Diagram Model Choice¶
Artifact — instantiates Representation Fit Selection
Chooses among diagrams such as flows, dependency graphs, state diagrams, causal loops, architectures, or process maps based on the relation the task needs to preserve.
Diagram Model Choice treats a diagram not as a picture but as a claim about which relation matters. A flowchart asserts that sequence is the point; a dependency graph asserts that it is coupling; a state machine asserts that it is legal transitions between conditions; a causal-loop diagram asserts that it is feedback. The mechanism's defining move is to pick the diagram genre whose native structure carries the one relation the task depends on — and to reject genres that would render the situation legibly while quietly dropping that relation. It is a choice among static, notational forms; the object produced is a drawing you read, not a thing you operate.
Example¶
An insurance company is redesigning how disputed auto-claims move through adjudication, and the modeling team's first instinct is a swimlane process map — boxes and arrows showing the claim passing from intake to adjuster to reviewer. It reads cleanly. But the actual pain the redesign must fix is that claims bounce: a claim can be "under review," kicked back to "needs documents," escalated, reopened after closure, and re-escalated — and the process map's left-to-right flow makes these look like rare detours rather than the dominant behavior. Diagram Model Choice steps back and asks what relation the redesign must preserve: not the happy-path sequence, but the set of states a claim can occupy and which transitions between them are legal. That names the fit as a state diagram. Redrawn as states-and-transitions, the reopen-and-re-escalate loop becomes the visible center of the picture, and the team can see exactly which transition to forbid. The process map wasn't wrong; it preserved the wrong relation for this task.
How it works¶
The method starts from the relation to be preserved, then maps candidate diagram genres to the structure each is built to carry: flows for ordered sequence, dependency and call graphs for coupling, state machines for condition-and-transition, causal loops for feedback and delay, layered architecture diagrams for containment and interface. The selection rule is explicit — pick the genre whose primitives are the task-relevant relation, and reject genres that can only depict it as an afterthought. Where a familiar default (the swimlane, the org-chart tree) is in play, it must be beaten by a materially different candidate, not merely restyled. The output is a single diagram type committed to, with a stated reason of the form "this genre makes the X relation primary."
Tuning parameters¶
- Candidate breadth — how many genres are seriously considered; wider guards against default capture but costs deliberation time.
- Relation priority — which single relation wins when several matter; sharper choice yields a cleaner diagram but forces one structure to the foreground and others to the margin.
- Notation formality — free sketch versus a strict notation like UML or BPMN; formality buys precision and tool support at the cost of reader literacy.
- Abstraction grain — how much the diagram collapses into each node; coarser reads faster but hides sub-structure a task might need.
When it helps, and when it misleads¶
Diagram Model Choice helps most exactly when a situation is being defaulted into the first familiar box-and-arrow form — it forces the question of which relation the drawing is really asserting, and swaps in the genre that asserts the right one.
Its failure mode is choosing on aesthetic or notational fluency: the genre the modeler knows best, or the one that produces the tidiest layout, wins over the one that fits, and a beautiful diagram then encodes a subtly wrong claim about what matters.[1] A causal-loop diagram drawn because it looks sophisticated can imply feedback that the system does not actually have. The guarding discipline is to state, before drawing, the single relation the diagram must make primary, and to check the finished picture against it — does the structure the eye sees first match the structure the task cares about?
How it implements the components¶
candidate_representation— the diagram genres (flow, state, dependency, causal loop, architecture) are the materially different candidates compared.preserved_feature— the chosen genre is the one whose primitives carry the task-critical relation (sequence, coupling, transition, feedback).representation_selection_rule— the explicit rule "pick the genre that makes relation X primary; reject genres that only depict it incidentally."
It stops at choosing a static form to read; it does not implement distortion_check, validation_case, or interpretation_context — those belong to Prototype Representation, its nearest artifact twin, which builds a form you operate and tests it against real behavior rather than reasoning about which relation to draw.
Related¶
- Instantiates: Representation Fit Selection — a diagram instantiates the archetype only when the genre is chosen for the relation it preserves, not for style or habit.
- Sibling mechanisms: Dashboard View Design · Graph–Table Comparison · Map–List Choice · Mathematical Model Selection · Multi-View Model · Narrative Representation · Prototype Representation · Schema Selection
Editorial Notes¶
Form Classification¶
Form family: Decision, Gate & Allocation
Rationale: Diagram Model Choice operates as a case-specific gate, selection, routing, prioritization, or resource disposition because it chooses among diagrams such as flows, dependency graphs, state diagrams, causal loops, architectures, or process maps based on the relation the task needs to preserve.
Independent corroboration: The frozen evidence defines Diagram Model Choice as 'Chooses among diagrams such as flows, dependency graphs, state diagrams, causal loops, architectures, or process maps based on the relation the task needs to preserve', so its operative form is Decision, Gate & Allocation.
Nearest alternative: Representation, Specification & Plan — The mechanism makes a bounded choice of diagram genre; the selected diagram is the downstream representation.
Review outcome: Independent reviewer agreement; medium confidence.
Origin Attribution¶
Primary origin: Human-Computer Interaction
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Information visualization cohered choosing a diagram notation by whether its visual primitives preserve the relations required for the user's reasoning task.
Related originating lineages:
- Cognitive Science — Perceptual and cognitive-semantics research supplied principles for matching visual variables to represented relations.
- Computer Science & Software Engineering — Software and information-systems modeling supplied standardized notations for state, flow, dependency, process, and architecture structures.
Review resolution: Moody's Physics of Notations and Mackinlay's presentation-design research frame notation choice around cognitive effectiveness and task-relevant visual encoding. That makes HCI and information visualization primary, with cognitive science and software modeling as constitutive lineages.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
- Moody (2009): The Physics of Notations
- Mackinlay (1986): Automating the Design of Graphical Presentations of Relational Information
References¶
[1] The Physics of Notations (Daniel Moody, 2009) — a set of principles arguing that diagram notations are cognitive tools whose visual primitives should map to the semantics they represent, and that "looks right" is a poor proxy for "represents the right relation." The clean warning against picking a diagram genre for fluency over fit. registry ↩