There can be little doubt that understanding how instructors instruct and how learners learn is important and challenging in higher education. There is currently much discussion regarding the subject of learning analytics and how meaningful results from such analyses can inform students, faculty, institutions, and other key stakeholders.
One dimension that is central to quality learning analytics, as information for students or information for instructors, is the ability to challenge core assumptions about the learning environment. Students may hold beliefs that may or may not align with instructor desires. Of course, instructors may hold beliefs that may or may not be linked to student outcomes.
In my experience, one of the most overlooked areas of evaluating and assessing student performance is the aspect of student networks. Of the many types of student networks, the understanding how students form self-initiated, study-partner networks are particularly important for improving academic and professional success. Ideally, this understanding should lead to interventions that can assist underserved and underrepresented student groups specifically.
With respect to out-of-class study partners, my observation is that most instructors don’t know which students are studying together (and, therefore, which are studying alone), much less other salient information such as the scope, nature, degree, technical discourse, technological exchange medium, and directional value-proposition of the relationship.
Fortunately, network theory, and just as important, network practice and readily-accessible tools, are available to assist in this classroom analytical endeavor. Understanding class networks in various ways is a natural complement to existing learning analytics regarding student performance, such as those provided by important courseware firms such as Café Learn.
Students are facing an increasing number of large lecture hall classes and an increasing number of hybrid or online classes. Higher education students should be spending at least as much time studying out-of-class as they are for contact hours in-class, and the students live a networked life on social media. Most instructors would like to assume that their students:
- know to form study-partner networks without a faculty-imposed requirement
- can demonstrate skills that are attractive to potential study-partners and choose study-partners well
- are able to leverage the strengths in the study-partner relationship and implement interventions to overcome any deficits in the study-partner relationship
The simplest approach might be to simply ask the following question: “Do you have a study partner?” Fortunately, in contemporary academic life, instructors have access to more electronic tools to gather such data. Also, merely asking the study-partner question reinforces that idea that study-partners matter. Perhaps the best use of such network data is to combine that data with final class grade data at the end of the current class and use the combination of both as a motivational tool for the subsequent class. Most instructors teach the same course in a following, often immediately subsequent semester or quarter. Effectively, the results from the learning analytics for one class can serve to inspire and motivate students in the next class. If the process of collecting, analyzing, and presenting the study-partner network data can be automated in whole or in part, then the benefits can accrue to the students with only a small cost on the part of the instructor.
I teach all levels of business students from undergraduate freshmen to graduate students; but mostly, I teach undergraduate juniors. Additionally, I teach at a large, urban, comprehensive university that has nearly two-thirds of its students as transfer students. For an introductory course in Principles of Management and Organizational Behavior, I simply add the results of the study-partner learning analytics to the first-day motivational lecture for students. A class such as this involves a visceral understanding theory and practice of human motivation; this topic is precisely aligned with the idea of an optimal study-partner network.
The slide that I used most recently as part of my first-day motivation presentation is shown in Figure 1 below. I’ve been using basic study-partner results for more than five years in this course. Initially, slightly more than 7% of the students chose a study-partner (data not shown); currently, as can be seen from Figure 1, almost one-third of the students selected a study-partner (the data shown is from Fall, 2015). In all semesters, the difference in final grade results between the two groups (study-partner or not) has been statistically significant (the test shown in Figure 1 makes a small adjustment to account for slightly unequal variances across the two groups). The details on such a slide can be altered to suit the level of prerequisite statistical literacy in the class.
Perhaps more important is that this increase in study-partner participation is associated with a gradual reduction in the number of low grades (DUFs) in the course. Presenting the results from this simple study-partner network learning analytic has helped reduce the DUF rate (or rather, increase the persistence rate) by approximately half (again, data not shown). Clearly, these results do not come from a rigorous, control-based study using formal experimental design. There may be context-, course-, or instructor-specific factors involved, but that discussion is beyond the scope of this brief article. Results such as the ones presented here are indeed the real world of learning analytics. There are patterns and practices in learning networks that matter. Recall that the goal here is to help students, sometimes in small ways, to achieve success. Part of that process, at least for this Management course, is not to require study-partner teams but rather to recommend them. It is that self-starting, self-efficacious initiative that is very valuable in class. And digital natives like visuals!
In the Basic case, we ask one simple question and implement one simple idea, and we observe improvements in one or more desirable outcomes, especially with respect to helping students that may need the help the most. Some instructors see the benefit and may just continue the established intervention even if the marginal improvements are minor (i.e., they asymptotically approach diminishing returns because it becomes increasingly more difficult to substantively improve every possible DUF instance). Other instructors see the benefit too but wish to embrace and extend the study-partner networking model. Collecting more data regarding the study-partner relationship does this.
I occasionally collect the following pieces of data regarding the study-partner relationship beyond the Yes/No response as discussed previously. I’ve collected data on the number and names of other partners, who initiated the relationship, which week of the semester the relationship began, the frequency and nature of electronic exchanges, the frequency and nature of face-to-face exchanges, the balance of the direction of the bulk of the learning between or among study-partners, what types of coursework was studied for (e.g., written assignments, exercises, quizzes, exams, etc.), and a subjective sense of the overall value of the relationship. In addition to self-reported data from a typical survey instrument, some faculty have experimented with collecting student communication data, for example, in email or chat sessions, or more persuasively, in deliberate on-line discussion forums in a Learning Management System (LMS). Each element of exchange is effectively a different network. Additional attributes can be collected too including Sex, Major, GPA, Transfer Status?, etc. Note that study-partners within the same team will likely differ as to the overall value of the relationship so that relationships can also be directed as well as undirected.
A range of visual diagrams can be generated and used in various student-, faculty-, or institutional-contexts. These diagrams are referred to as sociograms. A simple sociogram is displayed in Figure 2. Note that two of teams have three individuals in them, and one team consists of four individuals (the student names have been anonymized). There are open-source software tools—such as R, Python, Gephi, and SocNetV—that generate these diagrams. Note that each additional piece of network data beyond Yes/No is essentially a different part of the study-partner network relationship. Whether all the data should be included, weighted, and displayed together, or whether specific analyses should be done for specific individuals (referred to as ego networks) is a non-trivial matter but one that can be discussed among colleagues.
Even more advanced network analyses are possible. Instructors (or institutions) could generate sociograms answering questions such as; which students took which courses or which course-sections together previously? Beyond exploratory analysis there is an entire range of confirmatory analyses for network data including comparing empirical observations with random graph model expectations.
This article has discussed how class networks can complement learning analytics in the contemporary higher education classroom. Specifically, the study-partner relationship was discussed but other key learning networks abound.
Blog post by: Wayne Smith, Ph.D., Department of Management, California State University, Northridge email@example.com
Professor Smith currently teaches business and management course at Cal State Northridge. Professor Smith has also taught at Cal State Channel Islands, UC Irvine, and Santa Monica City College, and earned a Ph.D. from the School of Information Systems and Technology at Claremont Graduate University in 2008. You can read more about his work on his website: OCW (OpenCourseWare). Readers are welcome to contact the author for additional details or with questions.