Community of Inquiry Framework: The 3 Presences Explained

Go from storyboard to an interactive course in minutes

Start Free

Picture the second week of an online cohort for newly promoted claims team leaders. The discussion board is busy. Each leader has posted a storm claim that went badly, colleagues have replied with sympathy and a similar story of their own, and the thread stops there. Nobody has asked why those customers stopped hearing from the insurer, compared what the three offices do differently or tried anything new in their own team. The group is friendly and active, and it is not inquiring into anything.

I'm an instructional designer at Coursebox, where we make AI employee training software that builds a course from a brief, so I'll mention our product where I tested it. The Community of Inquiry framework from D. Randy Garrison, Terry Anderson and Walter Archer describes what that thread is missing: a problem worth working on, enough trust to disagree, and someone who designs the work and moves the discussion on. Below, I explain the three presences in the originators' own words, why online discussions tend to stall before anyone acts on what they found, and how to plan and check a cohort, with three real programmes, a 6-week plan for a workplace cohort and two free templates.

What is the Community of Inquiry framework?

The Community of Inquiry framework is D. Randy Garrison, Terry Anderson and Walter Archer's model of how a group learns together through discussion and reflection, first published in 2000 for online courses, in which learning comes from the interaction of three interdependent elements: cognitive presence, social presence and teaching presence.

View the full-size diagram

The three authors worked at the University of Alberta in Edmonton, Canada, on a research project about text-based computer conferencing, which the originators' site dates from 1997 to 2001. They set out the framework in a 2000 paper in The Internet and Higher Education. It starts from the claim that "a worthwhile educational experience is embedded within a Community of Inquiry that is composed of teachers and students", and that "learning occurs within the Community through the interaction of three core elements." The short definition on the originators' Community of Inquiry site today is: "An educational community of inquiry is a group of individuals who collaboratively engage in purposeful critical discourse and reflection to construct personal meaning and confirm mutual understanding."

The three elements are called presences:

• Cognitive presence is how far the group builds and tests understanding through sustained reflection and discussion.
• Social presence is how far people come across as real people, communicate with a purpose and trust the group enough to disagree.
• Teaching presence is the design, facilitation and direct instruction that hold the other two together and point them at the intended learning.

The framework is usually drawn as three overlapping circles. In the figure on the originators' site, social and cognitive presence overlap at "supporting discourse", social and teaching presence at "setting climate", and teaching and cognitive presence at "selecting content". The educational experience sits where all three meet.

A few points set the scope:

• Where it came from. The framework was built from university graduate courses that met through text-based, asynchronous computer conferences. The 2000 paper says the project "was situated in the practice of university graduate level programs". Using it for workplace cohorts, as this guide does, is an extension.
• Where it applies now. Garrison wrote in 2024 that while the framework has been applied largely in online and blended learning, "it should not be seen to be limited to such contexts".
• What kind of model it is. It describes how a group learns together, not which outcomes it reaches. Garrison wrote in 2017 that "the CoI framework is a process model and its relationship to actual learning outcomes has remained an open question".
• Where the name comes from. Garrison says the team "borrowed the term community of inquiry from Matthew Lipman", whose classroom idea draws on pragmatist philosophy going back to Charles Sanders Peirce. Cognitive presence draws on John Dewey's idea of practical inquiry, but the framework itself is Garrison, Anderson and Archer's.

Here are the three presences at a glance, with the question each one asks of a course.

How the framework has developed since 2000

The framework grew through a run of papers, most of them by the same small group of researchers.

Some category names changed along the way. In 2000, social presence's categories were emotional expression, open communication and group cohesion, and teaching presence's were instructional management, building understanding and direct instruction. From 2001, teaching presence's categories became design and organisation, facilitating discourse and direct instruction, and the survey calls social presence's first category affective expression. Use the later names when you describe the framework today.

Garrison's papers and books: which to read

• Garrison, Anderson and Archer (2000), the founding paper. Read this one first. Read it in The Internet and Higher Education. It defines the three presences and includes the original coding template.
• Garrison, Anderson and Archer (2001), on cognitive presence. Read it in the American Journal of Distance Education for the practical inquiry model, its four phases and the first coded results.
• Garrison, E-Learning in the 21st Century: A Community of Inquiry Framework for Research and Practice (3rd edition, Routledge, 2017). Read it for the book-length account. Garrison's seven design principles come from it.

For his current positions, read Garrison's editorials on the Community of Inquiry community site. A March 2026 editorial discusses artificial intelligence and shared metacognition.

The three presences, worked through one cohort

Here is the hypothetical example I use from here on. A home and contents insurer has claims teams in three regional offices. Twelve claims team leaders, each leading 8 to 12 claims consultants, join a 6-week facilitated cohort. Most were promoted from senior consultant in the last 12 months. They know the claims system and the policy wording, but none has led an improvement review or a peer discussion, and being a team leader does not show that they can.

The cohort is blended. Each leader spends about 2.5 hours a week: about an hour in a self-paced course, the rest in an online discussion space and three 60-minute live video sessions in weeks 1, 3 and 6. The leaders work in small groups of 4, mixed across the three offices, which stay together for all 6 weeks. One facilitator, a senior claims manager from another office who has been trained to facilitate, spends about 3 hours a week on the cohort.

The problem comes from the insurer's own complaints log. After a severe storm, most complaints about storm-damage claims were about delay and not hearing back, and the three offices handle claims that are "awaiting assessor report" in different ways. So that leaders can work from evidence, the cohort gets a de-identified training extract of 20 storm claims, with the contact timeline, assessor dates and each customer's complaint, plus each office's customer update procedure. Every figure in it is invented for training. Leaders never post customer details from their own teams in the cohort space.

The intended outcome is that by week 6 each leader has trialled one small, documented change in how their team keeps storm-claim customers informed, and can explain the evidence for and against it. It is not a proven drop in complaints, which 6 weeks and 12 teams cannot show, and it is not a certification. The example suits the framework because the leaders are spread across offices, so the cohort has to meet online, the problem has no single right answer, and each leader brings experience the others can question.

Cognitive presence

In Garrison, Anderson and Archer's 2001 paper, cognitive presence is "the extent to which learners are able to construct and confirm meaning through sustained reflection and discourse in a critical community of inquiry" (p. 11). The founding paper called it the element "most basic to success in higher education". It is a process, not a stock of knowledge: the 2001 paper says it "focuses on higher-order thinking processes as opposed to specific individual learning outcomes" (p. 8). It moves through four phases: a triggering event, exploration, integration and resolution.

• In the claims cohort: The 20-claim extract and the question "Why do customers on storm claims stop hearing from us, and what would you change in your team?" open week 2. Leaders explore in weeks 2 and 3, each group builds an explanation in week 4, and each leader trials one change in weeks 5 and 6.
• What to look for in posts: Posts that cite a claim number, connect two observations, or propose a change and test it against the extract. A story with no evidence attached is exploration at best.
• Common mistake: Treating cognitive presence as how much content the course covers. A cohort can read every page and never inquire into anything.

Social presence

The founding paper defined social presence as "the ability of participants in the Community of Inquiry to project their personal characteristics into the community, thereby presenting themselves to the other participants as 'real people.'" Its job is to serve the inquiry: "The primary importance of this element is as a support for cognitive presence, indirectly facilitating the process of critical thinking carried on by the community of learners." In a 2007 review of the research, Garrison moved the emphasis "from socio-emotional support to a focus on group cohesion (from personal to purposeful relationships)" (p. 61). The survey's categories are affective expression, open communication and group cohesion.

• In the claims cohort: Week 1 introductions are tied to the work ("Tell your group about one storm claim you still think about, and why"). The groups of 4 stay together, agree how they will disagree about each other's offices, and see the facilitator post first-person, specific examples.
• What to look for in posts: Leaders addressing each other by name, building on a colleague's point before questioning it, and saying where their own office's practice falls short.
• Common mistake: Treating social presence as icebreakers and fun. The aim is trust that lets people question each other's practice, and Garrison wrote that effective communication alone "is not sufficient for educational purposes" (2007).

Teaching presence

Teaching presence is "the design, facilitation, and direction of cognitive and social processes for the purpose of realizing personally meaningful and educationally worthwhile learning outcomes", in the words of Anderson, Rourke, Garrison and Archer's 2001 paper on teaching presence (p. 5). Its three categories are design and organisation, facilitating discourse and direct instruction, and it "begins before the course commences" (p. 5). The founding paper called it "the binding element in creating a community of inquiry for educational purposes", and said its functions "may be performed by any one participant in a Community of Inquiry; however, in an educational environment, these functions are likely to be the primary responsibility of a teacher."

• In the claims cohort: Design and organisation is the schedule, the claim extract and prompts with deadlines. Facilitating discourse is the facilitator's weekly summary and a probing question when a thread stalls at stories. Direct instruction is a short explainer on root-cause questions, and a correction when a post gets the complaints procedure wrong.
• What to look for in posts: Summaries, questions that ask for evidence, and corrections, from the facilitator at first and then from the groups' rotating leads.
• Common mistake: Reading teaching presence as a facilitator's energy on a video call, or as an AI tutor answering questions. The framework describes people designing, facilitating and directing a group's discussion. In my judgement, an AI tutor that answers one learner's questions can support the design, but it is not teaching presence in the framework's sense. That needs facilitation that responds to the group's discussion, rather than only answering an individual learner.

Facilitation was the weak point in a fully online master's for working doctors in London, where instructors were not always in the discussions.

Case study: Working doctors in an online master's in London

This programme was evaluated through the framework, not designed from it. A two-year postgraduate medical master's at an inner London university ran its first year fully online, with case-based asynchronous discussions and live web sessions. Every learner worked full time, from first-year doctors to established independent practitioners, and the question was whether a fully online course could produce the interaction that deep learning needs.

• What they did: About six months after the first year, the researcher gave the 34-item CoI survey to the cohort and coded 10 asynchronous discussion threads and 1 live session with the framework's coding template.
• Evidence and result: In Aylwin's 2019 study in JMIR Medical Education, 18 of 20 eligible learners answered the survey, with mean ratings of 3.9 for teaching presence, 4.1 for social presence and 4.4 for cognitive presence on a 5-point scale. Five teaching presence items did not reach agreement, mostly on course design and the two feedback items, which faculty acknowledged were "substandard in the early phases". Of 269 codings in the asynchronous threads, 135 were social presence, 82 cognitive and 52 teaching, and only 7% of the cognitive presence codes reached resolution. Cognitive presence "was generally higher in this study when instructors participated in the forum".
• Limitation: It was one small course in its first year, with administrative problems the author reports. The survey was taken once, about six months later, one researcher coded a sample of threads and there was no control group. Nothing was measured in clinical practice.
• Lesson: Busy professionals socialised readily, but discussion rarely reached resolution, and cognitive presence was lower where instructors stayed out. Put a facilitator into every case discussion. Fix feedback turnaround early, because the feedback items were among the lowest rated.

Categories and indicators: what to look for in a week's posts

The founding paper included a coding template with categories for each presence and example indicators, which the authors described as "a template or tool for researchers to analyze written transcripts as well as a guide to educators". The table uses today's category names, example indicators from the 2000 template and the 2001 teaching presence paper, and what each might look like in the claims cohort's posts.

To use it, read one week's thread and note which categories turn up. If a thread is all exploration with no integration, ask for an explanation. If nobody has summarised or probed by midweek, the facilitator needs to step in. The template is not a scoring rubric. Its header says "Indicators (examples only)", and coding was hard even for trained researchers: in the 2001 study, agreement between two coders on three transcripts ranged from a kappa of .35 to .74, where 1 means full agreement.

The four phases of cognitive presence, and why discussions stall

The four phases come from the practical inquiry model in Garrison, Anderson and Archer's 2001 paper. The model has two dimensions: one runs between deliberation and action, the other between perception and conception. Together they link the private world of reflection with the shared world of discussion. The phases are:

1. Triggering event: "an issue, dilemma, or problem that emerges from experience is identified or recognized" (p. 10). Any group member, not only the facilitator, can raise one.
2. Exploration: people exchange information and ideas, moving between their own reflection and the group's discussion.
3. Integration: people connect ideas into an explanation or a proposed solution.
4. Resolution: the group tests a solution "by means of direct or vicarious action" (p. 11).

The paper is explicit that the phases are not a lock-step sequence. They are "the idealized logical sequence of the process of critical inquiry and, therefore, must not be seen as immutable" (p. 9), and a group may make "an intuitive leap apparently shortcutting the logical inquiry cycle" (p. 11). In the claims cohort, a leader's two-week trial can raise a new question and send a group back to exploration in week 6. That is the model working as described, not a failure.

Resolution has a particular meaning at work. The 2001 paper says that "In most noneducational settings, this means implementing the proposed solution or testing the hypothesis by means of practical application" (p. 11). That is why the claims cohort ends with a trial in each leader's own team rather than a final post.

Getting to resolution is the framework's known weak point. The 2001 paper's first coded results came from one week of discussion in one graduate health promotion course: of 24 messages, 8% were coded as triggering events, 42% exploration, 13% integration and 4% resolution, with a third not coded as cognitive presence at all (pp. 19 to 20). That is one small transcript, not a general distribution, but later research found the same shape. Garrison wrote in 2007 that "inquiry invariably has great difficulty moving beyond the exploration phase" and that this "may have more to do with aspects of teaching presence" than with other factors (p. 65). The 2001 paper had already said why: "Often students will be more comfortable remaining in a continuous exploration mode; therefore, teaching presence is essential in moving the process to more-advanced stages of critical thinking and cognitive development" (p. 10).

Experienced professionals stall at exploration too. A doctoral study supervised by Garrison followed college instructors in Calgary who used the framework to redesign their courses.

Case study: College instructors redesigning courses in Calgary

This programme was designed with the framework and then evaluated through it. In 2003, Mount Royal College (now Mount Royal University) in Calgary, Canada, ran a faculty learning community to help instructors redesign one of their own courses for blended delivery, rather than move the content online unchanged.

• What they did: Twelve instructors, chosen from 25 applicants across nine departments including nursing, journalism and biology, met for six 90-minute face-to-face sessions with lab time. Between sessions they discussed online in two groups of six, starting from "take away questions for online discussion", with the coordinator as facilitator. Two trained coders coded six online and six face-to-face discussions for all three presences.
• Evidence and result: In Vaughan's 2004 doctoral thesis at the University of Calgary, "almost two-thirds of the messages (61% and 60%) were coded for exploration" online and face to face, with "virtually no resolution/application phase (1% and 0%)", across 179 online messages and 2,009 face-to-face comments. Online, group cohesion rose from 3% to 71% of the social presence codes between the first and sixth discussions, and direct instruction rose from 13% to 43% of the teaching presence codes. Only one of 21 survey items on the instructors' own teaching changed significantly.
• Limitation: Twelve self-selected instructors at one college took part, the researcher was also the facilitator and there was no control group. Nothing was measured in the redesigned courses or in their students' results.
• Lesson: Even experienced professionals who chose to take part stalled at exploration. The author concluded that "Learning activities should be intentionally designed to attain this phase". Build an apply-it-to-your-own-work task into the plan, and expect the facilitator to move from prompting towards direct guidance as the group matures.

What moves a discussion on is design and direction. Garrison reported that "When questions specifically asked students to engage in practical applications, discussions did progress to the synthesis and resolution phase" (2007, p. 66). In a European online course for teachers, Holmes and Sime found that "Most of the messages coded at these levels [integration and resolution] were in the final reflection activity", where teachers reported back on what they had tried. For the claims cohort, that means an open problem with evidence, prompts that ask for explanations, a week 4 group task and a trial in each leader's own team, with a facilitator who steps in when a thread circles.

How the three presences work together

The presences are not stages. The originators call them "three interdependent elements", and Garrison wrote that "the presences do not operate in isolation. One presence cannot be considered absent of the influence of the others." A course plan that spends week 1 on social presence, week 2 on cognitive presence and week 3 on teaching presence has turned a set of overlapping circles into a timetable.

Each presence still has a role. Social presence supports the inquiry: trust in the small groups is what lets a claims team leader say that their own office's update routine lets customers down. Teaching presence holds the two together: the facilitator's Friday summary and a question about what the extract shows turn a pile of stories into the start of an explanation. And cognitive presence gives the other two something to be about, because a group with no open problem has nothing to inquire into, however warm it is.

Some things do come first in planning, because someone has to choose the problem and write the prompts before the cohort starts. That is the order of design work, not an order learners move through.

Teaching presence is also shared. Garrison has written that its dimensions "are the shared responsibility of the participants in the learning community". In the claims cohort, the facilitator leads at first, and from week 4 each group's rotating lead posts the group's summary.

Garrison's seven design principles, from his 2017 book and summarised on the originators' site in 2019, run across all three presences rather than one at a time:

1. Plan for the creation of open communication and trust.
2. Plan for critical reflection and discourse.
3. Establish community and cohesion.
4. Establish inquiry dynamics (purposeful inquiry).
5. Sustain respect and responsibility.
6. Sustain inquiry that moves to resolution.
7. Ensure assessment is congruent with intended processes and outcomes.

How to use the Community of Inquiry framework to design a cohort

The originators give principles rather than a step-by-step method, so this workflow is my own, built on Garrison's seven principles and carried through the claims cohort.

1. Check that the outcome needs inquiry. Garrison wrote that "a CoI approach (collaborative inquiry) is not required for simple recall of information (surface learning)". If the goal is to learn a fixed procedure, build a course with practice and feedback instead.
2. Choose an open problem and the evidence to work from. The triggering event for the claims cohort is the 20-claim extract and one open question, not a topic such as "customer communication".
3. Plan the resolution first. Decide what each learner will try in their own work and how they will report it, before you write a single prompt. Here, a two-week trial in each leader's team, reported at the week 6 live session.
4. Design the groups and the first week for purposeful trust. Small, stable groups mixed across offices, introductions tied to the work and an agreement on how to disagree.
5. Write prompts for each phase, with deadlines. Exploration prompts ask what people see. Integration prompts ask what explains it and what evidence supports that.
6. Plan the facilitator's week and give them the time. Name the person, set aside time each week (about 3 hours in the claims cohort) and write down what they do each week.
7. Hand teaching presence over as the cohort matures. Rotating group leads take on the summaries.
8. Assess the process you designed. Collect learners' views at the end, alongside the group explanations and trial reports.

The 6-week plan for the claims cohort

Here is the claims cohort as a plan, with one column per presence and the evidence the facilitator looks for each week.

Not every programme can afford one facilitator for each small group. A Commonwealth of Learning course for teachers shows one way to spread teaching presence across thousands of learners.

Case study: Teachers in 132 countries on a Commonwealth of Learning course

This course was designed with the framework. The Commonwealth of Learning, an intergovernmental organisation based in Canada, ran Introduction to Technology-Enabled Learning with Athabasca University ten times between January 2017 and July 2021, for teachers and educators mainly in the global south. The problem was facilitation at scale: open online courses rarely offer it, and the authors call facilitation a key element of a community of inquiry.

• What they did: Teaching presence came in three tiers: a lead instructor on recorded video, an "Inspirer" who gave live presence through informal video, announcements and forum posts, and roving facilitators, who were Athabasca graduate students, at roughly one per 250 active learners. The facilitators' main job became "making connections between learners". Each lesson followed the practical inquiry phases and ended in a "Respond" step on using the idea in the learner's own teaching, and the final task was an activity plan for the learner's own context. Certificates required forum posts as well as quizzes.
• Evidence and result: Ostashewski and Cleveland-Innes's 2022 report for the Commonwealth of Learning counts 18,248 registrations from 132 countries. In the exit survey, 1,610 of 1,851 respondents (87%) agreed with the statement "I felt like I was part of a community", and 73.0% of 1,572 agreed that the roving facilitators supported their learning. The course issued 3,709 certificates, which is 70.3% of the 5,274 fully active learners and 20.3% of all registrations.
• Limitation: The course designers evaluated their own course, exit-survey respondents (about 1,850 of 18,248) lean towards people who finished, the 70.3% uses a denominator the authors chose, and there was no control group. Changes in teaching practice beyond the activity plan were not measured.
• Lesson: When one facilitator per small group is unaffordable, teaching presence can be tiered, and peer-level facilitators can spend their time connecting people in similar jobs. Tie completion to discussion and to a plan for the learner's own work, and report completion against both denominators.

Moving a thread from exploration to resolution

Garrison has written that "Nothing may be more central to the success of a community of inquiry than facilitation." The 2001 paper says the integration phase "requires active teaching presence to diagnose misconceptions, to provide probing questions, comments, and additional information" (p. 10). These are the facilitator moves I would write into the claims cohort's guide:

• Open with the problem, not a topic. Post the week's question and what people should work from, such as the five claims each leader reviews.
• Summarise each week in one short post. Say what the group has found, where people disagree and what is still unexplained.
• Probe when a thread stalls at stories. Ask "What explains the difference?" or "What evidence in the extract supports that?"
• Draw in quiet learners privately. A direct message asking for one observation puts less pressure on someone than naming them in the thread.
• Explain when a misconception would derail the inquiry, and hold back otherwise. Correct a misread procedure straight away, but let the group reach its own explanation of why claims go quiet.
• Hand over as the group matures. From week 4, each group's rotating lead writes the summary.
• Close the inquiry. In week 6, ask each leader what they tried, what the evidence showed and what they will keep doing.

Measuring a community of inquiry: the CoI survey

The CoI survey is a 34-item questionnaire that asks learners how far they agree with statements about the three presences, on a 5-point scale from "1 = strongly disagree" to "5 = strongly agree". It has 13 teaching presence items (design and organisation, facilitation and direct instruction), 9 social presence items (three each for affective expression, open communication and group cohesion) and 12 cognitive presence items (three for each phase). Arbaugh and colleagues reported its development in a 2008 paper in The Internet and Higher Education, which found it "a valid, reliable, and efficient measure of the dimensions of social presence and cognitive presence". For teaching presence, the factor analysis "suggested that the construct consisted of two factors", one for course design and organisation and one for the instructor's behaviour during the course.

Swan and colleagues' companion validation paper reports that "287 students volunteered to complete the survey, yielding a response rate of 43%", from master's and doctoral courses at four institutions in the United States and Canada in summer 2007. Internal consistency was .94 for teaching presence, .91 for social presence and .95 for cognitive presence.

The survey is free to use: its page says "The CoI Survey is an open resource under Creative Commons license", so you can adapt it with credit, sharing any modified version under the same licence. Three things to know before you use it:

• It measures perceptions. Garrison quotes a 2018 review describing it as "a widely accepted instrument for revealing participants' perceptions of a learning experience". A high score shows learners saw the presences, not that they learned.
• Its teaching presence items are about "the instructor". Item 7 reads "The instructor helped to keep course participants engaged and participating in productive dialogue." Where peers share the facilitation, those items miss part of what happened.
• Some items translate well to work. Item 20 reads "I felt comfortable disagreeing with other course participants while still maintaining a sense of trust", and item 34 reads "I can apply the knowledge created in this course to my work or other non-class related activities."

For the claims cohort, I would use a handful of credited items at the end of week 6, with "facilitator" in place of "instructor", and read them next to the groups' explanations and the trial reports. Whether storm-claim complaints fall over the following months is a separate question, which belongs in your Kirkpatrick evaluation.

What course briefs leave out

To see whether people brief online courses with a community in mind, I looked at course requests typed into Coursebox in the 90 days to 7 October 2026: one per person, from English-language browsers, with no uploaded file. Of 266 prompts, 104 were course requests, and 24 of those were clearly training for employees or professionals. Two independent AI readers each classified all 104 against written definitions, and every positive call was re-read in full by hand. Counts combine the two readers' classifications and remain estimates. Checking positive calls does not establish how many requests were missed.

None of all 104 requests asked for facilitator notes or guidance on running a discussion. None of the requests that did ask for discussion or group work was workplace training: they included school lessons, university courses and a hobby forum. Five of the 104 under the broader definition, and 2 under the stricter definition, were built around an open problem or a driving question for learners to work through, and none asked for a facilitator, peer interaction and an open problem together. In the framework's terms, the briefs described content for one learner at a time, which is part of teaching presence's design work, and left out facilitating discourse and the social side. The sample is small and covers requests, not finished courses, and many organisations run live sessions or forums that never appear in a prompt. Still, if you want a cohort rather than a self-paced course, you have to write the problem, the groups, the discussion prompts and the facilitator's role into the brief yourself.

A separate sample of support conversations helps identify practical setup questions. It does not show how customers design their cohorts. In 44 recent support threads about learner interaction, from roughly the last one to three months, the largest themes were instructors reviewing and commenting on learner work (16 threads) and setting up groups or cohorts (10). None asked how to design a discussion or set up peer review.

Is there a fourth presence?

Other researchers have proposed more presences, and the originators disagree about them. None has been added to the framework as its originators present it.

• Learning presence. Shea and Bidjerano proposed it in a 2010 paper in Computers and Education to cover learners' self-efficacy and self-regulation. A 2012 follow-up by Shea and colleagues says it "reflects the proactive stance adopted by students who marshal thoughts, emotions, motivations, behaviors and strategies in the service of successful online learning".
• Emotional presence. Cleveland-Innes and Campbell defined it in a 2012 study of 217 students from 19 graduate courses as "the outward expression of emotion, affect, and feeling by individuals and among individuals in a community of inquiry".
• Shared metacognition. Garrison and Akyol's alternative is a construct, not a presence. The originators' site places it at "the intersection of the cognitive and teaching presence constructs", covering how learners regulate their own and each other's thinking.

Garrison has argued against adding presences. On learner presence he wrote that "this approach is incongruent with the basic premise of a collaborative community of inquiry", and that "the supposed need for learner presence has been addressed with the development of the shared metacognition construct." In the same editorial he notes that Terry Anderson, one of the three originators, supports adding learner presence as a fourth presence.

Community of Inquiry vs connectivism, situated learning and other theories

Two mix-ups come up often. A community of inquiry is not a community of practice: Lave and Wenger's communities of practice are groups who share a practice at work over time, while a community of inquiry is an educational group designed around a problem, usually for a set period. And the four phases of cognitive presence are not the stages of a lesson, such as the 5E model of inquiry-based learning. They describe the thinking a group does in discussion, in no fixed order.

The table is my synthesis for choosing a lens, not a ranking. I found no study comparing these theories head to head.

For the claims cohort, I would use the CoI framework to plan the six weeks, the cognitive apprenticeship model for the root-cause explainer and Gibbs' reflective cycle for each leader's trial log. If your learners sit in a changing network rather than one cohort, start from connectivism. If they learn inside a team's daily work, start from situated learning theory. Kolb's experiential learning theory and social cognitive theory cover individual reflection and learning from models.

Criticisms and limitations of the CoI framework

Critics and the framework's authors agree on one point: most of the evidence is about how learners see the presences and what their discussions contain, not about what they learned.

The sharpest criticism is that few studies measure learning at all. Rourke and Kanuka reviewed the research up to 2008 in a 2009 paper in the Journal of Distance Education.

Rourke and Kanuka concluded that "it is unlikely that deep and meaningful learning arises in CoI" (p. 19), and that "examples of groups of students engaging in a full cycle of cognitive presence have not been documented" (p. 22). Seven framework researchers, Garrison among them, replied that the review's search was selective, that the framework is "first and foremost a process model" (p. 124), and that failure to reach the later phases "very likely had to do with teaching presence issues (design, facilitation and direction)".

Later evidence has the same shape. A 2022 meta-analysis pooled 19 studies with 6,459 participants and looked separately at measured learning and at learners' own ratings of their learning.

In Martin and colleagues' meta-analysis, teaching presence had the strongest link with measured learning (r = .353) and social presence the weakest (r = .199). The links with measured learning are small, and the links with learners' own ratings of their learning are stronger. These are correlations, mostly from surveys, so they show that the presences go with learning, not that designing for them causes it.

The other limitations are narrower:

• Social presence's role is contested. David Annand argued in a 2011 paper in the International Review of Research in Open and Distributed Learning that "social presence does not impact cognitive presence in a meaningful way" (p. 40). Garrison replied on the originators' site that the critique "seems to be more of an argument for independent study (an objectivist paradigm) and against a collaborative, cohort based approach (a constructivist paradigm)".
• Many courses studied are not communities of inquiry. In the same reply, Garrison names "the tendency to consider every online/blended learning environment is a true community of inquiry design when, in fact, there is little teaching, cognitive or social presence".
• It came from graduate university courses. I found no study that measured job performance after a programme designed with the framework, and none of the three programmes described had a control group.
• It costs facilitation time. Without it, discussion tends to stay in exploration.

A study of one corporate setting shows how self-paced courses can fall outside the framework's collaborative scope. In a 2013 study of 462 employees at a Korean electronics, chemicals and communications group, learners took four-week, self-paced job courses built on video lectures, and "There were no online projects or discussions." The researchers left social presence out of their survey because, they wrote, "learners find it difficult to communicate with colleagues and creating social relationships in Korean corporate e-learning environments, where individualized learning prevails". They chose to measure teaching and cognitive presence. Adding a forum to a course like that does not turn it into a community.

The framework is the wrong main lens when:

• nobody can facilitate, because without teaching presence discussion tends to stay in exploration;
• learners cannot work together over weeks, such as staff who join one at a time and finish at their own pace.

Using the Community of Inquiry framework with an AI course builder

Two product details shape a cohort before you build anything. First, a discussion activity is added in the Coursebox editor after the course is generated, not at the outline stage. Second, our pricing page notes that a self-hosted SCORM export (a packaged course file you upload into another LMS) leaves out features that need a live connection, including assignment submissions and discussions, so if your cohort runs from another LMS, the discussion space sits outside the course.

On 8 October 2026 I briefed Coursebox to build the course side of the claims cohort. Coursebox is our product, so weigh my account with that in mind. The brief described the 12 team leaders, the three offices, the groups of 4, the facilitator and the 20-claim extract. It asked for eight sections in order:

• a short "Before you start";
• week 1 introductions;
• the week 2 storm claims problem;
• the week 3 comparison of office practices;
• the week 4 group explanation;
• the weeks 5 and 6 trial and report back;
• a facilitator guide addressed to the facilitator;
• a wrap-up as the very last section.

It also asked for open-text posts and discussion prompts, not multiple-choice quizzes.

Most of it came back as briefed. Week 2 opened as a puzzle from the training extract, with pages called "A Problem Worth Investigating" and "Trace Five Claim Timelines". Week 3 had a page on "Evidence Beyond Stories", week 4 a group explanation, then the trial and the report back, and the wrap-up was last. The course also kept the facilitator's role in the cohort.

The facilitator guide was where it went wrong. The first outline turned it into nine sections of the learners' course, with quizzes addressed to the facilitator, such as "Facilitator Judgment Reflection" and "Facilitator Closing Reflection". The course grew from the eight sections I asked for to 17, with 12 quizzes, and team leaders would have worked through the facilitator's playbook as course content.

One follow-up message fixed it. I asked for one facilitator-only section with a page per week and no quizzes, the wrap-up kept last, and every learner quiz item turned into a discussion prompt where leaders post and reply to two colleagues. Then I reloaded the outline and checked it rather than trusting the chat reply. It showed 8 sections, 28 pages and no quizzes, with the wrap-up still last. Where the quizzes had been, the learner weeks now have reply scenarios, such as "Discuss the claim evidence" in week 2, where a leader practises answering a colleague who reads the evidence differently. That is practice, not a discussion, so the real discussion needs a discussion activity added in the editor or the cohort's own space. The facilitator guide was now one section, with a page for each week from week 1 to weeks 5 and 6 and items such as "When the room stalls", "Drawing in a quiet learner" and "When to explain". I opened the generated week 3 facilitator page, and it spoke to the facilitator, starting "The aim is not to rush the group to a neat answer, but to keep them anchored in what they can actually see in the storm-claim records". One of its moves is to separate "What do we know?" from "What do we think it means?"

That check covers the outline and the generated pages. I did not run the cohort with team leaders, and a facilitator guide is not facilitation: teaching presence and social presence come from the facilitator and the team leaders. With the guide back in one section and the quizzes gone, the course now gives the leaders a problem to work on each week and gives the facilitator a plan for each week.

Putting the Community of Inquiry framework to work

The Community of Inquiry framework answers a practical question: what does a group need in order to learn by working through a problem together? It needs a problem worth inquiring into, enough trust to disagree, and someone who designs the work, keeps the discussion moving and steps in when it circles. Use it when a cohort can work together over weeks on a problem with no single right answer, and when someone has time to facilitate. Use something else for recall or a fixed procedure. The evidence is mostly about perceptions and discussion, so judge a cohort by what learners produce and try, not by how busy the discussion board looks.

Before you write any discussion prompts for your next cohort, write the resolution task: what each learner will try in their own work, and how they will report what happened. If you cannot fill that in, the discussion is likely to stop at exploration.