By Cait Etherington May 19, 2017
You can now take courses for free from some of the world’s top-ranked universities. Whether you’re searching for a course in biochemistry, book history or botany, you are bound to find a MOOC that meets your needs and in many cases, the MOOC will be led by an internationally renowned faculty member. Of course, this raises a critical question: Why are some of the world’s top universities, including those that typically charge over $50,000 annually in tuition, willing to give away their courses for free? While there are many reasons why universities continue to invest in MOOCs, big data and its potential contribution to learning science remains the main force driving the ongoing institutional enthusiasm for MOOCs.
For many years, psychologists and educational researchers struggled to fully understand what makes students tick and to understand the impact of different pedagogical approaches in context. Why do some students excel while others fall behind? Why do some students work well in groups and others work better alone? How long do students actually spend completing homework assignments? Do students really read assigned articles? Do skimmers do more poorly on tests than students who read texts thoroughly? Moreover, do tests have a quantifiable impact on one’s knowledge? What alternative evaluation methods are effective? In the past, answering these questions was difficult, if not impossible, for at least two reasons. First, educational researchers were often forced to rely on subjective methods of inquiry (e.g., focus groups, surveys, and interviews with learners). Second, since Institutional Research Boards generally consider students to be a “captive population,” gaining access to students, especially large groups of students, for research purposes was difficult. Then, big data presented educational researchers with a new way to explore how, why and when students learn and this, rather than accessibility, made MOOCs of interest to a some of the world’s top universities.
As reported back in 2012, when Harvard University first agreed to participate in MIT’s already established MITx project, big data and its potential insights on learning were a critical component. As Harvard University President Drew Faust emphasized at the time, by embracing MOOCs, “We will learn more about learning. We will refine proven teaching methods and develop new approaches that take full advantage of established and emerging technologies.” On this basis, MIT President Susan Hockfield characterized online education not as an “enemy of residential education, but instead a powerful and inspiring ally.”
For universities, MOOCs may not bring in income, but they do bring in something just as and perhaps even more important. By generating data rather than income (e.g., from tuition), MOOCs offer universities something they have previously not been able to access: Metrics about what actually makes instruction effective.
To illustrate, consider the following example from a course at MIT. A physics professor at MIT ran an edX version of his undergraduate quantum physics course. The course attracted a relatively large number of participants (approximately 2000). Simultaneous to running the MOOC, the course was offered to 18 users on campus. While both groups of students struggled with the questions, there was one notable difference. The professor found that in some cases, his questions were too difficult for the level or had other problems (e.g., unclear language). With 18 on-campus students, he may have assumed that the troubles were simply about the caliber of the students in his class. The MOOC, with a student sample 100x larger than his on campus class, provided the statistical power needed to identify and resolve item-level difficulties. As a result, some of the problems on the course were revised, and in the end, the professor was able to produce a higher-quality and fairer on-campus version of his course.
Of course, universities are not simply using the data mined from MOOCs to refine questions on exams. As data accumulates, it seems likely that MOOCs will increasingly be used to test-run large-scale courses to determine other factors (e.g., the choice of readings included on a course) or types of evaluation methods. Big data will also increasingly be used to automate some tasks once carried out by teaching faculty (e.g., responses to routine student questions about assignments or course content).
However, the road ahead may not be as transparent as many institutions assume. As Candace Thille, a pioneer in big data-based learning solutions who now works in Stanford University’s Graduate School of Education, warns, “Go ahead, experiment with some of these systems. Just don’t get locked in for a long-term contract and develop your whole approach around a particular system, because it’s going to change, or it should change.” Thille’s primary concern is that as MOOCs proliferate with a large number of private companies taking the lead, the true educational benefits of the marriage between big data and learning may get lost along the way. Whatever happens in the future, it seems likely that MOOCs are here to stay not because they will ever replace residential learning but rather because they hold the potential to finally pry open the mystery of learning itself.