps.
I am a physics and computational mathematics, science, and engineering PhD student at Michigan State University and the founder of PERbites. I’m interested in applying machine learning to analyze educational datasets and am currently studying the physics graduate school admissions process.