Transcription

The Past, Present, and Future of MOOCsBalakrishnan DasarathyKevin SullivanUniversity of MarylandUniversity College, Adelphi,MarylandUniversity of Virginia,Charlottesville, VAAbstractMassive open online courses (MOOCs) are a recent development in online education aimed at unlimited participation and open access via the Web. They are a potentiallydisruptive technology, changing how education is delivered and funded around the world. MOOCs are relevant tosoftware researchers and practitioners, not only becausethey will increasingly receive lifelong education throughMOOCs and related technologies, but also because contentcreation, delivery, and enhancement of MOOCs is evolvinginto a new form of socially- and cognitively-embeddedsoftware development.In this paper we discuss the means by which education isenhanced by MOOCs and other digital learning technology. In particular, we distinguish between the free educational content provided by MOOCs and the emerging collaborative processes that MOOCs are enabling, which ishaving more transformative impact on education thanthe content itself. We discuss blended models of highereducation to suit different learner communities, as well asa nascent move towards instructional communities ofeducators that transcend institutional boundaries. Wealso explore MOOCs and their evolution as a subject forresearch in the learning sciences and the implications forR&D in software and systems engineering.1 IntroductionSoftware engineering is an incredibly complex business.Successful engineers must draw on broad knowledge,skills, and practices to solve complex problems: including domain knowledge, requirements elicitation, formalmethods, design and architecture, human computer interaction, accessibly and usability, programming languages, concurrency and synchronization, network anddatabase programming, security, virtualization, resource management, static analysis, testing, projectmanagement and more.Attaining mastery in these areas has typically involvedphysically attending courses at universities with a critical mass of faculty who specialize in these topics. Oncethere, all students must conform to the university’s classschedules, course offerings, teaching styles, and pace,some or all of which can be inflexible and sub-optimal—especially for non-traditional students, such as parentsor working adults. Likewise, the rising cost of residential education—particularly in the USA—is making itprohibitively expensive, even for traditional studentsDouglas C. Schmidt andDouglas H. FisherVanderbilt University,Nashville, TN[Rosenberger:12].Adam PorterUniversity of MarylandCollege Park, MarylandAn alternative education model that has received a greatdeal of attention over the past several years is the massive open online course (MOOC) [NYT:12]. A MOOC is aweb-based class environment aimed at large-scale global participation and open access via the Web. In contrastto traditional on-campus education, MOOCs enable flexible learning styles where students can pick and choosewhich classes they take, and when and where they dotheir work. Although much attention on MOOCs hasfocused on their flexibility and low cost, MOOCs alsoallow faculty to choose with whom to teach their courses, even across institutional boundaries.While MOOCs are still in their infancy, they have nevertheless already yielded improvements in the productionand delivery of high-quality digital lecture material,benefitting both online and traditional on-campuscourses [Schmidt:13]. MOOCs have also generated considerable interest from research and education communities on developing and evaluating scalable learningenvironment methods and technologies [UK:13].Despite the popularity of MOOCs, there has been significant debate about their limitations, particularly amonghigher education faculty, administrators, and trustees.These debates originate for various reasons including(1) unprecedented—and arguably unwarranted—hypeabout MOOCs in the press [Fox:13], (2) concerns thatMOOCs will de-skill and disintermediate faculty, therebycompromising the quality of education and student experience [Rees:13], (3) high non-completion rates inMOOC offerings [Hill:13b, NYT:13], and (4) pressure oninstitutions to reduce costs, largely in response to massive withdrawal of public funds for higher education[Rosenberger:12], and resulting inflation in higher education costs borne by students and families [Wyner:13.The disruptive potential of MOOCs was clear in a boardlevel governance crisis at the University of Virginia[UVA:12], a highly respected public university, triggeredby an ill-informed response to articles in the popularpress [WSJ:12] over the threat posed by MOOCs.To the extent MOOCs might appear to provide a “silverbullet” for financial pressures, some may be tempted tosee them as substitutes for courses taught by highlytrained—but also professionally paid—professors. Asthe debate has unfolded, several questions have arisenon the role of MOOCs in higher education and lifelonglearning, including:

What are appropriate ways to characterize andunderstand MOOCs? They don’t fit neatly into established education categories, making it hard to evaluate them and to predict their future impact. Who are the key MOOC stakeholders, what are theirvalue propositions, and how will their (sometimescompeting) interests interact to produce new phenomena over time? What are alternative strategies for using MOOCsand digital learning technologies in education? How will future MOOCs differ from the first generation of MOOCs offered during the past several years?This paper presents our assessment of these questions,based on experiences [Schmidt:13] teaching (1) MOOCson the Coursera platform [Schmidt:POSA], (2) hybridcourses that blend MOOC content with on-campus lectures [Fisher:12, Bruff:13], and (3) online courses inInformation Assurance at the University of MarylandUniversity College (UMUC), the largest public onlineuniversity [UMUC:13].2. Key Characteristics of MOOCsAlthough our experience with MOOCs [Schmidt:13,Bruff:13] indicate they are a valuable medium for conveying certain types of knowledge, they don't fall neatlyinto established education categories. To understandthe potential and limits of MOOCs, therefore, it helps todraw analogies with earlier forms of media. Althoughthese analogies are imperfect they help us to understand and evaluate MOOCs and their future evolution.For example, MOOCs can be viewed as multimedia textbooks for Internet-based distance education. They canalso be viewed as akin to community-based educationalTV shows. Another perspective is that they are the nextstage in learning management systems, such as Blackboard or Desire2Learn, albeit more open and globallyaccessible. Yet another view of MOOCs is as contentintensive, socially and cognitively embedded, cloudbased computing systems around which communities ofcontent publishers and consumers assemble and interact, e.g., akin to YouTube combined with Wikipedia andFacebook. Future MOOCs are likely to evolve with allthese media tools and services, resulting in richer learning and teaching environments.Although there are various ways to categorize MOOCs,the first generation of MOOCs generally shared the following characteristics: They have been delivered to students and the public online openly and freely—often taught (and taken) by professors at top universities around theworld. Some MOOCs provide open access to coursecontent, which helps promote reuse and repackagingof material for other faculty. Other MOOCs don’t permit reuse of the material without permission, thoughstudents can still access the material freely. They don’t enforce prerequisites. In contrast, conventional on-campus courses follow well-defined sequences to ensure that students have the necessarybackground knowledge and experience before attempting to learn the new material.Various consequences result from the characteristics ofconventional MOOCs described above. For example, thecompletion rate of MOOCs is low: typically only 5-10%of MOOC registrants actually attain a statement of accomplishment. In contrast, completion rates at the University of Maryland University College (UMUC)—an established online program—are greater than 50% at thegraduate level after the second semester of enrollment(it would be even higher if UMUC had a more selectiveadmission policy). Completion is greater than 85% atselective residential universities, such as Vanderbilt andthe University of Virginia. There are various explanations for the low completion rates in MOOCs [Hill:13b],e.g., because MOOCs have no pre-requisites and are free,many students aren’t adequately prepared to master thematerial or aren’t motivated to finish the entire coursesince there’s no penalty for non-completion.Judging from analysis of the completion rates, MOOCsoffered to date appear better suited for highly selfmotivated learners than those less self-motivated (andwho may benefit more from conventional residentialeducation models [NYT:13)]. The first generation ofMOOCs tended to couple text and video material thatcan be easily browsed separately from other course material, such as assignments and exams. Students oftenfind value in these text and videos, without having toparticipate in all the trappings of the full course format.In this sense, MOOCs are used like books, where it maynot be necessary to read every page cover-to-cover, inprinted order. A significant fraction of non-completersmay simply benefit as they would by sitting with a bookin a library or at home: picking, choosing, with no senseof obligation to report to anyone or to be assessed. Inthis view, students are “active auditors” who use MOOCslike streaming, on-demand, educational TV shows, delivered by inspiring teachers.Another consequence of first generation MOOCs is thatthey aren’t well-suited for fields (such as Chemistry andBiology) with labs that require preparation, experimentation, specialized equipment, observation, and resultcapture. Likewise, it’s hard to assess sophisticated assignments accurately and scalably in first-generationMOOCs due to limitations with the current state-of-thepractice of auto-grading and peer-grading methods andtools[Schmidt:13]. These limitations may be ameliorated over time as research on state-of-the-art automated tools and methods transition into practice [Fox:13]and as MOOCs incorporate capabilities for online labor-

atories found in other online learning contexts[Rivard:13]. In general, however, drills, exercises, andinteractions don’t yet scale in the same way as lecturedelivery via current MOOC platforms.3. MOOC Stakeholders and Value PropositionsAnother way to think about the evolution of MOOCs is interms of stakeholders in the MOOC ecosystem and theirvalue propositions. Stakeholders include (1) MOOC content publishers and curators, (2) professors, (3) traditional and non-traditional students and their families,(4) universities, (5) the public, and (6) scholarly communities. This section develops this perspective.3.1 Key MOOC StakeholdersMOOC publishers and curators (e.g., Coursera, Udacity,EdX, and others) currently provide courses for free, butare ultimately profit-driven enterprises. Their chief value proposition is eventual financial profitability andwealth creation for their investors and owners. Onereason why content publishers engage in the MOOCmarket stems from the enormous amounts of moneyspent on education, at all levels, each year.For example, Americans spent about 461 billion forthe post-secondary education in 2009, 42% more thanthe amount spent in 2000 [CNN:11]. These costs arebecoming prohibitive in many countries; e.g., it can costover 50,000 a year to attend a selective private collegein the USA. The content publishers and curators perceive that MOOCs can help (1) universities by reducingsome of their operating costs, (2) students and theirfamilies by making education more affordable, better,and accessible [Rosenberger:12], (3) and future employers of students by providing data on student performance and credentials in designated areas of study.Professors have several value propositions. Faculty canbenefit by folding MOOC content into their own coursesvia blended learning [Bruff:13] and flipped classroommethods [Fisher:12]. Such use involves learning MOOCcontent quickly enough themselves to teach new material or adopting “shows” for use in their own courses.Opportunities to get up to speed more quickly and toincorporate free, high-quality material into classes havereal value for professors. Likewise, MOOCs and onlinecontent generally enable instructional communities ofeducators to create and share content readily, on platforms ranging from Coursera and EdX, to YouTube.Coursera, Udacity, and EdX have also created an impressively successful—albeit only quietly recognized— “starprofessor” award system that helps MOOC faculty burnish their personal brands. For example, faculty invitedto produce and deliver MOOCs receive acclaim and visibility, which are coins of the realm in academia. Facultywho teach MOOCs may also find numerous ways tomonetize their fame, for example, through speaking engagements, by teaching with for-profit content publishers, such as Pearson’s LiveLessons video series, aimedat corporate training, and more.Universities are major stakeholders. They value theirown financial health, which makes them sensitive toeducational costs and quality. They value their studentsand the student experience. They also greatly valuetheir reputations, on which student demand and tuitionrevenues depend. In part for this reason, in conventional MOOC models (such as Coursera and EdX), universities play a vital role as arbiters of what MOOC content gets produced and delivered by their faculty.Traditional and non/traditional students are stakeholders, both as consumers of the MOOC content and aspotentially scalable labor pools for peer assessmentsand crowd-sourced content review and improvement.Likewise, families/employers are stakeholders to theextent that their loved ones/employees education willbe enhanced (or not) by MOOC-related changes in education, or by the impact of MOOCs on higher educationcosts. Ironically, families paying the tuition of their children who attend top universities may be indirectlyfunding MOOCs (the so-called “MOOC tax”) since thecosts of their faculty and staff who produce and deliverMOOCs may come from the tuition paid by residentialstudents and their families.The availability of MOOC materials is a boon to people around the world. In some cases for the first time,people around the world have opportunities to accessthe materials and processes of quality higher education.While MOOCs have initially focused on individual courses that are closely related to traditional on-campuscourses—MOOCs in-the-small—there are opportunitiesfor MOOCs in-the-large, as well, where coalitions orcommunities of scholars document whole bodies ofknowledge in MOOC-like form. This possibility raisesquestions regarding the architecture of knowledge andthe means for its large-scale production and evolution.3.2 Implications of this Value-proposition AnalysisThere are a number of implications of our valueproposition analysis presented above.MOOC content publishers must become profitable,evolve into something different, or go out of business.Current MOOC providers are multimedia publishers(who focus on content creation, integration and dissemination), rather than higher education institutions (whofocus on scholarship, assessment, and credentialing).Coursera, EdX, Udactity et al. are thus in the same market as companies like Pearson, i.e., publishers seeking todevelop and promulgate new educational uses of the

Web and related technologies, such as mobile devices,cloud computing, and data analytics.These publishers have so far operated with investors’capital, but their business models will have to becomesustainable soon. As they do, there will be changes inthe tradeoffs involved in collaborating with these publishers, as seen by other university, faculty, and studentstakeholders. These providers may also find themselvesthreatened by rapidly falling barriers to entry into theMOOC market. For example, the availability of the EdXMOOC platform (mooc.org) coupled with consumerpriced video production tools (e.g., cameras, audio, light,green screens, teleprompters, screen capture, editingsoftware, and video hosting) makes it relatively easyand cheap to produce satisfactory content.If MOOC content publishers can undercut universities,even smaller players may undercut the big MOOC publishers and commercial course management platforms,such as Blackboard and Desire2Learn, that provide environments for rich content creation and maintenance,as well as support for essential classroom features, suchas gradebook facilities for creating quizzes and examstied to the gradebook, as well as discussion sessions andchats. As production costs decrease it’s possible that wewill see a decentralization of MOOC creation, andbroader adoption of open-source alternatives to proprietary MOOC publishers and platforms.Universities will have to become savvy about providing private enterprises valuable intellectual property, aswell as the valuable data produced by student engagement with these interactive systems. The value exchanges will have to make sense financially and withrespect to pedagogical goals, privacy, and legal and ethical obligations to students. If universities choose to implement their own MOOC production and delivery systems, it may become feasible to move production inhouse and to control distribution. To the extent thatUniversities own course content, pressures will militateagainst unlimited free distribution. As with newspaperslike the New York Times, universities will struggle tobalance financial self-interest against the public goodand other stakeholders’ interests.If universities, instead, choose to focus on the parts ofeducation that are not easily delivered online (such asindividualized face-to-face instruction and coaching,exposure to cutting edge facilities and research projects,access to study abroad programs and internships, andintegrated learning opportunities that span diverselearning fields) they may treat MOOCs as simply anotherservice they can use to further their educational missions and outreach.The benefits that professors have garnered from being among the first MOOC stars will fade, as barriersto entry and novelty diminish. Professors may also cometo believe that the benefits of creating MOOCs are outweighed by the significant costs in time and lost opportunities. Of course, a small number of popular star professors could be backed by investors with resourcesthat neither individual faculty nor institutions would beable to provide. It’s also not inconceivable that actorswill eventually play professors, to deliver content withprofessionally trained voices, familiar faces, and actingskills. Prof. Larry Sabato at the University of Virginia hasalready reportedly hired professional film-makers toproduce his MOOC. On the other hand, high productionvalues aren’t always necessary if the content is useful toconsumers, as Kahn Academy has clearly shown.Access to free, high-quality educational materials ispositive for students. A MOOC can be of enormous value by providing access to materials that might otherwise be inaccessible. High quality MOOC courses willalso pressure all professors to up their game, as students raise their expectations based on exposure to professionally produced MOOCs.A key question is whether in the long run students willbe disadvantaged if educational institutions try to combine MOOC content with low-cost teaching assistants toreplace classes traditionally taught by professors. In thisregard, the question isn’t necessarily whether MOOCsare as good as traditional on-campus courses, butwhether they are so much cheaper that business-oriented governing boards and legislatures will see them(with or without adequate rationale) as good enough tojustify sweeping changes. Something valuable for students and society may be lost if students can no longerexperience the benefits of education and mentoring bycommitted, physically present, highly qualified—andoften inspiring—professors.It’s not yet clear, however, that thoughtful universityleaders see acceptable ways to use MOOCs to reducecosts substantially in the short term. For example, justas existing instructional content represents an enormous investment, MOOCs will be costly to produce (often thousands of dollars per hour of finished material).Moreover, interaction, feedback, testing, and creating alearning culture are all labor- and knowledge-intensiveand do not (yet) scale as well as lecture delivery (whichis only one part of teaching). For now, top institutionsare using MOOCs to increase their students’ “bang forthe buck” (i.e., increasing “the bang” and not by reducing“the buck”) by customizing MOOC offerings, in additionto providing all the richness of traditional courses, asdiscussed in [Fisher:12, Bruff:13] and explored furtherin Section 4.2. Others are incrementally creating MOOCs,adding content bit-by-bit over a long period of time, andintegrating this new material with traditional lecturesas parts of flipped classrooms.

Visibility into the materials and processes of highereducation is wonderful for the broad public, both in theU.S. and around the world. Making high-quality coursematerials available to the public is an enormous contribution to the public good. Universities and professorswill have to decide, however, whether they’ll remaincommitted to this goal given the costs and the need forsustainable business models. It’s not clear that we’ll seea continued proliferation of high quality open accessMOOCs, due to the lack of a clearly articulated return oninvestment. We may see MOOCs used as “loss leaders”for profitable online degree programs, i.e., as a naturalspecializations of a strategy that uses MOOCs for branding, recruiting, and outreach.Communities of knowledge are also likely to recognizethe potential of MOOC-like encodings of knowledge.Professional, societies (such as the IEEE) have historically documented bodies of knowledge (e.g., the Software Engineering Body of Knowledge available at www.swebok.org). A carefully architected set of MOOClike courses may be a viable—and potentially far moreuseful and accessible—representation of the stableknowledge of a field, such as computing or software.As stakeholders begin to understand and adapt in thisecosystem—and as the technology and methods mature—the first generation of MOOCs will likely be replaced by more evolved tools and methods that computationally enhance and enable education, includingmethods for better organizing the bodies of knowledge inmany fields into more readily accessible digital form. Learner accountability Are students who they saythey are and is their work original? Curricular diversity and depth The depth andbreadth of disciplines covered. Well-roundedness Education is more than a mastery of a field’s content; does the model supportlearning many soft skills, including social, networking and leadership skills? Support structures Guidance and mentoring inacademic and non-academic aspects of real or virtualcampus life. Scalability How well the model support increasingthe number of learners without having detrimentaleffect on the quality of learning? Cost Does the model support affordability by learners and their parents and employers—are the benefits worth the cost? Flexibility Refers to freedom in time and space. Accessibility – Refers to the degree to which educational resources are available to physically-challenged learners. Diversity Does the model provide opportunities fordiversities in various dimensions, such as socio-economic, cultural, and geographic?4. Alternative Strategies for Using MOOCsThis section describes our vision of how the digitallearning capabilities—leveraged and promoted byMOOCs—are evolving in higher education based on ourexperience. In our view, there’s no one-size-fits-allmodel for digital learning. Instead, we expect thatMOOCs, small private online classes (SPOCs) [Fox:13],and traditional residential education will blend togetherto support different types of students in different lifesituations, as shown in Figure 1. This figure depicts different digital learning adoption strategies ranging fromtraditional residential education (which is largely synchronous and delivered face-to-face) to online education programs (which are largely asynchronous).4.1 Evaluation Criteria for Digital Learning ModelsTo determine how digital learning is evolving, we firstassess what attributes a particular learning model supports. The following attributes provide a basis for understanding and evaluating various learning models: Quality of real-time feedback and motivation How well (and soon) do teachers/learners assessand provide feedback to alter directions or missteps?Figure 1: Strategies for Digital Learning Adoption4.2 Alternate Digital Learning Adoption StrategiesBelow we use the attributes described in Section 4.1 tocharacterize and evaluate the models across the spectrum shown in Figure 1. Likewise, Table 1 applies theassessment criteria presented in Section 4.1 to compareand contrast the following two models at opposite endsof this spectrum: Traditional residential education. In its idealized(and sometimes romanticized) form, this model ischaracterized by an immersive education experiencethat places a premium on frequent, in-depth, andsynchronous face-to-face interactions between faculty and students. Universities are communities thatcater to many aspects of student scholastic and inter/intra-personally development, including sports

and recreation, social engagement, political activism,student organizations and clubs. Although the residential education model has a long history of success, it also incurs a premium cost. For example, during the past several decades tuition costs in the USAhave risen much faster than the rate of inflation—especially as public support for institutions of highereducation has been radically reduced—increasingU.S. student debt above 1 trillion [Wyner:13]. Online education programs. This model is characterized by students who are working professionalsinterested in advancing their careers or enteringnew fields. These programs typically offer career-oriented degrees and professional certificates in accounting, finance, Management, IT, cyber security,etc. Students are generally adult learners who arenot looking for social and networking skills and opportunities. Unlike MOOCs, class sizes are typicallysmall to make online learning experiences personalized. There’s a potential to leverage MOOC-developed lectures in this model to reduce costs, especially in core courses. Conversely, Udacity is focusingon making inroads into higher education not withMOOC-developed content, but with technology andsupport services for this online education market forcredit at traditional universities[Kolowich:13a].There’s a need to leverage authentication and identification biometric techniques to minimize impersonation which helps improves credibility ofonline degrees [Bailie:09].As discussed in Section 2, MOOCs can be viewed as amicrocosm of digital learning, so they share many characteristics of the online education model outlined inTable 1. MOOCs can scale better (at least the deliverypart) and may be even cheaper and more diverse thantraditional online education models. First generationMOOCs, however, perform poorly on the first five rowsof attributes listed in Table 1, even more so than traditional online education model due to their massive size.While the two models evaluated in Table 1 are stereotypic education models—albeit at opposite ends of thespectrum—hybrid approaches are also becoming popular. For example, residential education institutions areadopting technologies and concepts (such as flippedclassroom) that leverage MOOCs without going fullyonline. Likewise, online education programs may require some face-to-face interactions with students (e.g.,to authentic their identities in a physical testing center).The remainder of this section discusses ways in whichdifferent strategies for adopting digital learning (all ofwhich both leverage and inform MOOCs to some extent)can yield different blends of synchronous and asynchronous models of faculty/student interaction: Tightly-coupled blended learning environments.This hybrid model is being explored and adopted in anumber of top-tier institutions [Fisher:12, Bruff:13].The main goal is not cost reduction, but rather increasing teaching effectiveness by leveraging multimedia content (including content from MOOCstaught by faculty at other universities), innovativestudent/faculty engagement methods (such as theflipped classroom), and online discussion forums(such as Piazza) that encourage asynchronous interactions between faculty, course staff, and students.For example, all Vanderbilt faculty who have offeredCoursera MOOCs thus far are applying the digitallearning material from their MOOCs to augment theiron-campus classes. The primary benefit of the blending learning model is increased flexibility in facultystudent and student-student interactions; the otherattributes of this model are essentially the same astraditional residential education.Classes co-taught locally and remote satelliteclasses. In this hybrid model local classes are faceto-face and lectures are broadcasted using streamingvideo and audio to remote campuses. Remote campus classes are largely staffed by adjuncts and/orjunior faculty. The course delivery to satellite campuses is predominately synchronous. Concepts suchas flipped classroom and asynchronous discussionsessions from MOOC can be applied to enhancelearning/teaching effectiveness. On

there, all students must conform to the university’s class . (it would be even higher if UMUC had a more selective admission policy). Completion is greater than 85% at . MOOCs due to limitations with the current state-of-the-practice of auto-grading and peer-grading metho