L@S investigates large-scale, technology-mediated learning environments that typically have many active learners and few experts on hand to guide their progress or respond to individual needs. Modern learning at scale typically draws on data at scale, collected from current learners and previous cohorts of learners over time. Large-scale learning environments are very diverse: evolving forms of massive open online courses, intelligent tutoring systems, open learning courseware, learning games, citizen science communities, collaborative programming communities (such as Scratch), community tutorial systems (such as StackOverflow), shared critique communities (such as DeviantArt), and countless informal communities of learners (such as the Explain It Like I’m Five sub-Reddit) are all examples of learning at scale. A growing number of current campus-based courses in popular fields also involve many learners, relative to the number of course staff, and leverage varying forms of data collection and automated support. All share a common purpose to increase human potential, leveraging data collection, data analysis, human interaction, and varying forms of computational assessment, adaptation and guidance.

Research on learning at scale naturally bring together two different research communities. Learning scientists are drawn to study established and emerging forms of knowledge development, transfer, modelling, and co-creation. Computer and data scientists are drawn to the specific and challenging needs for data collection, data sharing, analysis, computation, and interaction. The cornerstone of L@S is interdisciplinary research and progressive confluence toward more effective and varied future learning.

The L@S research community has become increasingly sophisticated, interdisciplinary and diverse. In the early years, researchers began by investigating proxy outcomes for learning, such as measures of participation, persistence, completion, satisfaction, and activity. Early MOOC researchers in particular documented correlations between easily observed measures of activity – videos watched, forums posted, clicks – and these outcome proxies. As the field and tools mature, however, we have increasing expectations for new and established measures of learning. As MOOCs morph into a more varied and provocative medium and L@S research expands, we aim for more direct measures of student learning, accompanied by generalizable insight around instructional techniques, technological infrastructures, learning habits, and experimental interventions that improve learning.