Cognitive apprenticeship holds that traditional apprenticeship worked because the work was visible, that thinking is not, and that the reasoning therefore has to be spoken aloud or nothing transfers. Collins, Brown and Newman set it out in 1989, and it names the mechanism that the current argument about juniors and AI keeps circling without naming.
The answer, in one line
A framework set out by Collins, Brown and Newman in 1989, arguing that apprenticeship works because the work is visible, that cognitive work is not visible, and that the reasoning therefore has to be spoken aloud, coached and gradually handed over.
Definition#
Cognitive apprenticeship: an instructional framework holding that expertise in cognitive work transfers only when the expert's reasoning is made explicit, through six methods: modelling, coaching, scaffolding, articulation, reflection and exploration.
The observation it starts from#
A weaver's apprentice can watch the weaving. A trainee lawyer cannot watch the thinking that produced the advice, only the advice. The output of cognitive work carries almost none of the process that made it, so a junior who reads a hundred finished memos learns what a memo looks like and not how to decide what goes in one. Traditional apprenticeship did not have to solve this problem, because the craft was on the bench in front of the learner.
The six methods#
- Modelling. The expert performs the task while narrating the reasoning, including the false starts. The narration is the whole of it; silent demonstration of cognitive work transfers little.
- Coaching. The learner performs while the expert observes and intervenes.
- Scaffolding. Support is supplied and then deliberately withdrawn as capability grows.
- Articulation. The learner is made to explain their own reasoning aloud, which is how the expert finds out what the learner actually thinks rather than what they produced.
- Reflection. The learner compares their process against the expert's, which requires that both were made visible.
- Exploration. The learner sets their own problems, which is the step at which they stop being an apprentice.
The clinical version, with staging#
Medicine has built the most explicit machinery for the scaffolding step. Entrustable professional activities define units of work a trainee may perform at progressively lower levels of supervision, from observing, through performing with someone in the room, to performing unsupervised and eventually supervising others. Adoption is wide across health professions education. Validity evidence for the entrustment decisions themselves is mixed, and no other profession has anything comparable, which is a gap rather than a reason to dismiss it.
What AI changes#
Three of the six methods run on the junior producing something imperfect that a senior then corrects. Where the first draft comes from a machine and the junior edits it, the sequence changes in a way that is easy to miss because the output improves.
Modelling is unaffected in principle and tends to disappear in practice, because the senior has less reason to work through anything aloud when a draft already exists. Coaching loses its object: the senior is now correcting the machine's work with the junior's edits on top, and cannot tell which is which. Articulation is the one that collapses outright, because the junior can produce a defensible document without having formed a view, and nobody finds out, including the junior. That is the mechanism behind synthetic seniority, described here from the teaching side rather than the output side.
The adaptation is not complicated and it is not automatic. Ask for the reasoning separately from the document. Ask what the tool proposed and what the person changed and why, which is what the shared prompt review puts on the table. Keep some work unaided, not for the output but because articulation needs something of the learner's own to attach to.
What this does not establish#
Cognitive apprenticeship is a framework argued from observation of traditional apprenticeship, not a tested intervention. It has been widely cited for over three decades and has little controlled evidence behind it, and the six methods have not been separated out and measured against each other. It names a mechanism plausibly and does not demonstrate that following it produces expertise faster or at all.
Key sources
- Collins, A., Brown, J. S. and Newman, S. E. (1989). Cognitive apprenticeship: teaching the crafts of reading, writing and mathematics.
- Shorey, S. et al. (2021). Entrustable professional activities in health care education: a scoping review. Medical Education, 55(11).
Related SuperSkills research#
Explainer · SS-2026-325 · Graded against the published rubric
Hirji, R. (2026). What is cognitive apprenticeship?. The SuperSkills evidence base, SS-2026-325. https://thesuperskills.com/research/what-is-cognitive-apprenticeship. Last reviewed 26 September 2026.
An evidence review by Rahim Hirji, not peer-reviewed research. For a material claim, cite the underlying study as well; every study here carries its own permanent link.
How citations and IDs work