What Is the ICAP Framework?
Explain the ICAP framework by Chi and Wylie (2014)
The ICAP framework, developed by Michelene T. H. Chi and Ruth Wylie (2014) — Educational Psychologist, 49(4), 219–243, classifies learning activity by the kind of cognitive contribution a learner makes. Its four modes are Passive, Active, Constructive, and Interactive. The framework predicts a general ordering of learning outcomes: Interactive > Constructive > Active > Passive. This is a theoretical prediction, not a guarantee that every discussion beats every lecture or that every learner should always work collaboratively.
Define passive, active, constructive, and interactive engagement
The useful question is not whether a student looks busy. It is whether the activity requires them to receive information, manipulate what is presented, generate something new, or build understanding with another person. A page covered in highlights can look productive while revealing little about what its reader can explain or apply.

Explain how cognitive engagement relates to learning outcomes
ICAP classifies engagement by the contribution learners make and predicts how those contributions relate to learning. Use the categories to design an activity and then assess what learners can explain, remember, or transfer.
What Are the Four ICAP Modes?
Passive learning activities
Learners receive information without producing an observable knowledge-generating response, such as watching an explanation or reading a summary.
Active learning activities
Learners manipulate information already supplied, for example by highlighting, labeling, sorting, or navigating a completed map.
Constructive learning activities
Learners generate something beyond the presented material, such as a prediction, causal explanation, analogy, question, or independently created representation.
Interactive learning activities
Learners respond to and extend one another’s reasoning to develop or revise an explanation. Talk or group work alone does not guarantee this kind of engagement.
| Mode | What the learner does | Example | What would count as evidence? |
|---|---|---|---|
| Passive | Receives information without an observable knowledge-producing action | Watches an explanation or reads an AI summary | Can identify what was presented, but understanding has not yet been demonstrated |
| Active | Physically or mentally manipulates information already supplied | Highlights a passage, labels a diagram, sorts provided cards, or clicks through a mind map | Correct selection, annotation, or manipulation of given material |
| Constructive | Produces an idea or representation that goes beyond the supplied material | Explains why a process works, predicts an outcome, creates an analogy, or draws a causal map from memory | An explanation, prediction, question, or relationship not simply copied from the source |
| Interactive | Two or more people build on one another’s contributions to develop or revise an idea | Partners propose explanations, challenge assumptions, and combine evidence to solve a problem | A contribution responds to and extends the other person’s reasoning |
The modes describe the activity, not a fixed personality type or permanent ability. One lesson can move through all four. A learner may first watch a demonstration, label its parts, explain the mechanism, then debate an alternative explanation with a partner.
Two distinctions matter. First, visible activity is not automatically constructive: highlighting is Active when it marks supplied information, while writing a causal explanation is Constructive. Second, talk is not automatically Interactive. A conversation in which students merely take turns repeating answers may not build on anyone’s reasoning.
What Are ICAP Framework Examples in Education?
ICAP framework examples in classroom lessons
Consider a lesson on why a plant wilts.
- Passive: Students read a short explanation of water loss.
- Active: Students label the parts of a supplied diagram of the plant.
- Constructive: Students predict what will happen if water uptake slows, then explain the causal chain in their own words.
- Interactive: Partners compare their explanations, challenge a missing step, and agree on a revised model.
Examples of constructive vs. interactive learning
The plant-wilting example distinguishes an individual explanation (Constructive) from partners who build on and revise one another’s explanations (Interactive).
The goal is not to eliminate Passive or Active activity. Learners sometimes need a clear explanation and a chance to inspect or label a representation before they can construct a useful explanation. The instructional design question is what each step prepares them to do next.
Map common student activities to ICAP modes
| Common activity | ICAP classification depends on… | A stronger version |
|---|---|---|
| Reading or watching | Usually Passive if learners only receive the explanation | Pause and ask for a prediction before revealing the next step |
| Highlighting and annotating | Usually Active when learners mark supplied information | Ask learners to explain why a selected passage supports a claim |
| Making notes | Active when copied; Constructive when reorganized into a new explanation | Close the source and write a short causal account from memory |
| Mind mapping | Active when exploring a finished map; Constructive when creating relationships independently | Hide selected nodes or links and ask learners to reconstruct and justify them |
| Group discussion | Interactive only when contributions respond to and extend prior reasoning | Require each participant to build on, question, or revise a partner’s claim |
How Can Teachers Apply the ICAP Framework?
Redesign tasks to increase cognitive engagement
Use questioning, explanation, and collaborative problem-solving
A productive peer-learning sequence begins with individual thought. Each learner commits to an answer, explains the reasoning, listens to an alternative, and revises or defends the original claim. The initial response makes reasoning visible; discussion supplies competing explanations; a second vote or new problem tests whether understanding changed.
Use a simple routine:
- Present a conceptual question that requires reasoning rather than recall alone.
- Ask everyone to answer independently before discussion.
- Have partners explain why they chose their answers.
- Ask each partner to identify the strongest evidence for the other person’s view.
- Vote again and explain what changed.
- Finish with a new but structurally related problem.
The final problem matters. Agreement during discussion may reflect persuasion, imitation, or genuine understanding. A transfer question helps distinguish these possibilities.
Align activities with learning objectives and assessment
Choose an engagement mode based on what learners should be able to explain or do independently. Use assessment to check the intended outcome, including delayed recall or transfer.
Use ICAP alongside related learning frameworks
| Framework | Main question | How it complements ICAP |
|---|---|---|
| Generative learning theory | How do learners select, organize, and connect information to existing knowledge? | Explains why producing an explanation or representation can be more meaningful than copying one |
| Generation effect | Is information remembered differently when learners produce it themselves? | Focuses more narrowly on memory benefits of generating responses |
| Levels of processing | How does the kind of processing affect later memory? | Helps distinguish superficial manipulation from meaningful semantic processing |
| Dual coding theory | How can verbal and visual representations support learning? | Helps design representations that learners can inspect, explain, and reconstruct |
| Scaffolding in education | What support helps a learner perform a task they cannot yet complete alone? | Guides how to support and gradually fade constructive or interactive activity |
| Bloom’s taxonomy | What kinds of learning objectives or cognitive demands are being targeted? | Helps specify outcomes; ICAP instead classifies the learner’s engagement with the material |
ICAP is not a replacement for direct instruction. A concise explanation can supply essential information, especially for novices. Nor does ICAP replace learning objectives, assessment design, or subject expertise. It helps answer a narrower but useful question: what is the learner being asked to do with the information?
Self-explanation: a practical bridge from Active to Constructive
Self-explanation asks learners to account for why a step works, how a claim follows from evidence, or how new information relates to what they already know. It can turn a worked example from something to inspect into something to reason about.
For example, after reading a worked algebra solution, do not ask only, “Can you follow these steps?” Ask, “Why is this operation valid here, and what would break if we used it one step earlier?” The learner must produce a reason that is not supplied by the worked example itself.
A useful explanation rubric has four checks:
- Accuracy: Is the claim correct?
- Causality: Does the learner explain why the result follows?
- Connections: Does the explanation link the new idea to relevant prior knowledge?
- Transfer: Can the learner use the same principle in a different example?
Do not reward length alone. A short, accurate causal explanation is more informative than a page of copied terminology.
Use ICAP to design a lesson, organizer, or AI tutor
A practical lesson can move learners through the modes without treating them as rigid steps.
| Stage | Prompt or activity | ICAP emphasis | Check for learning |
|---|---|---|---|
| Orient | Provide a short explanation, diagram, or advance organizer | Passive, with a clear purpose | Learner identifies the central question |
| Inspect | Label components, sort examples, or annotate a supplied map | Active | Learner correctly identifies key elements |
| Generate | Predict, explain, complete missing links, or create a causal map | Constructive | Learner produces a defensible idea beyond the source |
| Compare | Discuss competing explanations and revise the model | Interactive | Contributions build on prior reasoning |
| Retrieve | Reconstruct the explanation without the source | Constructive | Accuracy and completeness of unaided recall |
| Transfer | Solve a new problem using the same underlying principle | Constructive or Interactive | Correct application in a changed context |
A compact lesson template
Learning objective: What should learners explain or do independently?
Prior-knowledge check: What prerequisite idea must they already understand?
Initial input: What concise explanation or worked example will orient them?
Generation prompt: What can they predict, explain, or draw before seeing the answer?
Feedback: What evidence will correct errors without doing all the reasoning for them?
Interaction: How will learners compare and build on one another’s explanations?
Assessment: What delayed recall or new problem will show whether learning transferred?
For AI tutors, the design principle is ask before telling, then give feedback. An AI system can invite a prediction, ask the learner to explain a step, compare that explanation with a sound model, and offer a hint targeted at the gap. Simply generating a polished summary for the learner is not evidence that the learner has learned it.
A 2025 study by Simsek and colleagues, reported in the source material, compared AI-generated summaries with transcripts across participant groups of 101 and 215. The summary condition did not show a retention or transfer advantage over transcripts. This finding cautions against assuming that an AI-produced summary automatically improves learning; it does not establish that all summaries or all AI tutoring systems are ineffective.
Adapt support to prior knowledge
Constructive activity requires something to construct with. A novice who lacks prerequisite knowledge may struggle to generate a coherent explanation; an experienced learner may find a fully worked example unnecessarily restrictive. Instruction should therefore adapt to prior knowledge rather than forcing every learner through the same sequence.
A 2006 computer-based electrical-circuit study by Jana Reisslein, Robert Atkinson, Patrick Seeling, and Martin Reisslein compared example–problem, problem–example, and fading sequences. Low-prior-knowledge learners benefited most from example–problem sequencing, while high-prior-knowledge learners benefited most from problem–example sequencing. The overall comparison did not identify one universally best sequence. This pattern is consistent with the expertise reversal effect: support that helps novices can become redundant as expertise grows.
A 2009 study by Gabriele Cierniak, Katharina Scheiter, and Peter Gerjets examined 60 university students learning nephron physiology from integrated or separated text-and-picture formats. Low-prior-knowledge learners performed better with integrated materials. This suggests that a well-designed representation can help learners coordinate unfamiliar information before they are asked to generate more independently. See cognitive load theory.
A useful progression is show → label → explain → solve → discuss → transfer. Begin with enough structure to prevent aimless guessing, then fade prompts as learners demonstrate prerequisite understanding. Restore support when errors reveal a missing concept, not merely because the learner is working hard.
How Does Technology Support the ICAP Framework?
Map digital tools to learning activities
Use discussion forums and collaborative platforms
Discussion tools support Interactive engagement when learners respond to, challenge, and extend one another’s reasoning. A forum that only collects isolated posts may not create this pattern.
Design interactive assignments and formative assessments
Use digital prompts that require predictions, explanations, reconstructed maps, misconception checks, and feedback. Assess the learner’s contribution and later performance, not only clicks or time spent.
How mind maps fit ICAP
A mind map is not automatically a Constructive activity. Viewing a completed map is usually Passive; clicking, expanding, or highlighting supplied nodes is generally Active; creating and justifying links is Constructive; and negotiating a shared map by responding to a partner’s reasoning can be Interactive. Classification depends on the learner’s contribution, not the software feature.
With y2map, an AI-generated mind map can provide an initial structure. To move beyond viewing, hide several nodes or relationships and ask the learner to restore them from memory. Then ask why each connection belongs. Finally, compare the reconstruction with the source and correct errors. For collaborative work, have partners defend different links and revise the map together.
Use a simple checklist when reviewing a learning artifact:
- Did the learner only receive or navigate the information?
- Did they generate an explanation, prediction, question, or new relationship?
- Is that contribution accurate, or does it preserve a misconception?
- Did another person’s reasoning change or extend the learner’s explanation?
- Can the learner reconstruct the idea later and apply it to a new problem?
This checklist is a design aid, not a validated diagnostic instrument. It helps make the intended engagement visible and suggests what to assess next.
Are AI summaries bad for learning?
No. A summary can reduce search costs, orient a learner, and provide a concise first explanation. The limitation is that reading a summary and learning from it are different events. The summary supplies an answer; a generation prompt reveals what the learner can reconstruct, while feedback corrects the gaps.
Try this sequence: read the summary, close it, write three key claims from memory, draw the relationships between them, check the source, and explain one relationship in your own words. For a video lesson, pause before a key result and predict what comes next. Afterward, test the principle on a new example.
The point is not to make every interaction difficult. It is to ensure that passive exposure is followed by an activity that provides evidence of understanding.
Retrieval practice, spacing, and misconception checks
ICAP classifies engagement; it does not prescribe a review schedule. For retention, combine constructive activity with retrieval practice and spaced review. Reconstruct the main explanation without looking, check it against a reliable source, correct errors, and return to it after an interval. A finished map can serve as the answer key, but repeatedly viewing it is not equivalent to reconstructing it.
Use a simple misconception check after the learner’s first attempt:
- Record the learner’s explanation before showing the solution.
- Identify the exact missing, reversed, or unsupported relationship.
- Give a targeted hint or counterexample.
- Ask the learner to revise the explanation.
- Test the corrected idea on a new case and again after a delay.
Track accuracy, explanation quality, delayed recall, and transfer—not clicks, time spent, map size, or confidence alone. Those process measures can be useful, but they do not independently establish learning.
What Are the Benefits and Limitations of the ICAP Framework?
Use engagement as a guide to activity design
Distinguish visible activity from constructive thinking
ICAP emerged from research distinguishing the receipt of information, manipulation of given material, generation of new ideas, and collaborative knowledge building. Chi and Wylie’s 2014 framework organizes these distinctions into a hierarchy.
Research evidence and limits
ICAP brings together research on how different forms of engagement relate to learning. Work on self-explanation showed that learners can benefit from explaining worked examples rather than merely reading them. Later engagement taxonomies distinguished manipulating presented information from generating new ideas. Chi and Wylie’s 2014 framework organized these distinctions into the four-mode hierarchy.
The framework is best treated as a way to classify instructional activities and make predictions, rather than a complete history of education or a universal ranking of teaching methods. Earlier classroom practices varied widely, and the presence of a lecture or discussion alone cannot establish how deeply learners processed the material.
Consider challenges in measuring cognitive engagement
A 1994 study by Alison King examined paired science learning in fourth- and fifth-grade classrooms. Students generated questions and explanations, with a condition that also prompted them to use prior knowledge. The study supports the idea that asking learners to construct explanations can deepen comprehension. The practical lesson is to prompt reasoning, not simply to add more questions.
A 2009 undergraduate genetics study by Michelle Smith, William Wood, Wendy Adams, Carl Wieman, and colleagues examined peer instruction: students answered a conceptual question individually, discussed it with peers, voted again, and then answered a related question. Performance improved after discussion, including when no one in the group initially had the correct answer. Among students who changed from an incorrect to a correct answer, 77% answered a later isomorphic question correctly. This is evidence that peer discussion can support learning and transfer in that setting; it does not mean all group work is effective.
A broader 2014 meta-analysis by Freeman and colleagues synthesized 225 studies of undergraduate STEM courses. Active-learning courses showed performance about 0.47 standard deviations higher on examinations and concept inventories than traditional lecturing, and failure rates were higher under traditional lecturing (33.8% versus 21.8%). This supports active learning broadly. It is not a direct test of the entire ICAP hierarchy: “active learning” in those studies covered a range of methods and should not be equated with ICAP’s specific Active mode.
Research on note-taking also cautions against equating quantity with quality. In three experiments reported in 2014, Pam Mueller and Daniel Oppenheimer found that laptop note-takers recorded more words, while longhand note-takers performed better on some conceptual questions. The finding is consistent with a distinction between transcription and constructive processing, but it should not be turned into a blanket claim that laptops always impair learning.
Taken together, the evidence supports a practical principle:
design for meaningful learner contributions and assess what learners can do afterward. The evidence does not justify treating the hierarchy as an inflexible law or assuming that a more demanding activity will help regardless of prior knowledge, subject matter, and support.
A key measurement challenge is that observable actions are imperfect proxies for internal thought. Highlighting, typing, talking, or producing a large diagram does not by itself establish conceptual understanding. Pair activity classification with evidence such as explanation accuracy, delayed recall, and transfer.
Conclusion: Design for the Learner’s Contribution
The ICAP framework offers a practical way to inspect learning activities. Passive reception can introduce a topic. Active manipulation can help learners inspect it. Constructive work asks them to generate an explanation or representation. Interactive work requires people to build on one another’s reasoning. The strongest design chooses the mode that fits the learner’s current knowledge and the objective, then checks what the learner can do afterward.
Use the hierarchy as a design prompt, not a scoreboard. A well-timed worked example can be more useful than unsupported discovery; a discussion is only interactive when reasoning genuinely develops; and a mind map is valuable when learners use it to explain and reconstruct relationships, not merely admire its branches.
References
- Chi, M. T. H., & Wylie, R. (2014). The ICAP framework: Linking cognitive engagement to active learning outcomes. Educational Psychologist, 49(4), 219–243.
- King, A. (1994). Paired science questioning and explanation in Grades 4–5 classrooms.
- Smith, M. K., Wood, W. B., Adams, W. K., Wieman, C., et al. (2009). Peer instruction with individual vote, discussion, and isomorphic transfer question.
- Freeman, S., et al. (2014). Meta-analysis of 225 undergraduate STEM studies (active learning vs. traditional lecturing; not a direct test of the full ICAP hierarchy).
- Mueller, P. A., & Oppenheimer, D. M. (2014). Laptop vs. longhand note-taking (three experiments).
- Reisslein, J., Atkinson, R., Seeling, P., & Reisslein, M. (2006). Example–problem vs. problem–example sequencing in electrical circuits.
- Cierniak, G., Scheiter, K., & Gerjets, P. (2009). Integrated vs. separated text-and-picture formats in nephron physiology (n=60).
- Simsek et al. (2025), as reported in source material — AI summaries vs. transcripts (groups of 101 and 215); no retention/transfer advantage reported.
Frequently Asked Questions
What is ICAP in education?
ICAP is a framework for classifying learning activities by the cognitive contribution learners make: Passive, Active, Constructive, or Interactive. It predicts the general ordering Interactive > Constructive > Active > Passive, while recognizing that outcomes depend on task design, prior knowledge, and support.
What are examples of the ICAP framework?
For a lesson on plant wilting, reading an explanation is Passive; labeling a supplied diagram is Active; predicting and explaining why a plant wilts is Constructive; and partners who revise their explanation in response to each other are Interactive.
What is the ICAP framework by Chi and Wylie (2014)?
Chi and Wylie’s 2014 framework links observable cognitive engagement to predicted learning outcomes through four modes: Interactive, Constructive, Active, and Passive. It is a guide for classifying and designing activities, not a guarantee that one mode will always outperform another in every setting.
What is the difference between constructive and interactive learning?
Constructive learning involves an individual generating an explanation, prediction, question, analogy, or representation beyond the supplied material. Interactive learning requires learners to respond to and extend one another’s contributions so that their shared reasoning develops.
How can teachers apply the ICAP framework to improve classroom learning?
Teachers can start with concise input, ask learners to label or inspect key information, prompt them to explain or predict, and then structure discussion so peers build on one another’s reasoning. Align each activity with the learning objective and check understanding through explanation, retrieval, and transfer.





