Two years ago, Sebastian Thrun, a computer science researcher at Stanford, claimed that in fifty years there would be only ten institutions in the world delivering higher education.1 That’s a pretty ambitious statement, given that there are over 9000 universities across the globe today! Thrun’s confidence in his claim stemmed from his recent work in morphing traditional college lectures into Massive Open Online Courses, also known as MOOCs. Instead of restricting knowledge to the privileged few, these courses aimed to bring the highest levels of teaching and scholarship to students everywhere. In a way, Thrun succeeded: instead of his course on artificial intelligence reaching around 200 university students, he was able to digitally distribute lectures and assignments to 160,000 students across hundreds of countries.
Recently, however, MOOCs have begun falling out of favor. Critics have raised questions about the quality of online classes, and several universities which had originally jumped on the bandwagon have quietly cancelled their plans to move courses online. Thrun now states that “we were on the front pages of newspapers and magazines, and at the same time, I was realizing, we don’t educate people as others wished, or as I wished. We have a lousy product.”2 What caused this abrupt change of opinion? Various academics and teachers have placed blame on a variety of issues, such as high dropout rates and a lack of diversity in student populations, but we can look for answers ourselves by examining the online education of the past.
Online Education: Not So Novel!
Online educational resources have been around for almost as long as computers themselves; in fact, as the graphic above shows, the first online course was offered in 1968, before computer networks had been established outside of universities and military research. More recently, many teachers have experimented with contributing their materials to open databases online and adding online instruction to their in-person classes. These teachers have run many studies showing that students can learn just as well online as in person; in fact, the US government ran a meta-analysis of the research3 in 2010 (two years before MOOCs!) which showed that students learn better when online resources are used. There have even been online sites providing open university-level courses since the early 2000’s, such as MIT’s OpenCourseWare and CMU’s Open Learning Initiative.
The main difference between these sites and the new MOOCs seems to be the “massive” part of the name. Perhaps there’s something inherently exciting or motivating about the prospect of taking a class at the same time as thousands of other people. In evaluating what makes a MOOC a success or a failure, then, we primarily need to study the diverse range of students taking the course and how they are affected by the larger scope.
One benefit of the massive scale that MOOCs work in is that it lets us use methods from the field of “big data,” where very large corpuses of logged data (such as when students log in, look at lectures, and submit assignments) are statistically analyzed to find deep trends and relations. The research that’s already been done has investigated a range of research questions, including how students benefit from talking to each other in online forums,4 how the design of lecture videos impacts how students watch them,5 and how students progress through a variety of different courses.6 One of the biggest outstanding questions, however, has been about the students themselves – where do they come from, what are their backgrounds, and what are they getting out of the course?
This brings us to one of the main controversies around MOOCs: the high dropout rates. Reports have shown that, on average, around 90% of students who sign up for a MOOC do not complete the course. This sounds shockingly high, but recent analysis of student data has shown that it might not be such a cause for concern. Phil Hill developed a set of student patterns (shown at right) classifying the different types of students who sign up for MOOCs, based on data collected from various courses. He and other researchers have found that the vast majority of students (60-80%) will leave the course within two weeks of signing up.7 Perhaps these students are just lurkers – internet users who are mildly interested in finding out what the course will cover with no real intention of committing to completing all the work. If this is the case, then the high dropout rates may not be negative after all.
While use of big data can tell researchers a great deal about the types of students who participate in the courses and how they participate, it is harder to show whether the students are making significant learning gains from beginning to end of the course; after all, the students might enter the course knowing the material already, or they might be completing the assignments without retaining any of the information. Because MOOCs are still so young, there has been little experimentation within them at this point, though new results are coming out all the time. In the near future, perhaps we’ll see more work that follows traditional experimental setups, like the randomized control trial (which randomly assigns students into a variety of experimental conditions) or pre- and post-tests (which measure the change in knowledge that occurs in specific areas with targeted questions assigned before and after the experiment). Through these experiments, we’ll be able to assess the effectiveness of our new massive online courses through hard data, which will hopefully lead us to findings about how to improve all courses.
So, if we haven’t been able to gather hard data to find an answer yet, what caused that abrupt change in public opinion about MOOCs? My theory is that the innovators who started the MOOC revolution reviewed their attempts to teach thousands of students and realized what I too have found out after years of learning about what causes learning: education and the learning sciences are messy, intractable, and extremely complicated fields. There is no silver bullet that can immediately improve the learning of all people; there’s only an immense collection of factors which influence how an individual learns. That’s a very frustrating thing to hear when you want to make the world a better place! But the cause of improving education is not hopeless: by researching how these individual factors affect learning across a diverse set of people, all of us can make small changes that may incrementally improve education for the multitudes.
1. The Stanford Education Experiment Could Change Higher Learning Forever 2. Udacity’s Sebatian Thrun, Godfather of Free Online Education, Changes Course 3. Evaluation of Evidence-Based Practices in Online Learning: A Meta-Analysis and Review of Online Learning Studies 4. Development of a Framework to Classify MOOC Discussion Forum Posts: Methodology and Challenges 5. Understanding in-video dropouts and interaction peaks in online lecture videos 6. The Life Cycle of a Million MOOC Users 7. Emerging Student Patterns in MOOCs: A Graphical View