As online courses become more common, practitioners are in need of clear guidance on how to translate best educational practices into web-based instruction. Moreover, student engagement is a pressing concern in online courses, which often have high levels of dropout. Our goals in this work were to experimentally study routine instructional design choices and to measure the effects of these choices on students’ subjective experiences (engagement, mind wandering, and interest) in addition to objective learning outcomes. Using randomized controlled trials, we studied the effect of varying instructional activities (namely, assessment and a step-through interactive) on participants’ learning and subjective experiences in a lesson drawn from an online immunology course. Participants were recruited from Amazon Mechanical Turk. Results showed that participants were more likely to drop out when they were in conditions that included assessment. Moreover, assessment with minimal feedback (correct answers only) led to the lowest subjective ratings of any experimental condition. Some of the negative effects of assessment were mitigated by the addition of assessment explanations or a summary interactive. We were surprised to find no difference between the experimental conditions in learning outcomes, but we did find differences between groups in the accuracy of score predictions. Finally, prior knowledge and self-rated confusion were predictors of post-test scores. Using student behavior data from the same online immunology course, we corroborated the importance of assessment explanations. Our results have a clear implication for course developers: the addition of explanations to assessment questions is a simple way to improve online courses.
We report an experimental implementation of adaptive learning functionality in a self-paced HarvardX MOOC (massive open online course). In MOOCs there is need for evidence-based instructional designs that create the optimal conditions for learners, who come to the course with widely differing prior knowledge, skills and motivations. But users in such a course are free to explore the course materials in any order they deem fit and may drop out any time, and this makes it hard to predict the practical challenges of implementing adaptivity, as well as its effect, without experimentation. This study explored the technological feasibility and implications of adaptive functionality to course (re)design in the edX platform. Additionally, it aimed to establish the foundation for future study of adaptive functionality in MOOCs on learning outcomes, engagement and drop-out rates. Our preliminary findings suggest that the adaptivity of the kind we used leads to a higher efficiency of learning (without an adverse effect on learning outcomes, learners go through the course faster and attempt fewer problems, since the problems are served to them in a targeted way). Further research is needed to confirm these findings and explore additional possible effects.
In order to understand potential applications of collaborative problem-solving (CPS) assessment tasks, it is necessary to examine empirically the multifaceted student performance that may be distributed across collaboration methods and purposes of the assessment. Ideally, each student should be matched with various types of group members and must apply the skills in varied contexts and tasks. One solution to these assessment demands is to use computer-based (virtual) agents to serve as the collaborators in the interactions with students. This article proposes a human-to-agent (H-A) approach for formative CPS assessment and describes an international pilot study aimed to provide preliminary empirical findings on the use of H-A CPS assessment to inform collaborative learning. Overall, the findings showed promise in terms of using a H-A CPS assessment task as a formative tool for structuring effective groups in the context of CPS online learning.
Whiteboard animations are increasingly used in education despite little evidence of their efficacy. In this study, we measured the impact of whiteboard animations and other common instructional formats on learning outcomes, experience, and motivation. We recruited participants from Amazon’s Mechanical Turk (N=568; 326 females). Participants were randomly assigned to view online lessons about popular topics in social science from wellestablished scholars in one of five common instructional formats: whiteboard animation, electronic slideshow, stage lecture, audio, and text. Results showed a benefit of whiteboard animations in terms of learning and subjective experiences of enjoyment and engagement.
Text components of digital lessons and problems are often static: they are written once and too often not improved over time. This is true for both large text components like webpages and documents as well as the small components that form the building blocks of courses: explanations, hints, examples, discussion questions/answers, emails, study tips, motivational messages. This represents a missed opportunity, since it should be technologically straightforward to enhance learning by improving text, as instructors get new ideas and data is collected about what helps learning. We describe how instructors can use recent work (Williams, Kim, Rafferty, Maldonado, Gajos, Lasecki, & Heffernan, 2016a) to make text components into adaptive resources that semiautomatically improve over time, by combining crowdsourcing methods from human computer interaction (HCI) with algorithms from statistical machine learning that use data for optimization.