Massive open online courses (MOOCs) collect valuable data on student learning behavior; essentially complete records of all student interactions in a self-contained learning environment, with the benefit of large sample sizes. Here, we offer an overview of how the 108,000 participants behaved in 6.002x - Circuits and Electronics, the first course in MITx (now edX) in the Spring 2012 semester. We divided participants into tranches based on the extent of their assessment activities, ranging from browsers (constituting ~76% of the participants but only 8% of the total time spent in the course) to certificate earners (7% of participants who accounted for 60% of total time). We examined how the certificate earners allocated their time among the various course components and what fraction of each they accessed. We analyze transitions between course components, showing how student behavior differs when solving homework vs. exam problems. This work lays the foundation for future studies of how various course components, and transitions among them, influence learning in MOOCs.
Though free online courses are not new,8 they have reached an unprecedented scale since late 2011. Three organizationsCoursera, edX, and Udacityhave released MOOCs13 drawing more than 100,000 registrants per course. Numbers from these three initiatives have since grown to more than 100 courses and three million total registrants, resulting in 2012 being dubbed "The Year of the MOOC" by the New York Times.16 Though there has been much speculation regarding how these initiatives may reshape higher education,6,12,20 little analysis has been published to date describing student behavior or learning in them.
6.002x, Procedures, Data Analysis
With some modification for online delivery, the 14 weeklong units of 6.002x largely mirrored a traditional on-campus course in both format and timing. The course sequence (see Figure 1, left navigation bar) involves lecture sequences consisting of lecture videos (annotated PowerPoint slides and actual MIT lectures) with embedded lecture questions, tutorial videos (recitation substitute), homework (three to four multi-part problems), and lab assignments (interactive circuit toolbox). Overall grades were determined by homework (15%), labs (15%), a midterm (30%), and a final (40%). Supplementary materials (see Figure 1, top navigation bar) included a course textbook (navigable page images), a staff- and student-editable wiki and moderated student discussions. For further exploration of course structure and available resources, see the archived course at https://6002x.mitx.mit.edu/.
Parsing tracking logs. Analysis of tracking logs is an established means for understanding student behavior in blended and online courses.5,14 In the 6.002x tracking logs, each interaction (click) contained relevant information, including username, resource ID, interaction details, and timestamp. Interaction details are context-dependent (such as correctness of a homework problem submission, body text of a discussion post, and page number for book navigation). The edX software is distributed through the cloud; meaning interaction data is logged on multiple servers. In total, approximately 230 million interactions were logged in 38,000 log files over the initial Spring 2012 semester.
The correlation of attrition with less time spent in early weeks begs the question of whether motivating students to invest more time would increase retention rates.
We preprocessed the logs into separate time-series for each participant, then compiled participant-level descriptive statistics on resource usage, including number of unique resources accessed, total frequency of accesses per resource type, and total time spent per resource. We also parsed problem submissions, generating a response matrix including correctness and number of attempts. Where possible, we crosschecked our event-log assessment data against a MySQL database serving the 6.002x courseware. All log parsing was performed through standard modules in Python and R.
Estimation of time spent on resources. Time estimation for each participant involved measuring the durations between a student's initial interaction with a resource and the time the student would navigate away. We accumulated durations calculated from each participant's time series for each separate course component, including homework, book, and discussion forums. We found evidence that durations shorter than three seconds represent students navigating to desired resources; hence, we do not count these intervals as activity. In addition, we did not accumulate durations longer than one hour, assuming users have disengaged from their computers. Using alternate values of the high cutoff (20 minutes to one hour) can change overall time by 10%20% but did not significantly alter relationships regarding time allocation among course components or total time spent by different participants.
An important point is that time accumulated is associated with the resource displayed at the moment; for example, if a student references the book while doing homework, this duration is accumulated with book time. In our case, only direct interactions with the homework are logged with homework resources. There are clearly alternatives to this approach (such as considering all time between opening and answering a problem as problem-solving time21). Our time-accumulation algorithm is partially thwarted by users who open multiple browser windows or tabs; edX developers are considering ways to account for this in the future.
The novelty and publicity surrounding MOOCs in early 2012 attracted a large number of registrants who were more curious than serious. We still take participation in assessment as an indication of serious intent. Of the 154,000 registrants in 6.002x in spring 2012, 46,000 never accessed the course, and the median time spent by all remaining participants was only one hour (see Figure 2a). We had expected a bimodal distribution of total time spent, with a large peak of "browsers" who spent only on the order of one hour and another peak from the certificate earners at somewhere more than 50 hours. There was, in fact, no minimum between these extremes, only a noticeable shoulder (see Figure 2a). The intermediate durations are filled with attempters we divided into tranches (in colors) on the basis of how many assessment items they attempted on homework and exams: browsers (gray) attempted < 5% of homework; tranche 1 (red) 5%15% of homework; tranche 2 (orange) 15%25% of homework; tranche 3 (green) > 25% of homework; and tranche 4 (cyan) >25% of homework and 25% of midterm exam. Certificate earners (purple) attempted most of the available homework, midterm, and final exams. The median total time spent in the course for each tranche was 0.4 hours, 6.4 hours, 13.1 hours, 30.0 hours, 53.0 hours, and 95.1 hours, respectively. In addition to these tranches, just over 150 certificate earners spent fewer than 10 hours in the course, possibly representing a highly skilled tranche seeking certification. Similarly, just over 250 test takers spent fewer than 10 hours in the course and completed more than 25% of both exams but did not earn a certificate.
The average time spent in hours per week for participants in each tranche is shown in Figure 2c. Tranches attempting fewer assessment items not only taper off earlier, as the majority of participants effectively drop out, but also invested less time in the first few weeks than the certificate earners. The correlation of attrition with less time spent in early weeks begs the question of whether motivating students to invest more time would increase retention rates.
In the rest of this article, we restrict ourselves to certificate earners, as they accounted for the majority of resource consumption; we also wanted to study time and resource use over the whole semester.
Frequency of accesses. Figure 3a shows the number of active users per day for certificate earners, with large peaks on Sunday deadlines for graded homework and labs but not for lecture questions. There is a downward trend in the weeks between the midterm and the final exam (shaded regions). No homework or labs were assigned in the last two weeks before the final exam, though the peaks persist. We plotted activity in events (clicks subject to time cutoffs) per active student per day for assessment-based course components and learning-based components in Figure 3b and Figure 3c. Homework sets and the discussion forums account for the highest rate of activity per student, with discussion activity increasing over the semester. Lecture question events decay early as homework activity increases. Textbook use peaks during exams, and there is a noticeable drop in textbook activity after the midterm, as is typical in traditional courses.18
Time on tasks. Time represents the principal cost function for students, so it is important to study how students allocate time among available course components.15,19 Figure 4 shows the most time is spent on lecture videos; since three to four hours per week is close to the total duration of the scheduled videos, students who rewound and reviewed the videos must compensate for those speeding up playback or omitting videos.
The most significant change over the first seven weeks was the apparent transfer of time from lecture questions to homework, as in Figure 4. Considering a performance-goal orientation (see Figure 5), it should be noted that homework counted toward the course grade, whereas lecture questions did not. But even on mastery-oriented grounds, students might have viewed completion of homework as sufficient evidence of understanding lecture content. The prominence of time spent in discussion forums is especially noteworthy, as they were neither part of the course sequence nor did they count for credit. Students presumably spent time in discussion forums due to their utility, whether pedagogical or social or both. The small spike in textbook time at the midterm, a larger peak in the number of accesses, as in Figure 3, and the decrease in textbook use after the midterm are typical of textbook use when online resources are blended with traditional on-campus courses.18 Further studies comparing blended and online textbook use are also relevant.3,17
Percentage use of course components. Along with student time allocation, the fractional use of the various course components continues to be an important metric for instructors deciding how to improve their courses and researchers studying the influence of course structure on student activity and learning. For fractional use, we plotted the percentage of certificate earners having accessed at least a certain percentage of resources in a course component (see Figure 5). Homework and labs (each 15% of overall grade) reflect high fractional use. The inflection in these curves near 80% might have been higher but for the course policy of dropping the two lowest-graded assignments. The low proportionate use of textbook and tutorials is similar to the distribution observed for supplementary (not explicitly included in the course sequence) e-texts in large introductory physics courses,16 though the 6.002x textbook was assigned in the course syllabus. The course authors were disappointed with the limited use of tutorial videos, suspecting that placing tutorials after the homework and laboratory (they were meant to help) in the course sequence was partly responsible. (The wiki and discussion forums had no defined number of resources so are excluded here.)
To better understand the middle curves representing lecture videos and lecture problems, it helps to recall that the negative slope of the curve is the density of students accessing that fraction of that course component (see Figure 5b and Figure 5c). Interestingly, the distribution for the lecture videos is distinctly bimodal: 76% of students accessed over 20% of the videos (or 24% of students accessed less than 20%), and 33% accessed over 80% of the videos. This bimodality merits further study into learning preferences; for example, do some students learn from other resources exclusively? Or did they master the content prior to the course? The distribution of lecture-problem use is flat between 0% and 80%, then rises sharply, indicating that many students accessed nearly all of them. Along with the fact that the time on lecture questions drops steadily in the first half of the term (see Figure 2), this distribution suggests students not only allocated less time to them, some abandoned the lecture problems entirely.
Resources used when problem solving. Patterns in the sequential use of resources by students may hold clues to cognitive and even affective state.2 We therefore explored the interplay between use of assessment and learning resources by transforming time-series data into transition matrices between resources. The transition matrix contains all individual resource-resource transitions we aggregated into transitions between major course components. The completeness of the 6.002x learning environment means students did not have to leave it to reference the textbook, review earlier homework, or search the discussion forums. We thus had a unique opportunity to observe transitions to all course components accessed by students while working problems. In previous studies of online problem solving this information was simply missing.21
Figure 6 highlights student transitions from problems (while solving them) to other course components, treating homework sets, the midterm, and the final exam as separate assessment types of interest. Figure 6 shows the discussion forum is the most frequent destination during homework problem solving, though lecture videos consume the most time. During exams (midterm and final are similar), previously done homework is the primary destination, while the book consumes the most time. Student behavior on exam problems thus contrasts sharply with behavior on homework problems. Note that because homework was aggregated, we could not isolate "references to previous assignments" for students doing homework.
This article's major contribution to course analysis is showing how MOOC data can be analyzed in qualitatively different ways to address important issues: attrition/retention, distribution of students' time among resources, fractional use of those resources, and use of resources during problem solving. Among the more significant findings is that participants who attempted over 5% of the homework represented only 25% of all participants but accounted for 92% of the total time spent in the course; indeed, 60% of the time was invested by the 6% who ultimately received certificates. Participants who left the course invested less effort than certificate earners, with those investing the least effort during the first two weeks tending to leave sooner. Most certificate earners invested the plurality of their time in lecture videos, though approximately 25% of the earners watched less than 20%. This suggests the need for a follow-up investigation into the correlations between resource use and learning. Finally, we highlight the significant popularity of the discussion forums in spite of being neither required nor included in the navigation sequence. If this social learning component played a significant role in the success of 6.002x, a totally asynchronous alternative might be less appealing, at least for a complex topic like circuits and electronics.
Some of these results echo effects seen in on-campus studies of how course structure affects resource use18 and performance outcomes4,11,19 in introductory (college) courses. This and future MOOC studies should further illuminate on-campus education generally. On the other hand, MOOCs could well take advantage of insights from existing research in on-campus education (such as frequent exams drive resource use and maximize learning outcomes11).
Textbook use peaks during exams, and there is a noticeable drop in textbook activity after the midterm, as is typical in traditional courses.
Finally, we emphasize that MOOCs provide a unique view into the learning of a large, diverse population of students, allowing research based on detailed insight into all aspects of a course. In contrast to most previous studies of on-campus educational environments, we have time-stamped logs of essentially all student behavior and associated learning throughout the entirety of a course, all with solid statistics and the ability to study specific student cohorts (such as based on effort, learning habits, and demographics9). Combining time-on-task observations with measures of learning paves the way for measuring learning valuethe amount learned per unit time spent on a given course componentpossibly extending previous studies of online learning.7,15 This, in turn, will allow a process of cyclic improvement based on research development, experimentation, and measurement of learning outcomes, supporting improvement of educational content and delivery. Since many MOOCs largely mirror traditional on-campus courses in types of resources, format, and chronology, we anticipate insights into, and improvements of, learning in traditional on-campus courses as well.
This work is supported, but is not endorsed, by National Science Foundation grant DUE-1044294; additional support provided by a Google Faculty Award. We thank MITx for data access and J. deBoer and other members of the Teaching and Learning Laboratory and the Research in Learning, Assessing and Tutoring Effectively groups at MIT for their helpful suggestions and comments.
2. Baker, R.S., D'Mello, S.K., Rodrigo, M.M.T., and Graesser, A.C. Better to be frustrated than bored: The incidence, persistence, and impact of learners' cognitiveaffective states during interactions with three different computer-based learning environments. International Journal of Human-Computer Studies 68, 4 (2010).
13. McAuley, A., Stewart, B., Siemens, G., and Cormier, D. The MOOC model for digital practice. Social Sciences and Humanities Research Council, Knowledge Synthesis Grant on the Digital Economy, 2010.
18. Seaton, D.T., Bergner, Y., Kortemeyer, G., Rayyan, S., Chuang, I., and Pritchard, D.E. The impact of course structure on etext use in large-lecture introductory-physics courses. In Proceedings of the Physics Education Research Conference, 2013.
19. Stewart, J., Stewart, G., and Taylor, J. Using time-on-task measurements to understand student performance in a physics class: A four-year study. Physical Review Special Topics-Physics Education Research 8, 1 (2012).
Figure 6. Transitions to other components during problem solving on (a) homework, (b) midterm, and (c) final. Arrows are thicker in proportion to overall number of transitions, sorting components from top to bottom; node size represents total time spent on that component.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2014 ACM, Inc.