8341639815_b7864c3743_z

Image courtesy palphy/Flickr

As the school year gets into full swing, folks might be starting to think about how MOOCs (massive online open courses) can help them on their own educational journeys – whether towards a degree or simply for growing their own knowledge for personal or career reasons. After a meteoric rise, MOOCs such as those offered by CourseraEdX and Udacity, have taken a few hits. Early results from a study last year by the University of Pennsylvania, for instance, said that MOOC course completion rates average just 4 percent across all courses, and range from 2 to 14 percent depending on the course and measurement of completion. The New York Times reported on some other setbacks here – but also noted that while MOOCs may be reshaped, they’re unlikely to disappear.

Some of that reshaping is underway. Among the efforts is a project announced this summer to take place at Carnegie Mellon University, in a multi-year program funded through a Google Focused Research Award. The announcement says the project will approach the problem from multiple directions, including a data-driven effort that will use machine-learning techniques to personalize the MOOC learning experience.

“Computer programs will evaluate each student’s work, identifying subject matter that has been mastered and areas where additional study or different types of exercises could be beneficial,” the release noted.

More for the MOOC

Recent months also saw MIT’s Computer Science and Artificial Intelligence Lab working to improve MOOC course videos (and other video learning experiences) with the help of a platform it developed called LectureScape – data-driven interaction techniques to improve navigation of online videos. The platform was developed by analyzing collective and individual learners’ click behaviors related to video content, and offers users alternative timelines, search interfaces and automated summaries to help emphasize interaction peaks; visualize and rank occurrences of search terms on the timeline and recommend salient keywords in each video section; and capture the frames frequently viewed by others.

To improve searching within the video, for example, it generates visual highlights that point to the frames frequently watched by other viewers, while a word cloud at the top of the video displays automatically-extracted topics for a particular section. “To support novice learners who do not necessarily have the vocabulary to translate their information needs into a direct search query, we suggest major topics discussed in each section of the video in a word cloud,” the researchers write in a paper about the platform.

A pinning algorithm also pins relevant slides relating to previously introduced concepts and formulas so that the learner can refer to them even after the course video lecture no longer displays those slides onscreen.The effort was funded by edX, the online learning platform from MIT and Harvard.

Meanwhile, The Bill And Melinda Gates-funded MOOC Research Hub earlier this summer published nearly two dozen reports (including the final version of the aforementioned University of Pennsylvania study) exploring the topic from a number of perspectives. That includes one on Developing Data Standards and Technology Enablers for MOOC data science, based on the MOOCDB data standard description that harmonizes and unifies raw data streams from various MOOC platforms in order to facilitate intra-platform and inter-platform collaboration among the MOOC education science community.

The concept envisions analysts as referencing the data to “visualize, descriptively analyze, use machine learning or otherwise interpret some set of data within the database. Analysts extract, condition…and create higher level variable for modeling and other purposes from the data.”