Abstract
Many universities have general education requirements. However, Biomedical Engineering departments rarely offer courses to fulfill this requirement. No published benchmarks exist for engagement and demographics for a general education course in Biomedical Engineering. This paper characterizes an online general education course in Biomedical Engineering. Engagement was measured by monitoring the viewing and reading patterns of students. All instructional videos were delivered through the YuJa platform, which records how often each segment of each video is viewed. Readings were assigned through the Perusall platform, which records how long each student viewed each page of an assigned reading. The demographics of students enrolled in the class were provided by the university's institutional research office. The number of views for most instructional videos was measured to be about 75% of the number of enrolled students. This percentage did not vary after the first week. Viewing dropped gradually over the length of a 30-min video, the final minutes were viewed about 75% as many times as the first few minutes. Reading engagement dropped precipitously with the length of the assigned reading. The final pages of a 13-page reading assignment had about 10% as many views as the first pages. The average view time of the final pages was about 20% of the first pages. A significantly higher percentage of students enrolled in the general education class were domestic, first generation, low income, and female compared to the general undergraduate population at the university.
Introduction
General education courses are a hallmark of a liberal arts education. These courses are meant to introduce students to a range of ideas outside of their major area of concentration. Accrediting agencies in the United States require colleges and universities to include a general education requirement as part of their degree requirements for all students [1,2]. These requirements vary from campus to campus but typically require all students earning a bachelor's degree to take a few courses in various categories including humanities and arts, science and engineering, mathematics, and social sciences. At this university, students can choose from dozens of courses in the science and technology category with titles like Introduction to Ecology, Diseases of the 21st Century, The Atmosphere, and Classical Physics. However, no engineering courses were previously offered in this category.
A 10-week (one quarter) biomedical engineering course was designed to fulfill the university's science and technology general education requirement. The course, Engineering Innovations in Treating Diabetes, followed a historical path describing how physicians, scientists, and engineers went from the pre-insulin days of the early 1900s to the modern treatment of diabetes with engineered insulins and integrated glucose measurement/insulin pump systems. A more detailed description of the course is included in Appendix. The course was offered online to encourage broad enrollment. Online courses are offered asynchronously; thus, the class does not conflict with any other classes, work schedule, or athletic events.
The course consisted of three main components all tied together with the Canvas learning management system. First, instructional videos were delivered to students by way of the YuJa content management platform. The instructional videos ranged in length from 5 to 40 min. These videos were produced in the style of a low-budget documentary (with historical photos, videos, and a narrator) rather than in the style of an unedited recording voice-over PowerPoint. Second, collaborative readings were assigned on the Perusall social reading platform [3]. Perusall allowed the readings and discussion to take place online asynchronously. Finally, quantitative problems were assigned through Canvas. These quantitative problems taught the key engineering concept of optimization to students. The exercises were designed to be performed with Microsoft Excel's Solver since it was widely available to students. The optimization exercises provided a sampling of typical engineering mathematics problems (finding the shortest path and scheduling problems) without requiring any prerequisites in advanced mathematics.
The purpose of this paper is to quantitatively characterize both student engagement in the course and demographics of students enrolled in the course. This characterization will provide a benchmark against which other Biomedical Engineering (BME) courses and changes to this course can be evaluated.
Methods
Videos of the instructor were shot mostly in a faculty office with a Logitech c920 camera. The audio was simultaneously recorded with a lapel microphone. The instructor served as the narrator and on the final videos his face was on screen for only about 25% of the time. These raw videos were edited with Davinci Resolve. Editing removed long pauses and overlaid photos, videos, and diagrams onto the video. A relatively quiet −35-dB background music track was added to the final video to hide distracting background noise and to provide continuity across cuts. Closed captions were added to the final videos by a professional service to meet accessibility requirements.
Finished videos were hosted by YuJa and presented to students on a Canvas page. YuJa is an enterprise video content management system that records detailed analytics. YuJa reports the number of views for every 5-s segment of video. YuJa can track individual student's viewing time for each 5-s segment of video if logins are required. Students were not required to log in to view the videos, so the video analytics were aggregated across the entire class.
Collaborative reading was assigned on Perusall. Perusall is a social reading and annotation platform. Students were assigned one 10- to 20-page document per week to read and discuss. The documents were uploaded to Perusall in PDF form. Students had to be logged into Perusall to read and post comments; thus, Perusall has analytics available at the level of individual students. Perusall recorded the amount of time each student viewed each page in the readings. Perusall records both the total viewing time and the active viewing time. An “active view” is recorded whenever there is a key press or mouse movement within 2 min.
Demographic information was obtained through the campus office of institutional research with a custom-built web-based tool. The tool reports aggregate statistics by class, major, or for all enrolled undergraduates by term and year. Class standing (freshman, sophomore, etc.) is based on units completed; many 5-year students arrive on campus with a large number of units completed from high school advanced placement credits. Low-income students are defined by the university as students who requested and were granted a fee waiver on their initial application or by several other similar criteria. This “low-income” status is only determined during the initial application and stays the same throughout the student's enrollment.
This research was determined to be exempt from IRB regulations based on the finding that it involves normal educational practices and will not impact students' opportunity to learn. All students were adults, and there are no financial interests involved.
Results
Engagement was measured with both video analytics from YuJa and reading analytics from Perusall. Figure 1 presents results measured by video analytics. Three videos of varying lengths that were assigned to students during the first, fifth, and the ninth weeks of the 10-week quarter were analyzed. Measures were not taken to ensure that students watched the videos. Individual views were not attributable to a particular student, no grades were awarded for watching, etc.
This plot shows how viewership varies over the length of a video. The number of views was recorded during each 5-second segment of a video by YuJa, the video enterprise management system. The three videos were very similar in content and difficulty. Three different videos of varying lengths are shown, one from the first week of class, one from the fifth week of class, and one from the ninth week of the ten-week class. There were 198 students who were enrolled in the class throughout the quarter. The first week may be significantly higher because some students dropped from the class.
This plot shows how viewership varies over the length of a video. The number of views was recorded during each 5-second segment of a video by YuJa, the video enterprise management system. The three videos were very similar in content and difficulty. Three different videos of varying lengths are shown, one from the first week of class, one from the fifth week of class, and one from the ninth week of the ten-week class. There were 198 students who were enrolled in the class throughout the quarter. The first week may be significantly higher because some students dropped from the class.
Variation in student engagement across the length of the 10-week course was also quantified. Figure 2 presents the total number of views for each instructional video lesson sequenced from the beginning of the quarter to the end. Each week had three–four instructional video lessons for students to watch. Videos assigned during the first week had substantially more views than the other weeks.
This figure presents the number of views, as reported by the YuJa video management system, as a function of the video number. The videos are numbered and assigned sequentially, thus this graph can be interpreted as a measure of how student engagement varies across the 10-week quarter. A total of 34 videos were assigned for students to watch, about three or four per week. The relatively large number of views during the first few lectures is probably due to the large number of students who “shop” for general education courses, joining a few different courses for a week before dropping the class. The class had 198 students enrolled throughout the quarter.
This figure presents the number of views, as reported by the YuJa video management system, as a function of the video number. The videos are numbered and assigned sequentially, thus this graph can be interpreted as a measure of how student engagement varies across the 10-week quarter. A total of 34 videos were assigned for students to watch, about three or four per week. The relatively large number of views during the first few lectures is probably due to the large number of students who “shop” for general education courses, joining a few different courses for a week before dropping the class. The class had 198 students enrolled throughout the quarter.
Reading engagement was measured in two separate ways. One of the ten reading assignments, a 13-page description of the discovery and manufacture of insulin [4], was analyzed to quantify the total views per page, the average time spent on each page, and the total time spent by each student on the assignment. The specific assignment was for the student to read the 13-page document and provide three comments, or replies to comments, on specific sections of the document.
Figure 3 presents a breakdown of the student reading engagement as a function of the page number of the document. The figure shows that students viewed the earlier pages much more than the later pages. Even when they viewed later pages, they spent less time, on average, than on the earlier pages.
This assignment asked students to read and comment on a 13-page document through Perusall during the fourth week of the class. Students were required to supply three substantial comments, or replies to comments, throughout the document to obtain full credit. Of the 198 students in the class, 17 did not supply any comments on this assignment.
This assignment asked students to read and comment on a 13-page document through Perusall during the fourth week of the class. Students were required to supply three substantial comments, or replies to comments, throughout the document to obtain full credit. Of the 198 students in the class, 17 did not supply any comments on this assignment.
Figure 4 shows a histogram of active reading times for the same 13-page assignment. Perusall records both the total time a reading assignment was open and the active reading time for an assignment.
This histogram shows the distribution in active reading times for the 198 students in the class. The reading/commenting assignment was the same 13-page document assigned in Fig. 3. Active reading requires some mouse or keypress at least once every two minutes. This histogram groups student active reading times into ten-minute bins, ending at the labeled value, 0–10 min, 11–20 min, etc. Twelve students had zero active reading times on this assignment. These are included in the 10-min bin.
This histogram shows the distribution in active reading times for the 198 students in the class. The reading/commenting assignment was the same 13-page document assigned in Fig. 3. Active reading requires some mouse or keypress at least once every two minutes. This histogram groups student active reading times into ten-minute bins, ending at the labeled value, 0–10 min, 11–20 min, etc. Twelve students had zero active reading times on this assignment. These are included in the 10-min bin.
Demographics of students who chose to take this course were compared to both the population of BME majors and the overall university undergraduate population.
The students enrolled in this BME general elective course came from a wide variety of majors. Table 1 shows the distribution of the ten most common majors of students who enrolled in this class. These ten only account for about 60% of the class. The table also compares this distribution to the distribution of majors for all undergraduate students at the university. This BME general education class was relatively most popular among students majoring in Biomedical Engineering, Criminology, Education, and undeclared students who have not yet settled on a major.
The ten most common majors of students enrolled in the BME general education class
| Major | BME General Education Class | University undergraduate population | Ratio of BME GE class to University undergraduate population |
|---|---|---|---|
| Criminology, Law and Society | 10.2% | 3.9% | 2.6 |
| Biomedical Engineering | 9.7% | 1.5% | 6.3 |
| Undeclared | 8.2% | 5.0% | 1.7 |
| Education Sciences | 6.1% | 2.6% | 2.4 |
| Biological Sciences | 5.6% | 11.5% | 0.5 |
| Business Administration | 4.6% | 3.2% | 1.4 |
| Business Economics | 3.6% | 5.7% | 0.6 |
| Political Science | 3.6% | 2.8% | 1.3 |
| Psychological Science | 3.6% | 3.0% | 1.2 |
| Major | BME General Education Class | University undergraduate population | Ratio of BME GE class to University undergraduate population |
|---|---|---|---|
| Criminology, Law and Society | 10.2% | 3.9% | 2.6 |
| Biomedical Engineering | 9.7% | 1.5% | 6.3 |
| Undeclared | 8.2% | 5.0% | 1.7 |
| Education Sciences | 6.1% | 2.6% | 2.4 |
| Biological Sciences | 5.6% | 11.5% | 0.5 |
| Business Administration | 4.6% | 3.2% | 1.4 |
| Business Economics | 3.6% | 5.7% | 0.6 |
| Political Science | 3.6% | 2.8% | 1.3 |
| Psychological Science | 3.6% | 3.0% | 1.2 |
Note: For comparison, the “University undergraduate population” column shows the percentage of students across the university who have chosen that major. The final column indicates the ratio of student majors found in the class to those found in the general university population.
Table 2 describes the detailed demographics of students enrolled in the BME general education class and compares these students to both the general university undergraduate population and to students who have declared BME as their major. Overall, students who enrolled in the general education course were substantially more likely to be first-generation college students and from low-income families compared to students who have chosen BME as their major. When compared to the general undergraduate university population, students who chose to enroll in this BME general education class were substantially less likely to be international students and more likely to be in the early stage of their university education. Students in the BME general education class were also more likely to be from low-income families and the first generation of their family to attend college than the general university undergraduate population. Female students enrolled in the BME general education class with a higher percentage than both the general undergraduate population and the population of BME majors.
Demographic details of the students enrolled in the BME general education class
| Classification | BME GE Class | University undergraduate population | BME Majors |
|---|---|---|---|
| Fresh | 17% | 13.1% | N/A |
| Sophomore | 23% | 17.7% | N/A |
| Junior | 16% | 28.0% | N/A |
| Senior | 44% | 41.1% | N/A |
| International | 6% | 16% | 7% |
| First generation | 59% | 49% | 34% |
| Low income | 38% | 31% | 23% |
| Female | 72% | 52% | 51% |
| Classification | BME GE Class | University undergraduate population | BME Majors |
|---|---|---|---|
| Fresh | 17% | 13.1% | N/A |
| Sophomore | 23% | 17.7% | N/A |
| Junior | 16% | 28.0% | N/A |
| Senior | 44% | 41.1% | N/A |
| International | 6% | 16% | 7% |
| First generation | 59% | 49% | 34% |
| Low income | 38% | 31% | 23% |
| Female | 72% | 52% | 51% |
Note: Students in the class were substantially more likely to be early in their university education compared to the general university population. Compared to students who are pursuing a BME degree, students enrolled in the GE class were about the same fraction of international and much more likely to be first-generation college students, from low-income families and female.
Discussion
The National Survey of Student Engagement (NSSE) has shown that students who participate in their education with a high level of engagement retain material better, and this effect is even stronger with students from underrepresented groups [5,6]. The NSSE measures student engagement at an institutional level, using benchmarks like how much time students spend on reading assigned books, writing papers, and preparing for class. The assumption is that these benchmarks relate directly to a student's learning and development [7] However, measuring these benchmarks is not easy, requiring either trained observers [8–10] or self-reporting [11,12]. The use of video and reading analytics should make engagement benchmarks more widely available to instructors.
Multiple measures of engagement can be extracted from video analytics. Figure 2 shows that the number of video views was consistently between 50 and 100% of the enrollment. The number of views declined after the first week but did not change significantly throughout the rest of the class. The decline after the first week is probably caused by students who were shopping for classes. One limitation is that although initiating a view is clearly a measure of engagement and terminating a view is a measure of ending engagement, continuing to view does not necessarily indicate the student is engaged. Students could be multitasking or simply ignoring the video.
Figure 1 shows that student viewing declined slightly during the length of a video. An informal rule widely passed on for instructional videos posits that students have limited attention span. This rule suggests that instructional videos should always be shorter than 10 min or so. The data did not support this rule. A small decline in viewership was observed with time, but typical viewership after 30 min was within 20% of the viewership after 5 min. Others have also noted that this rule is not supported with data [13].
In contrast to the gradual decline in viewing found in longer instructional videos, reading declined precipitously page by page with longer reading assignments. The final pages of a 13-page reading assignment had about 10% as many views as the first pages. In addition, the average view time of the final pages was about 20% of the first pages. Few students read to the end of the assignment, and those who did reach the final page spent little time on it. Very little published data exists on the reading behavior of students in online classes, but there are studies of researchers reading scholarly literature [14]. Web analytics from academic publishers and libraries shows that professors act in a comparable manner, reading many articles shallowly and rarely spending more than a few minutes on any one article. This skimming behavior is attributed to several reasons including boring articles, multitasking, conditioning by PowerPoint/Twitter, and the small screens of smartphones [15].
The demographics of students in this BME general education class were quite different than in classes for students who chose Biomedical Engineering as a major. The online general education class had a higher percentage of both first-generation and low-income students. This higher percentage is probably due to a combination of both the course being offered as a general education class and the course being offered online. During the quarter, this general education BME class was offered, and students could choose from 59 other classes that would also fulfill the same requirement. Of the 60 different choices, only four were offered online. Online courses taught asynchronously may be more appealing to low-income and first-generation students since schedule flexibility might increase employment opportunities. Underrepresented minorities are underrepresented in STEM disciplines compared to other university majors. The reasons behind this are thought to be complex and multifactorial, related to academic, financial, and social challenges that these students face [16].
The online general education BME class also had significantly more females, compared to both the university's undergraduate population and BME majors. Marketing research has consistently shown more women than men choose online colleges. Annual surveys of online college students found that 70–75% of students were female between 2012 and 2015 [17]. However, those surveys were for fully online colleges and this paper is about a single online class in a residential college. It might be that a higher percentage of women prefer online learning compared to men.
An alternative explanation for the excess number of women enrolled in this class is that women might prefer to enroll in groups. In this class, a group of about a dozen women who were all members of a university athletic team appeared to have chosen to take the class together. This single group is not large enough to account for the exceptionally large observed difference between males and females enrolled in the course, but there may be other groups of women who acted jointly to enroll.
In conclusion, student engagement in this course dropped gradually with the length of instructional videos up to 30 min long, but it dropped precipitously with reading assignments just a few pages long. The demographics of this online general education had significantly more women, low-income, and first-generation students than expected. These findings indicate that online courses might be one way to increase diversity and inclusion in biomedical engineering classes.
Conflict of Interest
There are no conflicts of interest. This article does not include research in which human participants were involved. Informed consent not applicable.
Data Availability Statement
No data, models, or code were generated or used for this paper.
Appendix
Outline of the Course:
Week 1. Getting Started. Students become familiar with the technology and routine needed to be successful in the course.
Week 2. Pre-insulin era. The importance of basic research. Diabetes discovery pre1800. The discovery of the pancreas. Improvements in surgical technique. Oscar Minkowski's removal of the pancreas. Animals in testing, anti-vivisection.
Week 3. Discovery of insulin. Fred Banting's early life. Banting's idea and move to Toronto. Banting, Best, Macleod discovery of insulin. Treating Leonard Thompson, the first patient.
Week 4. Industrialization and purification of insulin. Alcohol purification works, then fails at scale-up. The importance of pH. Arnold Beckmann develops the first integrated pH meter in the 1930s. Patents, licensing, and Eli Lilly. George Walden discovers isoelectric precipitation.
Week 5. Complications of diabetes appear. Complications from diabetes. Peripheral arterial disease and foot, kidney, and eye problems. Neuropathy and other complications. Examples with particular historical patients: George Minot and case studies.
Week 6. Glucose measurement. Glucose measurements from urine. Ames Reflective Meter. Detlev Müller and the isolation of glucose oxidase. Leland Clarke invents the oxygen electrode. Today, continuous glucose monitors. Control of glucose by changing food and insulin. NPH insulin.
Week 7. Clinical Trials. Understanding the structure of clinical trials. The Tolstoi control controversy resolved by DCCT and fundus photography. The importance of modern clinical trials illustrated by beta carotene.
Week 8. Recombinant DNA and production of insulin. Projected insulin shortage in the 70's. Synthesis of insulin. Axel Ulrich clones insulin from an insulinoma. Ethical concerns in the United States about recombinant DNA research. Eli Lilly and Irving Johnson mass-produce recombinant insulin. Modern engineered insulins.
Week 9. Innovations in insulin delivery. Alternatives to the bottle and syringe: prefilled insulin “pens,” continuous insulin pumps, and even breathable insulin. Does the c-peptide have any value as an additional drug for patients with diabetes?
Week 10. Future of Technology for Diabetes and Diversity in Engineering. Artificial pancreas, autoimmune diseases, and vaccines. Why would a company need a diverse group of engineers? Story of Margaret Crane, the inventor of the home-based pregnancy test.
