MOOC Research

This proposal bridges research on MOOCs and online education with research in the disciplines of cognitive science and psychology, including both theoretical insights about motivation and learning, and methodological expertise in computational modeling and experimental design. The proposal focuses on how to improve student motivation when learning from online exercises, by providing messages that encourage students to believe intelligence is malleable, and therefore work harder. An experimental comparison that ran on Khan Academy’s platform inserted different kinds of motivational messages into online mathematics exercises accompanied by worked-out example solutions, in order to increase student engagement and learning. We propose to conduct in-depth statistical modeling and analysis of this data set to statistically confirm and characterize what these precise learning benefits are, how they are influenced by the kind of message provided, and identify potential moderator and mediational variables. We also propose a separate line of experiments that will be conducted as laboratory-style online experiments (using convenience samples of participants from the general Internet population). Since such experimental paradigms have complementary affordances to the design and data collection features of Khan Academy’s platform, these experiments will also allow us to easily investigate novel questions about the effects of motivational messages, such as the impact of providing such messages *after* incorrect responses are made, in order to directly motivate students who have made mistakes.

Students in Massive Open Online Courses (MOOCs) exhibit a remarkable diversity in terms of location, gender, demographics, and educational attainment level; they have different learning goals and patterns of interaction with MOOC content. In this study, we aim to develop clustering of different MOOC learners in order to enable customization of learning experiences. We intend to 1) classify student learners by learning goals; 2) cluster learners by engagement with the platform, comparing various groups by learning outcomes (i.e. certificate attainment), and aiming to predict user transition from one cluster to another; 3) study how the clustering could be used for platform customization and personalization of learning experience. This research will proceed in the context of HarvardX, which is Harvard’s division for online learning. Analysis will be based on the data on 17 HarvardX courses running on the edX platform, focusing on 5 courses that will be completed by December 2013.

The Friday Institute for Educational Innovation has recently created the Massively Open Online Course for Educators (MOOC-Ed) Initiative to explore the potential of scaling and delivering personalized, high-quality professional development. Among the core principles of MOOC-Eds are collaboration and peer-supported learning. This study will leverage the Friday Institute’s research expertise in online communities of practice and the rich social network data generated by MOOC-Ed participants to better understand networked learning. Specifically, this multiple-case study will use a mixed-method approach that combines Social Network Analysis (SNA), advanced network modeling techniques, and in-depth qualitative analysis to examine the development of social networks, mechanisms governing their structure, and their impact on learning. Findings from this study will contribute insight into the knowledge MOOC researchers may gain by using the social network perspective as a lens, as well as the potential of social network analytics to aid practitioners in assessing and improving peer-supported learning environments in MOOCs.

Much attention has focused on the impressive number of individuals who are enrolling in MOOCs. But, to date, little is known about the life cycle of these enrollments, including the pattern of persistence beyond initial enrollment to course completion. Without this knowledge, it is not clear how to productively assess access or success overall or for different groups or how to understand the forces that influence student progress through these courses. This project is designed to address this knowledge gap by producing a map of the basic ecology of the first-generation of MOOC enrollments and identifying key user-transition points. This project will produce a map of the basic ecology of the initial wave of MOOCs by studying the progression of users through the 17 Coursera courses that were offered for the first-time by the University of Pennsylvania between June 2012 and June 2013. The analyses will also explore differences in the patterns of progression through the identified transition points based on various attributes of Coursera course offerings. Potentially relevant course dimensions include course length, course requirements, course content, and more. Establishing a framework for describing and monitoring user behavior is a necessary building block for any empirical examination of the role of MOOCs in increasing access to and efficiency of higher education. The results of this descriptive but foundational research may also inform the design of future MOOC offerings, guide the revision of existing courses, help institutions assess the viability of MOOC courses, and inform user-selection of courses.

The image many have of the introductory science course is of a giant lecture hall filled with students watching from bleacher seats. While more engaging blended formats have shown tremendous gains in learning and student engagement, there are several key barriers to wider-spread adoption. First, the initial cost of creation of such courses can be tremendous. Second, many teachers are more comfortable in traditional formats, and find it difficult and time consuming to adapt to more modern formats. Third, with the abundance of materials and approaches available it can be difficult to sift through them to find the best possible resources. Our goal is run a small seminar/cMOOC-style on-line course for instructors of physics in which instructors will explore blended formats, physics education research, pedagogical content knowledge, and other relevant domains. As part of this instructor course, participants will work together in small groups to create, share and review videos, assessments, and other educational resources which they will then be able to use in their own classrooms. These resources will be assembled into a broader MOOC for students desiring sufficient skill in introductory physics to obtain advanced standing for college. We are interested in exploring both the potential of a community contributed/driven MOOC design and in evaluating the quality of the materials derived from it.

MOOCs fail to meet their full potential of “educating the world” (Pappano, 2012) when they fail to support students who underperform or drop out due to a fixed mindset (Dweck, 2007) – the mistaken and self-limiting belief that their intelligence is unchangeable. Psychologists have developed short interventions to both influence mindset and improve learning outcomes, but they lack detailed evidence of the specific behaviors that bridge the gap between the two. I propose to use tools from learning analytics to develop a “behavioral mindset metric” that can link self-reported mindset with gains in assessment scores and retention. If successful, this work will improve the learning outcomes of disengaged students, provide researchers with a better understanding of growth-mindset behaviors, and demonstrate the utility of learning analytics for assessing non-cognitive competencies.

This study explores the role of MOOCs in supporting and enabling professional learning, or learning for work. The study is contextualised within ‘Fundamentals of clinical trials’, a MOOC for health professionals designed and run by the Harvard Medical School, Harvard School of Public Health, and Harvard Catalyst, the Harvard Clinical and Translational Science Center, and offered by edX. The research is informed by contemporary theories of professional learning, which argue that conventional forms of learning are no longer effective in knowledge intensive domains. As work roles evolve and learning for work becomes continual and personalised, self-regulation is becoming a critical element of professional learning. Yet, established forms of professional learning generally have not taken advantage of the affordances of social, semantic technologies to support self-regulated learning. MOOCs present a potentially useful approach to professional learning that may be designed to encourage self-regulated learning (SRL).
The study addresses three key research questions: RQ1: How are Massive Open Online Courses currently designed to support SRL? RQ2 What SRL strategies and behaviours do professionals adopt? and RQ3 How can MOOCs be designed to encourage professionals to self-regulate their learning? The study will produce academic publications and new datasets and instruments as well as project reports and recommendations made available to the MOOC Research community and the wider community of MOOC designers, developers, providers and practitioners. Data from another MOOC, run by Kings College London, will be collected in a parallel project using the same instruments.
Follow the PL-MOOC project at: http://www.gcu.ac.uk/academy/pl-mooc/

Social learning theorists and educational constructivists have long proposed a central role for peer-to-peer interaction in learning, particularly in adult professional development (Brown, Collins & Duguid, 1989; Wenger, 1998, for example). This mixed methods study combines a quantitative analysis based on learner analytics and social network analysis (SNA) with surveys and interviews to explore and map peer-to-peer interactivity, its structure and segmentation within a large MOOC learning community. SNA is used to identify individuals highly central and influential within communication exchanges for a purposeful selection of participants in subsequent surveys and interviews. The purpose of this study is to describe the dynamics of peer-to-peer interaction in ad hoc, virtual learning activity in MOOCs, to build on existing theory regarding learning team productivity and cohesion, and to monitor and explore the behavior of key individuals, around whom ad hoc learning groups form online.

This research project will explore the prominent critiques and celebrations of MOOCs and their potential impact on existing educational systems. These issues will be addressed through a systematic discourse analysis of the popular debates surrounding MOOCs between September 2010 and September 2013. The project will take a discourse analysis approach to investigating these questions – drawing on established methodologies from linguistic studies and the social sciences. A large-scale corpus of text will be collated from popular and education-specific news media as well as the blogosphere. A combination of frame analysis and critical discourse analysis will provide a rich and rigorous account of what has been an important period in the recent history of educational technology – highlighting broader societal struggles over education and digital technology and capturing a significant moment before these debates subside with the normalization and assimilation of MOOCs into educational practice.

The nature of connectivist style MOOCs, where knowledge is created and shared within a distributed network, creates challenges in applying Learning Analytics techniques, primarily because activity data resides in a number of disparate systems. Unlike predominantly closed environments, such as Coursera, where researchers have already detected learner subpopulations, in open systems there is a likelihood that students are ‘analytically cloaked’, using different unique identifiers across multiple networks. This creates new challenges in detecting subpopulations, the detection of which might be beneficial in analyse patterns of engagement or used to identify active nodes to the benefit of tutors and students.
We propose a post hoc/retrospective investigation into learner subpopulation detection within the connectivist styled ‘Open Course in Technology Enhanced Learning (ocTEL)’ which ran between April and June 2013. Using free and open source tools we will attempt to resolve activity data from multiple sources to permit the analysis of any engagement patterns. Once any prototypical engagement populations are detected it is planned to apply these to similar connectivist MOOCs.

Massive Open Online Communities (MOOCs) are seen as an opportunity to gain an education; to gain new skills to prepare for high-paying jobs; and to gain upward mobility without incurring the increasingly crushing debt that comes with a university degree. Although MOOCs are seen as one possible path toward upward mobility, few studies have examined whether and/or how the populations with the most to gain leverage these resources. Our proposed research will investigate if MOOCs can help economically disadvantaged populations build the skills necessary to find employment and to achieve upward mobility. We will conduct a preliminary investigation of the experience of learners motivated to take MOOCs because of their inability to afford higher education; our goal is to better understand whether and how MOOCs can be a pathway to employment for these populations. Specifically, we propose to: 1) conduct an analysis of current University of Michigan (UM) Coursera-based course data to determine the number of these students enrolled in our courses; 2) provide an evaluation of student activities and performance in those courses; and 3) initiate interviews with targeted students to better understand their experience. Through these activities, we will begin to understand the ways in which the MOOC experiences may affect employment or potential employment for students who are not currently well served by more traditional forms of higher education.

Massive Open Online Courses (MOOCs) have the opportunity to significantly impact how students choose to engage with traditional institutions of higher education. This proposal seeks to shed light on the relationship MOOC students have with traditional degree-granting programs. In particular, we are interested in quantifying performance and patterns of usage of MOOC resources for those students who choose to enroll in or disenroll from traditional degree granting programs, or have previously graduated from traditional degree programs from the same institution offering the MOOC. We will do this through a data-driven investigation, looking at the natural groups of learners presented by historical usage data and customizing our survey instruments and methodologies as appropriate.

This project studies data gathered from the first Massive Open Online Course (MOOC) offered within the University of Wisconsin System (UW System). The course is titled the UW System College Readiness Math MOOC and is designed to help students gain or strengthen the mathematics skills needed in entry level college mathematics and science courses, helping to provide access to college and to avoid the need to take remedial (or developmental) math courses while in college. The demographics of the students in the course is very broad. Students from every state and from nearly 40 countries participated, and the ages ranged from 11 to 85 years. The study is analyzing data gathered from the MOOC using new tools and adapting existing tools to study a number of research questions. The student demographics is compared with successes, the activities that the students spent time on are analysed, and the value of each course component is being assessed. The research includes a study of the student experiences, learner outcomes, the learning design, and performance metrics through the use of learner analytics tools that are under construction. 

This project will examine MOOCs from two perspectives. The first is to use the open data available for a range of MOOCs to investigate the factors that relate to completion rate, including: enrolment numbers, MOOC provider, university reputation, course length, assessment, learner activity and gender. The data from 221 MOOCs has been gathered across a range of providers and MOOC types. This work builds upon the initial analysis of completion rates developed by Katy Jordan atkatyjordan.com/MOOCproject.html. This quantitative analysis is combined with a qualitative one focusing on a subset of MOOCs which represent the distributed (MOOCs that use a variety of tools) and centralised approaches (ones that are based on a central MOOC platform). This analysis uses two Learning Design tools developed at the Open University. The first of these, the Module Map details the type of resources used, and their main function within the course. The Activity Planner maps the types of activities learners are required to undertake across six pedagogic categories. The representations from the learning design tools will be shared at: bit.ly/ldmooc

As MOOC education researchers advance to analyzing cross-course data and broaden their single course analysis, the community requires a general organization for MOOC data which serves multiple specific uses in these contexts. The MOOCdb project centralizes and generalizes MOOC data organization and supports shared general purpose analytics for MOOC education research. This project will develop open-source, shared technology enablers for MOOCdb. MOOCdb’s schema definition will be broadened so it can accommodate Stanford U. data collected from the Coursera platform. A research MOOC data archive from multiple offerings of 2 MITx courses that were taught using the edX platform will be curated and a web based platform for sharing analytic visualization tools that exploit the MOOCdb schema will be developed.

The Behavioural Economics MOOC offered by the University of Toronto on the edX platform will be taken by a group of 32 secondary school students from the University of Toronto Schools who are studying grade 12 Economics.
The class will consist of students who are 15-17 years of age, and the course they are enrolled in is a university preparatory Economics course. The focus of the research will be to compare secondary school student achievement of
learning outcomes and levels of student engagement and persistence under two models of instruction:
a. through independent engagement with a MOOC
b. through a blended model involving teacher support and engagement with a MOOC

MOOCs are poised to transform access to higher education for millions of people worldwide. From a pedagogical standpoint, a central question surrounding MOOCs involves how a reliance on peer interaction and review impacts student learning. The main objective of this study is to evaluate how peer-to-peer interactions through writing impact student learning in two introductory-level Coursera MOOCs offered through Duke University: one in the humanities (English Composition I) and one in the natural sciences (Introduction to Chemistry). We will collect and analyze student writing from five different areas of our MOOC platforms: peer review, discussion forum posts, assignments, student self-reflections in the quiz section, and class generated wikis. We will use this data to measure learning gains against course learning objectives and analyze data about student retention and success in each course. We are particularly interested in how peer-to-peer interactions through writing in introductory-level MOOCs impact the learning outcomes for less academically advanced and/or low-income learners. This research will address several key areas of interest posed by the MOOC Research Initiative: Student experiences and outcomes, Learning design, and Learner analytics.

This study will compare the use of Massive Open Online Courses (MOOCs) as active, instructor-led, open-facilitated courses with their use as archived, self-directed learning resources. Having offered three popular introductory MOOCs in computer programming and statistics, we have discovered that after an open facilitated course cohort has progressed through scaffolded activities and graded assessment, many learners persist in the course and new registrants continue to join the archived course. This study will investigate the potential and purpose of archived MOOCs as learning tools for beyond and between scheduled instructor-led sessions. In order to understand potential differentiated usage patterns and learning outcomes, learner demographics, motivations, activities and completion, and levels of satisfaction will be examined and compared across the two models of content delivery.

The landscape of online learning has evolved in a synchronous fashion with the development of the every-growing repertoire of technologies, especially with the recent addition of Massive Online Open Courses (MOOCs). Numerous past circumscriptions associated with learning and teaching bounded by time and space dissolved accordingly, so does the parameters of the definition of “online learners”.
The proposed research here is a preliminary attempt to study how different motivational aspects of MOOC learners correlate with course completion rates by conducting a pre-course survey. The primary objective of this proposed project is to explore how various motivational aspects of MOOC learners corresponds with degrees of course completion. In the long run, research results expect to inform future interventions, research and design of MOOCs and reflecting potential emergent needs of MOOC learners.
Given the heterogeneity of motivational aspects of MOOC learners and the observed prevalence of emphasis on course completion rate, this proposed research intends to expand our understanding of MOOC learners by analyzing how MOOC learners’ motivation correlates with degrees of course completion. Specifically, two categories of motivational aspects will be taken into account in this initial research attempt. Both MOOC-specific motivational items including those tested by existing MOOC studies (e.g., Belanger & Thornton, 2013; MOOC @ Edinburgh, 2013) and two subscales of the PALS survey (Midgley, et al., 2000) measuring mastery goal orientation and academic efficacy will be included in a pre-course survey. The MOOC-specific items include questions such as the familiarity of the MOOC environment and course content; whereas the PALS subscales focus on learner orientations regardless of the learning context.
Methods including learning analytics and educational data mining techniques will also be applied following analysis of survey results accordingly. Specifically, drawing from past research in monitoring online learning and community (Dawson, 2006; Ming & Ming, 2012), this project will monitor course participation including discussion forums, quiz completion rate as well as correction ratio. In addition, learner’s video engagement activities will be coded and recorded. All the above-mentioned data collected will thereafter be linked to the MOOC survey employing FDR post-hoc correction and correction mining methods. Patterns of changes in participation will also be monitored throughout the course via questionnaire responses by means of sequential pattern mining. A similar paradigm drawn from Bowers’ (2012) study predicting high school dropout from grades and Romero’ (2011) work in predicting college dropout will be analyzed and applied.

One important hurdle that prevents MOOCs from reaching their transformative potential is that they fail to provide the kind of social environment that is conducive to sustained engagement and learning, especially as students arrive in waves to these online learning communities. The unique developmental history of MOOCs creates challenges that require insight into the inner-workings of massive scale social interaction in order to meet. Early attempts to organize the community into smaller study groups may be thwarted by such periodic growth spurts paired with attrition, as groups that initially had an appropriate critical mass soon fall below that level and then are unable to support the needs of remaining students. This proposed research seeks to lay the foundation for meeting this challenge through computational modeling of massive scale social interaction in order to yield new knowledge about the inner-workings of interaction in such environments so that support for healthy community formation can be designed and built. The ultimate goal of the eventual fully automated and context sensitive support would be to yield more resilient massive scale online learning environments. The methodological goal of the research is to contrast alternative computational modeling techniques including traditional social network analysis, simple topic modeling through Latent Dirichlet Allocation (LDA), and more sophisticated graphical modeling approaches that integrate these two lenses in a powerful way. This new approach is referred to as the Tensor Based Mixed Membership Stochastic Block Model, or tensor based MMSB.
In this MRI project, PI Rosé is partnering with colleagues at the University of Pittsburgh’s Institute for Learning (IFL) who are launching a MOOC on Accountable Talk through Coursera starting in September 2013 as well as Prof. Eric P. Xing of Carnegie Mellon University’s Machine Learning Department who specializes in large scale Graphical Modeling approaches to massive scale data analysis.

The role of employers in research on massive open online courses or MOOCs is a significantly under-investigated topic with great potential to shape MOOCs’ development and outcomes. Determining the extent to which taking and completing MOOCs can help individuals (particularly those who are less advantaged) advance in their careers is critical to understanding and capitalizing on the MOOC phenomenon. To explore the role of employers in the context of MOOCs, Duke University in partnership with RTI International, is conducting a quantitative and qualitative study. Major research questions include: To what extent do large employers know about MOOCs? To what extent are large employers using MOOCs as hiring and professional development tools? And how can MOOCs be more useful to employees and employers in hiring and professional development? Duke and RTI will collect and analyze two types of original data for this mixed-methods study: an online survey of about 300 of North Carolina’s largest employers and follow-up interviews with a subset of 30 employers that have been stratified based on their level of interest in using and interactions with MOOCs.

To date, faculty at the University of Toronto have offered more than half a dozen Coursera MOOCs. These MOOCs demonstrate the diversity of topics, disciplines, and teaching approaches that characterize the rich tapestry of academic life at the university. Early reflections on the experience of instructors and students have led us to consider the interplay of intent and expectations (corresponding to the “hatch” aspect of the proposal title), behaviour, and outcomes (“dispatch” in the proposal title) more closely. Our primary goal in the proposed research is to use survey, clickstream, and assessment outcome data from the MOOCs that have been offered to understand how those dimensions interact (the “match” in the proposal title). As well, the research project will focus on adaptation of analytic methods particular to large MOOC data sets and documentation of those methods.

MOOCs have garnered much attention from administrators looking to reduce the cost of education, and from educators looking to increase access and share their expertise. Rigorous research into real savings and efficacy, however, is in short supply. Recent reports suggest that the explosion of attention to MOOCs as an approach to moving traditional university teaching into an online environment may not have been grounded in good practice for online learning design. Thus the apparent collapse of several MOOC initiatives may lead to a widespread rejection of online delivery of courses for reasons which miss the point – that well-designed open and online courses can meet the learning needs of students who are often disadvantaged by traditional face-to-face education. Lead by a team of experienced distance and online educators, this project will consider the relationship between Massive Open Online Course (MOOC) design and open education resources and delivery, in reference to long-standing and well-researched principles of instructional design. This research seeks to answer the questions: Which instructional design and delivery principles are applicable to MOOCs, given increased class-size, unpredictable and wide ranging academic backgrounds of learners, availability of open education resources, and varied purpose among learners? Which ID principles are relevant and what might be added in consideration of unique MOOC characteristics?

MOOCs enable lifelong learners from around the world to interact with one another at unprecedented scales. Early literature on MOOCs has investigated the nature of learner interactions with their course environments. However, to date we know very little about the nature of interactions between learners or how these individuals exchange information with one another. Through a mixed method analysis of two MOOCs that emphasize collaborative problem-solving efforts, we aim to better understand who interacts with who in MOOCs, and how. We plan to interpret these interactions by contextualizing them according to the demographic characteristics and academic activities of each learner. These investigations will aid in analysing the formation of crowds versus communities in discussion settings; how information is aggregated and transmitted through interaction networks; and how participant backgrounds, course activities, performance, and communication tendencies are related. Using social network analysis and Bayesian inference in conjunction with insights derived from observations, participant interviews, and surveys, we hope to uncover how interaction patterns help us to understand how learning occurs through online interactions in ways that build on existing theoretical frameworks developed from previous learning and technology research. Ultimately, we aim to use this hybrid analytical framework to develop a typology that reflects the different ways in which MOOC participants communicate and interact in order to learn.

This study seeks to identify different types of participation, early predictors of high participation, and learning outcomes in a remedial writing English Massive Open Online Course (MOOC). It is estimated that 75% of students entering US community colleges require remedial education in at least one subject (NCPPHE and SREB, 2010). This need is also significant among students entering baccalaureate institutions and among the workforce. Further, students from around the world —especially those from racially marginalized and low-income populations— from around the world can improve their prospects by developing English communications skills.
The study will examine MOOC logfile data, outcomes from quizzes and peer-reviewed assignments, and pre/post course surveys. Patterns of participation will be identified, adapting the k-means clustering analysis used by Kizilcec, Piech, and Schneider (2013). Multivariate regression will identify early student behaviors that predict higher levels of final participation and improved achievement of learning outcome measures. These findings will be disaggregated by demographic and geographic criteria to identify differences between different student populations. Relevant findings from this statistical analysis will be further explored through qualitative analysis of surveys and discussion forums.
To date, there have been few MOOCs offered in remedial education subjects and little empirical research exists as to their efficacy. Findings from the project are relevant to educators and institutions considering developing MOOCs as an option for the large number of students who are not adequately prepared for college level courses. Additionally, MOOC platform providers and institutions developing other types of courses may benefit from this work.

The use of Massive Open Online Courses (MOOCs) to expand students’ access to higher-education has raised questions regarding the extent to which the course model can provide and assess authentic, higher-level student learning. In response to this need, many MOOC platforms have begun utilizing automated essay scoring (AES) systems that allow students to engage in critical writing and open-response activities. However, a limitation to these systems has been the lack of research investigating their validity, whether these assignments differentially benefit certain students, and best practices for utilizing this pedagogical approach. The research will examine these substantial issues through a three-phase investigation of the edX AES system and its adoption by faculty. These studies will test the effectiveness of the assessment method and examine student and instructor perceptions of this tool used in University of Texas at Austin edX courses (UTAustinX)

This research proposal brings together a multidisciplinary team with diversified expertise in distance education and educational technology. It uses a self-regulation and a socio-cognitive motivation approach theoretical perspective to analyze the impact of initial motivations on behavioral and cognitive engagement by learners and to identify the factors associated with MOOC course persistence. It uses a mixed methodology based on questionnaires and trace analysis of participant activity that reflects a learning analytics framework. The study will provide a better understanding of the factors that affect quantitative and qualitative engagement with course materials and design and should identify potential measures to improve MOOC completion rates.

“Bio Prep MOOC” is an online course to prepare university students for Introductory Biology. Though open to the public, it was specifically designed for incoming freshman who wish to become Biology majors but have insufficient background, many of whom are low-income, under-represented minorities, or first generation college students. Several hundred of the thousands who enrolled are incoming University of California, Irvine freshmen. The study will draw on a wide range of demographic and institutional data on these UCI students, including their performance in a subsequent Introductory Biology course, as well as course data for the broader set of students in the MOOC, including their access of course materials, quiz scores, participation in peer assessments, and peer assessment scores given and received.
Using this data and drawing upon statistical and machine learning regression techniques, the study will analyze what factors predict student participation in the MOOC, participation in peer assessments, peer scores given and received, and performance on quizzes. We will also examine how students’ performance on assessments and their peer assessment of others predicts their engagement in the MOOC, their quiz scores in the MOOC, and, for UCI students, their subsequent grades in Introductory Biology. The study will shed light on the value of a “gateway” STEM MOOC for facilitating success in subsequent university coursework, as well as the particular contribution of peer assessment activities toward this end.