Three years ago, UC San Diego computer science professor Pavel Pevzner predicted that today’s massive open online courses (MOOCs) would eventually give way to adaptive Intelligent Tutoring Systems (ITS). In doing so, they could also replace traditional classroom lectures to a large extent – by dramatically changing the role of the professor.
Benjamin Bloom, the giant of educational psychology, argued 30 years ago that the classroom is an inefficient way to teach because students learn the same material at different speeds. Pevzner agrees: “My feeling is that we are probably the last generation of professors who will use the traditional classroom as a teaching technology. Twenty years from now, packing students into a large classroom and keeping them there for an hour will probably be seen as cruel and unusual punishment.”
Lecturing to hundreds of students at the same time, argues Pevzner, is particularly objectionable when science, technology, engineering or math courses are involved. “Learning a complex idea is comparable to navigating a labyrinth,” he noted. “In a large classroom, once the student takes a wrong turn, he or she has limited opportunities to ask a question and get back on track. This results in a learning breakdown, or the inability for the student to progress further without individualized guidance. Rather than reproducing the impersonal experience of listening to a professor’s lecture in a large auditorium, online education should replicate the experience of receiving one-on-one tutoring in the professor’s office.”
Enter the first online advanced undergraduate course designed from the ground up as an adaptive Intelligent Tutoring System (ITS). Pevzner and colleagues are about to launch the adaptive bioinformatics course on edX.org, the leading open-source, nonprofit online learning destination. The course is open for enrollment and begins on October 15, when learners around the world will be able to enroll in what edX calls the “first adaptive system for a complex course with programming assignments.”
“We are thrilled to partner with UC San Diego to offer this innovative course to our over 13 million edX learners worldwide,” said Anant Agarwal, edX CEO and MIT professor. “The Introduction to Genomic Data Science course not only provides learners with access to content from leading experts in the field, it also showcases the latest in MOOC development technology, delivering a truly adaptive and personalized learning experience. We congratulate UC San Diego on this exciting milestone in online learning and we look forward to continuing our work together to increase access to high quality education for everyone, everywhere.”
The online course will lead students through individual learning paths. “It will also evaluate students at every stage of their learning to figure out how to maximize their performance and retention,” said Pevzner. “The goal is to represent each learner as a pathway through the tutoring system and to analyze those digital paths across thousands of students – producing an avalanche of data that can be used to continually adjust the coursework to meet the needs of all learners.”
Pevzner and his former student, Phillip Compeau -- now the Asst. Dept. Head for Education in the Computational Biology Department at Carnegie Mellon University (CMU) – have created eight bioinformatics MOOCs, accounting for more than 250,000 enrollments in the past three years.
Their previous online courses have been partially adaptive, and after each run of the courses, Pevzner and Compeau developed remedial modules heading off commonly occurring student issues based on new questions posed by students. As a result, the content keeps building for the benefit of future students. “The goal,” observed CMU’s Compeau, “is to set up a system that can be used both online and offline: in the online world, this system can mostly take our place as instructors; in the offline world, it allows us to drill down far deeper than a lecture-based class ever would dream into student understanding.”
As a result, Pevzner and Compeau have completely stopped lecturing in their bioinformatics classes. “We both now run a ‘flipped’ classroom, where students must watch short video modules and work through the Massive Adaptive Interactive Textbook (MAIT), which has integrated assessments, outside of class,” Compeau explained. “So when it’s time for class, we can supervise students working in groups on additional assessments and foster an environment that allows them to answer each other’s individual questions.”
Pevzner argues that teaching should be intimate, either one-on-one between student and professor, or between student and student. The solution to packed classrooms, he believes, is to create the illusion of one-on-one interaction via new technology capable of guiding each student through an individualized learning journey through the material.
“Computer science is a special discipline because it has a unique way to test students’ knowledge,” observed Pevzner. “My teaching assistants don’t spend a minute checking programming-based homework because the machine checks it automatically. As a result, they have more time to interact with students.”
“Our bioinformatics MAIT used in the edX course can prevent individual learning breakdowns, and even outperform a professor,” argued Compeau. “We will now be able to compare learning outcomes between more traditional MOOCs with this adaptive ITS.”
The effort expended to create an adaptive course is nontrivial. By their own assessment, Pevzner and Compeau estimate that they had to create five times more content than required for a printed textbook in order to take into account all learning breakdowns that students might encounter. “Much of the extra content grew out of our curation of 8,500 posts on the discussion forums of just the first run of our Bioinformatics MOOCs,” explained Compeau. “Imagine an editor who would provide such a gold mine of information on how your students actually learn your material.”
“Producing a traditional MOOC may take hundreds of hours, even if much of the content involves videotaped lectures in a classroom,” agreed Pevzner. “For our adaptive bioinformatics course, however, our team has invested close to 7,000 hours to generate the extra content needed to make a course truly adaptive.”
He added: “The courses of the future will be very polished works, not unlike modern movies, and creating a top-notch adaptive online course may easily cost $1 million.”
Pevzner literally wrote the book on bioinformatics. Published in 2004 by MIT Press, An Introduction to Bioinformatics Algorithms (co-written by Neil Jones) quickly became a standard reference in more than 100 universities worldwide. By 2012, however, Pevzner realized that the book had become outdated given the rapid growth of bioinformatics. Rather than simply create a new edition that would soon be outdated, Pevzner realized that he’d rather invest in a web-based “MOOCbook” that could be improved continually over a period of years based on the feedback of thousands of online students.
For this project UC San Diego is partnering with Stepik, an edtech startup in Cambridge, MA. The adaptive learning system on Stepik is designed for continual updating, and it has built-in detours and turning points as well as quizzes to gauge whether the individual student is understanding the content. If a student answers incorrectly, he or she is pointed to a refresher page with the missing knowledge. Compeau and Pevzner worked closely with Stepic CEO Nikolay Vyahhi and Mike Rayko at UCSD to adopt Stepic for their adaptive course.
Automated, individualized assessments built on the Stepik platform are meant to replace the rudimentary multiple-choice quizzes that pass for assessment modules in most existing MOOCs. “When a student suffers a learning breakdown, that student needs immediate help in order to proceed,” said CMU’s Compeau. “Quizzes can be replaced with robust assignments that require a sophisticated software system that automatically grades programming challenges in bioinformatics.”
The course provides hundreds of “just in time” exercises and coding challenges to assess each student’s progress. Each paragraph of the interactive text is linked to a separate discussion thread, which facilitates peer learning between students who encounter learning breakdowns at the same point in the class.
By creating an adaptive course, Pevzner says the edX course will circumvent one of the biggest problems facing most MOOCs. “If you’re dealing with video modules and you want to change anything about the content, it means going back into the studio to record an updated module,” he explained. “As a result, MOOCs rarely change – even if the professor has feedback from hundreds of thousands of students who have taken the course. Adaptive courses, on the other hand, can be easily modified online if you find that students are breaking down in a place you had not anticipated.”
Most of Pevzner’s existing MOOCs are partially adaptive because they share access to his online adaptive textbook, and more than 60 universities in 20 countries have adopted the content developed for his online courses for use in their classrooms. “The online revolution in education is still at an early stage,” said Pevzner, “but it is also changing the way courses are taught offline in the classroom.”
With the launch of the adaptive course next month and his ‘flipped’ approach to teaching in the classroom, Pevzner hopes to prove that the adaptive ITS will finally deliver on the promise of online graduate education that remains illusory despite the proliferation of traditional MOOCs available to learners around the world.