| Abstract |
It is our great pleasure to present the Proceedings of the Fourth Annual ACM Conference on Learning at Scale, L@S 2017, held on April 20-21 at the Massachusetts Institute of Technology, Cambridge, MA, USA. Learning at Scale investigates large-scale, technology-mediated learning environments. The conference was created by the Association for Computing Machinery (ACM), inspired by the emergence of Massive Open Online Courses (MOOCs) and the accompanying shift in thinking about education. While the conference was originally inspired by the emergence of MOOCs, large-scale technology mediated learning environments are very diverse and the conference series is a venue for discussion of the highest quality research on how learning and teaching can be transformed by that diversity of environments. Intelligent tutoring systems, open learning courseware, learning games, citizen science communities, collaborative programming communities, community tutorial systems, and the countless informal communities of learners are all examples of learning at scale. These systems either depend upon large numbers of learners, or they are enriched through use of data generated by the previous use of many learners. They share a common purpose--to increase human potential--and a common infrastructure of data and computation to enable learning at scale. Investigations of learning at scale naturally bring together two different research communities. Learning scientists are drawn to these innovative environments to study established and emerging forms of knowledge production, transfer, modeling, and co-creation. Computer scientists are drawn to the field as powerful site for the development and application of advanced computational techniques. At it's very best, the Learning at Scale community supports the interdisciplinary investigation of these important sites of learning and human development. The goal of the Learning at Scale community is the understanding and enhancement of human learning. In emerging education technology genres, researchers often use a variety of proxy measures for learning, including measures of participation, persistence, completion, satisfaction, and activity. In the early stages of investigating a technological genre, it is entirely appropriate to begin lines of research by investigating these proxy outcomes. As lines of research mature, however, it is important for the community of researchers to hold each other to increasingly high standards and expectations for directly investigating thoughtfully constructed measures of learning. In the early days of research on MOOCs, many researchers documented correlations between measures of activity (videos watched, forums posted, clicks) and outcome proxies including participation, persistence, and completion. As learning at scale research matures, studies that document these kinds of correlations should give way to studies of student learning that produce evidence of instructional techniques, technological infrastructures, learning habits, and experimental interventions that improve learning. As a community, we believe that that the very best of our early papers define a foundation to build upon but are not an established standard to which to aspire. The call for papers attracted submissions from all over the world, covering a broad range of topics from the theoretical to the pragmatic. We received submissions on a variety of topics including: novel assessments of learning, drawing on computational techniques for automated, peer, or humanassisted assessment; new methods for validating inferences about human learning from established measures, assessments, or proxies; experimental interventions in large-scale learning environments that show evidence of improved learning outcomes; domain independent interventions inspired by social psychology, behavioral economics, and related fields with the potential to benefit learners in diverse fields and disciplines; domain specific interventions inspired by discipline-based educational research that have the potential to advance teaching and learning of specific ideas, misconceptions, and theories within a field; tools or techniques for personalization and adaptation, based on log data, user modeling, or choice; usability studies and effectiveness studies of design elements for students or instructors; tools and pedagogy to promote community, support learning, or increase retention in at-scale environments; new tools and techniques for learning at scale; best practices in the archiving and reuse of learner data in safe, ethical ways; innovations in platforms for supporting learning at scale; and tools and techniques for managing privacy of learning data. In all topics, we encourage the use of best practices in open science, including pre-planning and pre-registration as well as a particular focus on contexts and populations that have been historically not well served. The overall submission numbers were greater than those of previous years. All papers were reviewed according to stringent criteria. Full Papers were reviewed by at least three program committee members, Work-In-Progress Papers and Demonstration Descriptions by two. Final decisions for acceptance of Full Papers were made by the program committee as a whole, often after extensive discussion of the merits of the paper. Whereas Full Papers present work that is innovative and mature, WiPs and Demos offer a forum for the newest and emerging work at earlier stages, offering pointers to future directions. As may be clear from the following analytics, the conference is highly selective in the work it accepts. |
