The Wall Street Journal published an opinion piece this week, “Higher Education’s Online Revolution,” by John E. Chubb and Terry M. Moe. The piece provided a balanced and insightful analysis, but whether online courses turn out to be a “revolution” or not remains to be seen. Delivering educational content to distant students is of course not a new idea. Correspondence courses have been around for decades, and while the delivery mechanism of online courses offers greater opportunities for self-pacing and feedback, the core concept is the same. There have been many attempts to commercialize computer-aided instruction over the years, but I think it is fair to say that so far none has really flowered.
To me, the key problem with online distance education is a human failing, not a technological failing. Students sign up for an online course with high expectations, but relatively few have the self discipline and habits of mind to stick it out. In a traditional residential college, there is strong social pressure to show up for class, turn in homework, and study for exams. Having peers to talk to and to study with keeps student interest from flagging. Interaction with a like-minded community helps fend off the other demands on one’s time. Remove these social incentives, and the odds of a typical student successfully completing a course drop precipitously.
The opinion piece mentions Stanford’s online artificial intelligence course, which has an enrollment of 150,000. At this scale, there is virtually no opportunity for a student to interact with a teacher or any other human being associated with the course. The authors don’t discuss Stanford’s course completion rate, but I can’t imagine the percentage is very high.
So, to me, the challenge is how to build into online courses a mechanism that socializes the experience for students and enhances their motivation to complete their studies. Here’s a practical idea.
Ask each enrollee to fill out an initial questionnaire that captures demographic information – age, gender, native language, employment, marital status, time zone, geographical location, educational background, extracurricular interests, and so forth, the goal being to create a snapshot of the student’s life and interests, analogous to the information admissions officers collect at traditional residential colleges. (Obviously, privacy issues would have to be carefully thought out.)
The school would then use this information to create a database about the profile of its enrollees. With this information, the school (or, more precisely, the school’s computers) could then create virtual classrooms of, say, 25 students, who have comparable backgrounds, interests, and goals. This virtual classroom would become the students’ personal social group for the course. Thus, a virtual class might consist of young working adults, or single mothers, or casual hobbyists, or PhD physicists – in other words, a group of traits that fosters a community that makes sense for the students. The algorithms for defining such a cohort would, with experience, be adjusted continuously to maximize course completion rates. If, for example, shared hobbies were found to have no impact on completion probabilities, then that field would be deleted from the database.
Once a virtual classroom was formed, students would then be given the opportunity to interact with their fellow students. Interaction could be via email, a Facebook-like page for each class, Skype-arranged conference calls, or whatever, all arranged automatically by computer. Students could ask each other questions, discuss course topics, debate, develop friendships, etc., all with the course being the common topic that binds them together. The computer could schedule regular “class meetings.” To preserve privacy, each student would be given the chance to opt in at various levels of involvement, e.g., first name only, personal information withheld, etc. After all, the goal is that classmates share traits only that contribute to course completion and interaction. Sharing personal information that goes beyond this level is not necessary.
One benefit of this model is that the results would be easily quantifiable. In a large course with thousands of enrollees, there would be many opportunities for experimentation. For example, the computer could create a virtual class of, say, 20-year-old working adults with a high school education, and benchmark their course completion rates against a similar cohort of enrollees who were not part of a virtual class.
As course providers refined their algorithms, improvements in completion rates could be directly quantified. In a for-profit environment, algorithms would be proprietary information, and those providers with the most successful algorithms would gain a significant competitive advantage over other providers. In this sense, the marketplace would resemble that for on-line dating services, where competing services vie with each to develop the most effective sorting strategies to match up their participants.
The key idea, however, is to use modern technology to create unique social networking sites tailored specifically to each student’s needs, in part to enhance student learning, and in part to provide the social and peer motivation essential for successful course completion. The enormous benefit of modern technology is that, having defined the algorithms and written the software, all of this can take place with minimal human intervention by the course providers.