Model Walkthrough Review¶
Procedure — instantiates Cognitive Representation Externalization
A structured review in which users or stakeholders step through an external representation and test its meaning.
A Model Walkthrough Review takes an already-built representation and marches a group of the right people through it step by step, asking at each point: does this actually mean what its author intended, to the people who must read it? Its defining move is that it is a test of a finished artifact, run from the audience's chair — not a way to create a model or to redraw one, but the inspection that turns a drawing or table into a validated intervention by surfacing where its meaning breaks down for real readers. The walkthrough consumes a representation someone else produced, exercises it against concrete cases or the reviewers' own knowledge, and its output is a record of mismatches: links that are wrong, pieces that are missing, terms that two readers understood two different ways. It does not fix them; it finds them.
Example¶
A transit agency has drawn a network model justifying a proposed light-rail extension, and the planners are sure it is clear. Before the public hearing, a facilitator runs a walkthrough with a mixed room — two engineers, a disability-access advocate, a neighborhood representative, a budget analyst — deliberately assembled so the model is read from every viewpoint that will judge it. They step through it node by node against a concrete question: "a wheelchair user travels from the east side to the hospital at 8 a.m. — trace it." Two stops in, the access advocate stalls: an interchange the engineers drew as a simple transfer is, from her viewpoint, a 300-meter walk with no elevator — a link that is true on the diagram and false in the world. The budget analyst catches a different thing: "operating cost" means annual to him and per-trip to the modelers. Neither is a drawing error; both are meaning failures that only surfaced because the model was walked from the reader's seat. The output is a logged list of these mismatches, handed back for revision.
How it works¶
- Assemble the viewpoints, not just reviewers. The room is chosen so every perspective that must trust the model is present, because a mismatch is invisible until the person who feels it is in the room.
- Walk a concrete path, step by step. Rather than admiring the whole artifact, reviewers trace a specific case or task through it, which forces each element to be interpreted rather than skimmed.
- Test meaning at each step. At every node the question is "what does this say to you?" — divergent readings, not just factual errors, are the target.
- Log every mismatch; resolve none in the room. Each break is recorded as a discrete item; fixing them is a separate act, kept off the review floor so the test stays a test.
Tuning parameters¶
- Reviewer composition — narrow-expert versus broad-stakeholder. Experts catch technical faults; a wider room catches interpretation gaps but spends more time.
- Traversal driver — free inspection versus scripted task-based traversal. Concrete tasks surface real-use failures; free reading is faster but skims.
- Probing depth — accept the first reading versus interrogate every element ("what would make this wrong?"). Deeper probing finds more but fatigues the room.
- Capture formality — informal notes versus a structured mismatch log with owners. Structured capture feeds revision reliably but adds overhead.
When it helps, and when it misleads¶
Its strength is that it is the step that actually validates an externalized model — it is the difference between a diagram someone drew and a diagram a team has confirmed means the same thing to all of them, and it reliably catches the silent interpretation gaps that authors are blind to precisely because they know what they meant.
Its failure mode is theater: a walkthrough run with the wrong people, or as a rubber-stamp after the decision is made, produces the reassurance of review without its substance — and a dominant voice in the room can convert a shared review into a recording of one person's reading. A related trap is critiquing polish (fonts, layout) instead of meaning. The guarding discipline, borrowed from the cognitive walkthrough in usability research,[1] is to anchor every step to a concrete task and a specific reader's viewpoint, and to protect quieter participants so the mismatches they see actually get logged.
How it implements the components¶
shared_review— the walkthrough is the inspection pathway: the structured act of stepping people through the artifact to question and challenge it.mismatch_capture— its product is the logged record of where the representation conflicts with a reader's knowledge, experience, or reading.audience_viewpoint— by seating the actual reading audience and tracing the model from their chairs, it tests the representation against the perspectives that must trust it.
It reviews an artifact it did not build — it does not produce the representation (external_representation) or encode its links (relation_encoding), which are Process Flowchart and Comparison Table, and it does not perform the fix (revision_loop) it hands off to Schema Draft and Whiteboard Modeling Session.
Related¶
- Instantiates: Cognitive Representation Externalization — the walkthrough is the review step that turns a built representation into a tested one.
- Consumes: Process Flowchart, Comparison Table, Schema Draft, or any built artifact — the walkthrough needs a finished representation to step through.
- Sibling mechanisms: Canvas Template · External Memory Aid · Process Flowchart · Schema Draft · Whiteboard Modeling Session · Concept Map · Causal Loop Diagram
Editorial Notes¶
Form Classification¶
Form family: Assessment, Review & Assurance
Rationale: The structured review tests a representation's meaning with relevant users, records divergent readings, and produces a mismatch ledger for correction.
Nearest alternative: Communication, Facilitation & Learning — Stakeholder interaction reveals the evidence, but the mechanism's defining result is an assessment of an existing model's interpretability.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Human-Computer Interaction
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Task-by-task cognitive walkthroughs are a canonical HCI inspection method for testing whether users can interpret and act through an external representation.
Related originating lineages:
- Cognitive Science — Models of attention, knowledge, and action selection provide the psychological basis for walkthrough questions.
- Education & Pedagogy — Think-aloud explanation and guided interpretation contribute evaluative elicitation.
- Engineering & Design — Design reviews and model walkthroughs contribute systematic artifact inspection.
Review resolution: Both independent reviews agree on primary origin human_computer_interaction; reconciliation resolves secondary fields (reported_ambiguity, alternate_origin_disagreement, origin_mode_disagreement, encyclopedia_synthesis_disagreement). Alternate origins retained (education_pedagogy, engineering_design, cognitive_science) are the union of reviewer-supported formative lineages with explicit rationales, not a list of later application domains. Present-day breadth is represented separately as domain_reach=multi_domain; origin_mode=cross_disciplinary_synthesis records the historical relationship among lineages. Confidence is conservatively reconciled to medium, and encyclopedia_synthesis=true preserves either reviewer's finding that the encyclopedia generalized the mechanism.
Attribution caveat: The mechanism is broader than the named cognitive walkthrough method.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Reconciled after independent review; medium confidence.
Notes¶
The walkthrough is where the archetype's whole claim — that externalization "invites correction" — is cashed out or exposed as empty. A model that survives a genuine walkthrough with the right people in the room has been tested; one that was never walked has only been drawn.
References¶
[1] The cognitive walkthrough (Wharton, Rieman, Lewis & Polson, 1994) evaluates a design by having reviewers step through concrete tasks and ask, at each step, whether a real user would understand what to do — the same "trace a task from the reader's viewpoint" discipline applied here to any externalized model. withdrawn registry ↩