Skip to content

Massive Open Online Course (MOOC)

Pattern · All patterns
Evidence · 33 claims (10 for, 18 mixed, 5 against) · 32 studies (7 review, 7 associational, 6 causal, 5 quant-synthesis, 3 design, 3 theoretical, 1 qualitative), q1–q4 · 5 of 32 report an effect size · 29 claims rest on one study

Description and scope

A reusable course-level configuration for a course that anyone may join, free or nearly so, at an enrolment so large that no teacher can respond to each learner: online-self-paced (or on a weekly release that learners follow at their own pace), over days-weeks, for mostly adult learners in postsecondary and adult-workplace bands. Its defining decision is what replaces individual teacher time: short video or reading segments with self-checks → a planning activity and recurring planning and monitoring prompts → discussion seeded and summarised by a small staff → automated scoring for closed items and rubric-guided, calibrated peer review for open work → an individual final task → evaluation by what each learner came for, not only by completion. The intended changes are in each learner's performance on the course's representative task (general-achievement, skill-performance) and in how far each learner gets towards their own purpose (persistence).

Two historical forms share the name. The first MOOCs (Downes and Siemens, 2008–2009) were built on Connectivism: learners were to build and navigate their own network of people and distributed content. Commercial "xMOOC" platforms (Coursera, Udacity, edX) adopted a broadcast model of video lectures and quizzes after 2011 (Weller, 2018), and Wiley (2014) argues that most of them fail the 5R openness test because their content cannot be freely copied, revised or redistributed. This page covers both, as two configurations of one course form; the choice between them is a row in the situation table below.

No claim in this wiki tests a MOOC against another way of teaching the same content, on learning or on completion. What the wiki holds is: second-hand reports that few registrants finish (about four percent in one report, under 10% in another) and that viewing falls off in videos longer than about nine minutes; two randomised field experiments in large online courses of a brief planning activity, which raised completion for some learners and not others; two quasi-experiments, one of them in a MOOC writing course, of rubric guidance for peer reviewers; syntheses of self-regulation prompts and interventions in online environments, few of whose studies were run in MOOCs; one case report of automated grading in a MOOC platform and a review warning of bias in such grading; and reviews and surveys of connectivist courses, video styles and institutional adoption. So the sequence below is the earlier page's six steps, expanded and kept as a default, with each step's evidence status stated; the response-dependent policy beside it is an untested design proposal.

How it sits beside its converted neighbours, each of which owns a relationship this page uses and does not repeat:

  • Online Course Design holds the model for an online course with a teacher behind it: constant weekly structure, a contact week, three kinds of interaction in a weekly cycle. A MOOC is that course with the teacher's individual time removed, so this page decides what replaces it. Where a MOOC has a tutor for each small cohort, use that page instead.
  • Self-Regulated Learning (principle) and its pattern hold the plan–monitor–act model this page leans on for planning and monitoring prompts.
  • Self-Directed Learning (principle) and its pattern own when learner-directed study helps and how to sequence it; the connectivist configuration here is one case of it.
  • Structured Peer Review owns the review-and-revise cycle; this page uses it as the scalable substitute for teacher feedback on open work.
  • Blended Learning decides how to split a course between online and in-person time, which applies when a campus or employer wraps a MOOC with local sessions.
  • Digital Learning holds that a medium helps only through the method it makes affordable; this page credits no gain to the MOOC format itself.

Inputs: establish the design brief

Field Record or elicit
Learner and starting observation Why each registrant enrolled and what they will count as success (a certificate, one module, a skill for work, browsing); prior knowledge of the content (novice, intermediate, advanced, mixed) from a short optional entry task; prior experience of studying online alone; device (computer, phone only), connection and data cost; language of instruction against first language; time available per week. Most registrants will not answer; record non-response as unknown, not as low motivation.
Context and activities Expected enrolment and its spread across countries and time zones; staff hours per week (lead teacher, teaching assistants, volunteer community mentors); fixed weekly release or fully self-paced; length in weeks; platform features (auto-graded items, peer-review workflow, forums, analytics, downloads); whether any learners take the course together in a room (a campus using it in a blended course, an employer's study group); licence of the content.
Learner-valued goal The registrant's own purpose, asked at enrolment and again at the midpoint. Record where it diverges from the designer's objective: a learner who wants one module and leaves after it has not failed the course.
Designer objective The capability and its knowledge type (concept, procedure, principle, complex-skill, metacognitive-strategy); the representative task; which parts a machine can score, which a calibrated peer can judge against a rubric, and which need an expert (embodied performance, clinical judgement, nuanced writing).
Outcome Performance on the individual final task, with its instrument and horizon; completion counted against learners who intended to complete, with the week each learner stopped (persistence); time spent (learning-time); accuracy of peer and automated scores against an expert sample (assessment-accuracy). Satisfaction and self-assessed competence are self-report and recorded as such.
Comparison being made What the MOOC replaces or adds to: nothing (new access), a campus course, a paid online course, a textbook. Record what else changes with it (cost, credential, pacing, staff), because no claim here separates those from the format.

A request to "turn our course into a MOOC" or "raise our completion rate" does not specify these. Ask what registrants come for, what the representative task is and who can judge it, and how many staff hours a week exist before choosing segment lengths, a peer-review load or a completion target.

Sequence and conditional policy

The default is the earlier page's six steps (produce recorded content; open enrolment; release content with low-stakes quizzes; run forums; assess by machine or peers; charge for an optional certificate), expanded into a course with the substitutes for teacher time made explicit. Steps 3, 4, 5 and 7 have neighbouring claims, carried by extrapolation; the others are untested here. Quantities are proposals to set locally and record. A typical frame: 4–8 weeks, 2–5 hours of learner time a week.

  1. Decide openness before production. Choose the licence (an open licence that allows copying, revision and redistribution, or a closed one) and whether there are start and end dates, paywalls on graded work, or a fee for the certificate. State the choice on the course page. Evidence status: untested; Wiley's (2014) argument that a closed, dated, fee-gated xMOOC is "open" only in the open-entry sense is an argument, not a claim tested here (see Open Educational Resources).
  2. Ask each registrant what they are here for, and offer routes. At enrolment, two optional questions: "what do you want from this course?" (one module, the whole course, a certificate, a look around) and "how many hours a week can you give it?" Offer an audit route (any module, no deadlines) and a completion route (all modules, a final task). Report outcomes by route. Evidence status: untested proposal. About half of registrants never viewed a lecture and about four percent finished, in one report [-W], and MOOC completion is typically under 10% [-W], both second-hand reports; neither says what those registrants intended, which is why the question is asked.
  3. In week 1, run an eight-minute planning activity. Each learner writes the best outcome of finishing, the most likely obstacle, and an if-then plan ("if (obstacle), then I will (action)"). Show it back at the midpoint. Evidence status: forming implementation intentions improves goal attainment [~S]: a meta-analysis of 94 tests (mostly university students, d = .65), and two randomised field experiments in online courses (n = 17,963) in which a brief mental-contrasting plus if-then writing activity made learners in individualist cultures 32% and 15% more likely to complete, left learners in collectivist cultures unaffected, and worked when the main obstacle was predictable and surmountable (work or family) rather than a lack of time. Expect it to help some learners, not all.
  4. Cut content into short segments, each followed by questions. Proposal: video segments of 3–6 minutes, never over 9, a talking head beside slides or a hand-drawn style for worked explanation; transcripts and captions; after each segment, 2–4 auto-scored questions with immediate answers and a one-line explanation. Evidence status: in 6.9 million MOOC viewing sessions, the fraction viewed fell as videos got longer, with median engagement under half the duration beyond nine minutes [+W] (a dissertation's report of a log analysis; viewing, not learning); short videos, a talking head on slides and the drawn style were the more engaging production styles [+W] (a review's report of one edX study, science and engineering courses only). The segment length and question count are proposals.
  5. Put planning and monitoring prompts into every week. At the start of each week, "what will you do this week, and when?"; after each segment's questions, "which question did you get wrong, and what will you re-watch?"; at the end of the week, a self-check with answers and "what will you change next week?" Make the prompts task-specific, and give feedback with them. Evidence status: metacognitive prompts improve learning [+M]: a meta-analysis of experiments in computer-based learning environments (learning outcomes g = 0.40 against unprompted controls; larger with feedback, task-specific and adaptive prompts) and a systematic review of 35 online-learning studies in which lower-prior-knowledge learners in one study needed training before prompts helped; self-regulated learning interventions had a moderate pooled effect in online and blended environments [+M] (15 studies, 5,519 participants). The online review found that support helped only when learners used it, that feedback alone showed no effect, and that very few experiments had been run in MOOCs [~M]; this step is carried to MOOCs from online courses generally.
  6. Seed the discussion, form small cohorts, and summarise weekly. One discussion question per week that has more than one defensible answer; optional study groups of 10–20 learners matched by time zone or language; one staff post per week summarising the main threads and the commonest error. Use volunteer community mentors (former learners) where staff hours are short. Evidence status: untested proposal. In early connectivist courses not all learners could direct their own learning, and some felt disconnected and demotivated [-W] (a survey's second-hand report of two course studies), which is a reason to give the discussion structure and a visible staff presence rather than to leave it self-organised.
  7. Score closed items by machine, open work by calibrated peers, and audit both. For open work (an essay, a design, a recorded performance), give a rubric with three or four criteria and level descriptors; before reviewing, each learner rates two or three anchor samples and sees the expert ratings; each submission gets three peer reviews; staff spot-check a sample. If open text is scored automatically, compare its scores with expert scores on a sample that includes non-native writers before relying on it. Evidence status: rubric guidance helped MOOC writing-course peers place samples in the right quality level, mainly on subjective, technically complex criteria [+M] (two quasi-experiments; the other, with 47 graduate students, found the rubric changed what reviewers commented on and restrained reflective comment); NLP grading in one MOOC platform cut grading time by 70% and raised satisfaction by 15% [+W] (a case report, no comparison of learning); automated grading may disadvantage international and non-Western students [-W] (a review's stated risk). The anchor-sample calibration and the number of reviews are proposals.
  8. Watch the first week and every week after, and act on the first miss. Record who opens each week and who completes its checks. Send one message (automated, personal in tone) after the first missed week, linking the learner's own plan from step 3 and a reduced catch-up route. Evidence status: untested proposal. In online courses (not MOOCs), attrition in the orientation week was twice that of onground classes' first week [~W], and a single "tipping point" lesson explained dropout in only about a quarter of 196 online courses [~W], so watch every week, with extra care in the first.
  9. End with an individual task and evaluate by route. The final task is the representative task, done alone; the optional certificate (step 1) rests on it. Report performance on it, completion among learners on the completion route, and the week each learner stopped. Do not report completion over all registrants as the course's result. Evidence status: untested proposal.

Progression across the course (untested proposal). Weeks 1–2: the planning activity; the shortest segments (3–5 minutes) with questions after each; prompts at the start and end of each week with model answers shown; peer-review calibration on anchor samples before any real reviewing; staff summary posts twice a week. Middle weeks: segments up to 6–8 minutes; one planning prompt a week, without model answers; real peer review, three reviews per submission, with staff spot-checking about one in ten; study groups take over the weekly question, with a staff summary once a week. Final week or two: no new content; learners rehearse the final task on a practice version with peer review, then do the final task alone. On the completion route a learner moves to the next week's material when they have done that week's checks (any score), not when they have passed them; a learner who has missed two weeks gets the reduced route (the two or three activities the final task needs), not the backlog. On the audit route there are no deadlines.

Observation Discriminating probe Proposed action, with its uncertainty
Large drop between enrolment and first video Compare by the intent each registrant gave at enrolment; look at whether non-starters opened the course page at all. If most non-starters said "a look around", it is browsing, not loss; report it as such. If completion-route registrants did not start, check access (device, data, start date) and send one message. The completion figures recorded here do not separate these.
Learners watch to the end of short segments but fail the questions after them Compare the wrong answers with the segment: is the point stated, shown in a worked example, or only implied? If only implied, add a worked example or a second segment; if stated and still missed, the questions may test something not taught. Watching does not show understanding.
Viewing drops part-way through a segment Check the segment's length and where the drop happens (a long derivation, a slide with no speaker). Split the segment at that point or change the presentation. The viewing data recorded here are about length, not about learning.
Peer ratings disagree widely with staff spot-checks Compare by criterion: are the disagreements on subjective, complex criteria or simple ones? On complex criteria, add anchor samples for that criterion and a worked rating; on simple criteria, score by machine instead. Rubric guidance helped most on complex criteria in the one MOOC study recorded.
Automated scores fall lower for non-native writers than staff scores of the same work Score a matched sample by staff and machine; compare by language background. Route that group's work to peer or staff scoring until the scorer is checked; do not rest a certificate on an unaudited scorer.
Active forum, but study groups fall silent after week 2 Ask a sample of group members why; check time-zone and language matching. Re-match groups, give each group a weekly product, or merge small groups. Silence fits mismatched groups, a weak question and learners who want to study alone equally.
A learner falls behind from week 3 Show them their week-1 plan and ask whether the obstacle has changed. Offer the reduced route and a new deadline. The planning experiments found no effect for learners whose obstacle was lack of time; for them, a smaller route matters more than a new plan.

Fitting the design to a situation

The three facts that most change the decision:

  • What registrants come for and how far they can direct their own study. A MOOC's substitutes for teacher time (prompts, planning, self-checks) carry learners who can use them; learners new to the domain or to studying alone need more structure, and many registrants never intend to finish. If unstated, ask: what do registrants say they want, and have they studied online on their own before?
  • Who or what can judge the representative task. Closed items can be scored by machine; complex written or designed work can be judged by peers with a rubric and calibration; embodied performance and expert judgement may need an expert. If unstated, ask: what is the final task, and could a trained peer with a rubric tell good work from poor?
  • How much staff time there is, and whether anyone is local. A lead teacher with a few hours a week for thousands of learners can summarise and spot-check, not respond; a campus or employer group studying together can add what the MOOC lacks. If unstated, ask: how many staff hours a week, and do any learners study together in one place?
If the brief says… Then change… Basis
Learners are novice in the domain, or have not studied online alone before Shorten segments to 3–5 minutes with worked examples; keep the start- and end-of-week prompts all course with model answers for the first three weeks; teach learners how to use the prompts in week 1 (one worked example of a plan and a self-check); give the reduced route early. Metacognitive prompts [~M]: in the online review, lower-prior-knowledge learners in one study needed training before prompts helped; carried to MOOCs
Learners are advanced or practising professionals (a teachers' PD MOOC) Offer a placement check to skip modules; replace content questions with an application task in their own workplace, with evidence of practice (a lesson plan, a recorded session, a reflection) reviewed against a rubric; consider stackable micro-credentials for the applied tasks. One issuer reported that micro-credential stacks closed the gap between MOOC-Ed content and classroom practice [+W] (a q1 case report, no comparison)
Learners are adolescent or younger, or a school uses the MOOC Do not rely on open forums or self-pacing: assign the MOOC within a class, with a teacher setting weekly targets and reviewing the self-checks; keep forums closed to the class or moderated. Untested; the claims recorded here are about adults and university students
The target is an embodied or performed skill (music, laboratory, clinical, a language spoken aloud) Add recorded performance submissions judged against a rubric by calibrated peers, and periodic live sessions with an expert for a sample of learners; do not rest the skill on video and quizzes. Embodied music learning resists short videos and self-graded quizzes [-W] (a review's argument); rubric-guided peer rating [~M], tested on writing, carried to performance
The final task is written or designed work too complex for machine scoring Calibrated peer review: a rubric of three or four criteria, two or three anchor samples rated before reviewing, three reviews per submission, staff spot-checks; tell reviewers to add one comment that is not on the rubric, since rubrics narrowed reflective comment. Rubric-guided peer rating [+M]: benefit concentrated on subjective, complex criteria; one study found less reflective feedback
Open text is to be scored automatically, and learners write in a second language Audit the scorer on a sample by language background before use; keep a human or peer route for anyone who disputes a score; do not let an unaudited score decide a certificate. Automated grading bias [-W] (a review's stated risk); one platform's time saving [+W] is not evidence of fairness
Learners come from many cultures, including collectivist ones Keep the week-1 planning activity, but do not count on it for completion everywhere; for those learners, add study groups formed by region or language, with a shared weekly goal. Implementation intentions [~S]: no effect for learners in collectivist cultures in the field experiments; the study-group alternative is untested
Adults in work settings with predictable obstacles (shifts, family) Make the if-then plan name the obstacle and a fixed weekly slot; show it back at the midpoint; allow a two-week extension without penalty. Implementation intentions [+S]: worked when the obstacle was predictable and surmountable, not for lack of time; extension is untested
The aim is networked, connectivist learning (learners build their own network of sources and peers) Use it only where learners can direct their own study; still give a weekly hub (a digest of learners' posts and the staff's pick), an orientation to the tools in week 1, and a named place to ask for help. Connectivist courses: some learners disconnected and demotivated [-W]; one structural model linked autonomy, openness and interactivity with achievement [+W] (a survey's second-hand report)
Learners on phones only, or with costly data Short downloadable segments with audio-only and transcript versions; quizzes that work on a small screen and offline; no task that needs a computer program the phone cannot run. Untested proposal. SuaCode, a smartphone coding course, drew mostly intrinsically motivated learners [~W], and free, accessible content was a weak reason for enrolling there [~W]; both are about motivation, not delivery
Mixed modes: some learners take the MOOC together in a room (a campus course or an employer's study group) while others are alone Give the local group a facilitator guide: one weekly meeting of 45–60 minutes to work the hardest problem set together and discuss the week's question; the local group posts its answer to the forum so remote learners see it. Untested; see Blended Learning for dividing work between online and room
One lead teacher, or a few hours of volunteer time a week, for thousands of learners Spend staff time only on a weekly summary of the commonest error, seeding the discussion question and spot-checking peer reviews; recruit former learners as community mentors; automate the reminder messages. Untested proposal
A fixed short time budget (e.g. 2 hours a week for 3 weeks) Plan 6 hours in total: per week, 50 minutes of segments and questions, 30 minutes of the main practice task, 20 minutes of discussion or peer review, 10 minutes of planning and self-check, 10 minutes slack; drop forums rather than shorten practice. Untested proposal
A single session only (a one-off open webinar or a single-module micro-course) Treat it as one segment-and-practice block, not a course: 3–4 short segments, questions after each, one practice task with a model answer, one planning question at the end ("when will you use this?"). Untested proposal
Learners with low confidence or low literacy in the language of instruction Captions and transcripts in plain language; a glossary per week; first-week tasks that cannot be failed (an introduction, a practice question with answer); allow audio or picture responses in the forum. Untested proposal; connectivist courses left some learners disconnected [-W], which is a reason to build early success in
A credential with real stakes (credit, a job requirement) Identity-checked, individually done final task (proctored or oral by sample), scored by staff or an audited scorer; keep peer review for formative work only. Untested proposal
Learners will share private or sensitive material (their workplace, their health, their students) Make reflections private by default; let the learner choose whether a piece goes to peer review; allow a display name in the forum; tell learners how many people can see a post. Untested proposal
Learners produce work for a real audience (an open resource, a performance, a public portfolio) Keep the final product public and whole; peer review against the brief's own criteria before publication; offer the open licence choice from step 1 for learners' own work. Untested proposal; Open Educational Resources for licences
The brief names its own routine, rubric or check-in Keep it; change only what the evidence speaks to: add anchor-sample calibration to its rubric, a planning prompt to its check-in, or a segment-length limit to its videos. If its rubric rewards use of a tool, keep that criterion and add one unaided check beside it. Rubric-guided peer rating [~M] for the calibration; the rest untested
The brief wants a whole performance or product kept whole Submit and review the whole performance (one recorded presentation, one full design), not split into quiz items; the rubric's criteria divide the judgement, not the task. Untested proposal
The brief reports an error learners actually made Use it as a worked contrast in the next week's first segment: the error, why it is tempting, and the correct version side by side; add one question on it. Untested proposal

Eleven of the twenty-one rows cite a claim (novice learners, professionals, embodied skills, complex written work, automated scoring, collectivist cultures, adults at work, the connectivist aim, phones only, low-confidence learners, and the brief's own rubric), each carried beyond the population or setting its claim page reports, as its Basis cell says; the phone-only and low-confidence rows cite claims that bear on motivation or experience, not on the design move, so those moves are untested; the other ten rows are untested proposals.

Choosing configurations from evidence

Do not rank a MOOC against a campus or paid online course from these claims: none compares them, and none measures learning in a MOOC against an alternative. Do not report completion over all registrants as a course's effectiveness, or engagement in logs as learning. A precise abstention names the missing comparison (the same content as a MOOC with and without a named substitute for teacher time, on the individual final task among learners who intended to complete) and the local observation that would stand in for it.

The evidence by question

Further evidence, not yet read against this model

Claims the earlier page cited, and neighbouring claims, that are not core evidence above. Each says what it is and how far it bears on this page.

Illustrative design instance and observation record

Invented illustration, not a tested course. A university offers a six-week open course on reading clinical research for health workers worldwide; 20,000 register; one lecturer has four hours a week and two former learners volunteer as mentors. At enrolment, 40% say they want one or two modules, 35% the certificate, the rest are looking around. The table changed the default: segments of 4–6 minutes with a talking head beside the slides, each followed by three questions; an eight-minute planning activity in week 1 that names the learner's shift pattern; weekly prompts with model answers in weeks 1–3; study groups by time zone and language, with a facilitator guide for the three hospitals that take it as a group; a final task (appraising one short trial report against a four-criterion rubric) peer-reviewed by three calibrated peers, with the lecturer spot-checking one in ten, and the certificate resting on a sample of finals re-scored by staff. No automated scoring of open text is used, because many learners write in a second language. A nurse says she wants to know whether to trust the drug leaflets on her ward; the designer's objective is appraisal of trial reports. They agree to add one leaflet as a practice item in week 5. Completion is reported among the 35% on the certificate route, with the week each learner stopped. None of the claims cited tested this population, course or outcome; these are local choices to observe.

Record: registrant's stated purpose and route → device and time available → segments watched and questions answered by week → planning-activity text and whether its obstacle recurred → prompts answered → peer ratings against staff spot-checks by criterion → automated scores against staff scores by language background, if used → week each learner stopped and the message sent → final-task performance by route → candidate interpretations (browsing, access, time, missing prerequisite, unclear content) and their basis → re-observation on the next run. Preserve each learner's stated purpose and any change in it. An agent may summarise this record but must not fabricate missing inputs or treat completion over all registrants as the course's effect.

Elements and limits

Open Educational Resources (5Rs) (when the content is openly licensed), feedback, discussion prompt and discussion. Theory: Connectivism. Principles and patterns the course draws on: Self-Regulated Learning, Self-Directed Learning, Segmenting, Multimedia Learning, Digital Learning, Structured Peer Review and Online Course Design.

This pattern is scoped to open courses at a scale where teacher time per learner is negligible. With a tutor for each small cohort, Online Course Design applies; with regular in-person time, Blended Learning. The policy supports design and observation; whether a MOOC built this way teaches as much as the same content taught another way, for which learners, at what cost and with what completion among those who intend to finish, remains untested in this wiki.

Claims

  • Online Course Design (Community of Inquiry) — shares concern for interaction design, but MOOC scale makes deep learner-instructor interaction impractical in a way a smaller online course does not face
  • Online Course Design (converted) — owns the model for an online course with a teacher behind it; this page decides what replaces the teacher's individual time at open scale
  • Structured Peer Review — the review-and-revise cycle used here as the scalable substitute for teacher feedback on open work
  • Self-Directed Learning and Self-Regulated Learning — the sequences for learner-directed study and plan–monitor–act support that a MOOC depends on
  • Blended Learning — when a campus or employer wraps a MOOC with in-person sessions

Examples

Key Sources

Page metadata
Field Value
type pattern
id massive-open-online-course
title Massive Open Online Course (MOOC)
description An open course at a scale where no teacher can answer each learner is expected to serve registrants who can direct their own study only if it replaces teacher time with short segments and self-checks, a planning activity and recurring prompts, calibrated peer review and audited automated scoring, and judges itself by each learner's purpose rather than completion over all registrants; no claim tests a MOOC against another course form, and the nearest tested components are an if-then planning activity that raised completion for some online learners and rubric guidance for peer raters.
status review
canonical True
generated by claude/unspecified on 2026-10-15
trust unverified — no one has confirmed this page's content

Frontmatter is the page's machine-readable record — see the Schema & Guide.