Identifying MOOC Learners on Social Media Platforms

We start the first paper session at WebSci 2016 with a paper by Guanliang Chen that examines learner engagement with Massively Open Online Courses (MOOCs). These generate a great deal of data about learner engagement during the MOOC itself, but there's very little information about learners before and after this experience. Can we use external social Web data to identify and profile these learners, in order to better customise the learning experience for them?

The biggest challenge in this is to identify the learners across diverse social media platforms. There are some possible approaches: explicit matching based on public email addresses; direct matching through cross-profile links between different social media platforms; and fuzzy matching of learners' login and full names against usernames on social Web platforms. The present paper drew on Gravatar, Twitter, LinkedIn, StackExchange, and GitHub as sources of social profile data, along with data from 18 MOOCs operated by TU Delft. Matching percentages using these approaches and data sources ranged between 1% and 42%; overall, some 5% of TU Delft learners could be identified on social media platforms.

Machine learning techniques were then used to predict leaners' demographics, focussing especially on Twitter profile information. The patterns found for different MOOC courses appeared to match the expected demographics of learners in each field. The study also explored the StackOverflow questioning and answering behaviours of learners before and after their MOOC participation. Using GitHub, it examined the extent to which learners transferred their acquired knowledge into practice.

So, on average 5% of learners could be identified on the five platforms; learners with specific social media traits preferred different types of MOOCs; their post-course behaviours appeared to differ from what they did before.