System archetypes

Status: Folded · Evidence: C · Family: Systems and consequences · Verdict: fold (2026-06-11)

Use instead: Causal Loop Diagrams

What it is

System archetypes are a small library of named, recurring feedback-loop structures - Limits to Growth, Shifting the Burden, Fixes that Fail, Tragedy of the Commons, Success to the Successful, Escalation, Drifting Goals, Growth and Underinvestment - each defined by a specific template of reinforcing and balancing loops with delays, a characteristic behavior-over-time signature, and a canonical intervention principle (“for Shifting the Burden, strengthen the fundamental response and wean off the symptomatic one”). The claimed move is diagnosis by template: instead of building the feedback structure of a troubling situation from scratch, you recognize it as an instance of a named pattern and import the pattern’s pre-derived trajectory and prescription. Wolstenholme (2003) formalizes this as paired structures - for every “problem archetype” there is a closed-loop “solution archetype” - and argues the proliferating set condenses to four totally generic archetypes, each an intended-consequence loop coupled to an unintended-consequence loop across an organizational boundary.

The honest description has to name what an archetype literally is: a small, pre-signed causal loop diagram with a name and a prescription attached. Senge’s appendix in The Fifth Discipline draws every archetype as a causal loop diagram; Kim and Anderson’s workbook teaches each one by walking its R and B loops; Braun’s reference document presents each as a CLD plus its behavior signature plus its management principle. The archetype library adds two things to a built CLD: the name (a vocabulary for communicating a dynamic quickly) and the canonical prescription (where the leverage usually is for that structure). Whether those two additions constitute a distinct cognitive move - rather than a reference table bolted onto loop diagramming - is the question the verdict section answers.

When it helps / when it misleads

As a vocabulary, the archetypes help: they compress decades of system-dynamics field experience into nameable patterns, they speed communication (“this is Fixes that Fail” is shorter than re-deriving the side-effect loop), they give a checklist of known failure structures to test a built diagram against, and they are a proven teaching device for introducing feedback thinking. Used as hypotheses - “could this be Limits to Growth? then there should be a balancing loop through some depleting resource; let me check” - they direct the loop-building toward structures worth looking for.

The method misleads when:

• The label substitutes for the structure. Matching a story to “Tragedy of the Commons” without verifying that there is actually a shared, eroding resource loop is superficial analogy, and the field’s own methodologists treat this as the central failure mode. Lane and Smart (1996) showed that “generic structure” is not even a single concept - it bundles three incompatible notions of what transfers between cases - so a label-first match can silently assert a transferability claim the matcher never examined. Wolstenholme (2003) condensed the archetype zoo to four partly because the proliferating, overlapping set confused users about what they had actually identified.
• The canonical prescription is treated as proven. “Where the leverage usually is” is practitioner judgment distilled from cases, not a measured effect. Importing the prescription without checking it against the actual loops inverts the tool’s value.
• The dynamic is genuinely novel, or genuinely linear. A finite template library cannot cover situations whose structure matches no archetype, and forcing the nearest-looking template manufactures false feedback - the same trap the shipped causal-loop-diagrams skill guards against with its “no closed loop found” stop rule.
• Archetypes inherit every CLD limitation. An archetype is a CLD, so the reliability critique recorded in the shipped causal-loop-diagrams dossier (Schaffernicht 2010: different modelers derive different loops and polarities from the same situation) applies with extra force when the loops are not even drawn but merely asserted via a label.

What the evidence says

The honest read is split P/C, and the governing grade is the conservative C: a large practitioner corpus exists, but the entry’s actual claim - that recognizing a named archetype improves diagnosis or locates real leverage - is exactly the kind of claim the field’s own literature says is untested, and the nearest controlled study does not credit it.

What the record supports. The archetypes are a real, durable practitioner tradition with peer-reviewed methodological literature. Senge (1990) coined the term and popularized the catalog; Kim and Anderson (1998) and the Pegasus Systems Thinker toolbox series built a substantial teaching corpus; Braun (2002) is the widely circulated field reference; Wolstenholme (2003, System Dynamics Review) gave the concept its most rigorous formalization (the four-core condensation and the problem/solution archetype pairing). Case applications continue to appear (flood-risk management, real-estate crises, urban water systems). That is a P-shaped pedigree for the archetypes as a descriptive vocabulary.

What the record does NOT support. No controlled study shows that archetype recognition improves judgment, diagnosis, or outcomes:

• Doyle (1997, System Dynamics Review) argued that claims that systems-thinking interventions (the archetype-and-CLD toolkit among them) improve the quality of thought were essentially untested, and called for the controlled cognitive research that mostly still has not arrived for archetypes specifically.
• The nearest controlled test cuts the other way. Green, Molloy and Duggan (2022, Sustainability 14(1): 394) ran a randomized controlled trial (n = 106) with systems thinking and system-dynamics simulation as separate factors, using two sustainability problems that share a common archetype precisely to test pattern-transfer. Simulation alone significantly increased learning outcomes; the transfer support that appeared was attached to simulation (with the archetype-structured material) and was weaker. The archetype-framed systems-thinking presentation was not the significant driver on its own.
• Lane and Smart (1996, System Dynamics Review 12(2): 87-120) is a direct conceptual critique: the “generic structure” idea underlying archetypes bundles three distinct, incompatible views of structural transferability, and the apparent unity of the concept is superficial. This is counter-evidence on the validity of the core move, not merely on its testing.
• Adoption is thinner than the fame suggests: Bureš and Racz (2016, Journal of Business Economics and Management 17(6): 1081-1096) surveyed 54 senior managers and found archetypes largely unknown and unused in practice, even though the managers recognized the dynamics from daily work.

Transfer caveat (required). All of the above is human-subject or conceptual work; none of it validates archetype matching performed by an AI agent. No effect size is quoted in this dossier because none has been verified against a primary source.

Why C and not P. The preliminary registry tag graded this C (“illustrative, not outcome-tested”) and that grade survives the deep run. The practitioner corpus alone would read P, like the shipped iceberg-model. But this entry’s stated move is the recognition-and-leverage claim, and on that claim the record is thinner than P: the method’s own peer-reviewed literature disputes what an archetype match even asserts (Lane and Smart), the improvement claim is flagged as untested by the field itself (Doyle), and the one RCT in the vicinity credits the simulation factor instead. A split honest read resolves to the conservative grade: C.

Why it is / is not a skill here

Verdict: Fold into causal-loop-diagrams. This overturns the preliminary cand / build / C tag - and does so by the test that tag itself set: “deep research must confirm that adds a move, or it folds into causal-loop-diagrams.” It folds.

The Build burden is to name one distinct, durable cognitive move no shipped skill produces and show that no mode of an existing skill produces it. The archetype move decomposes, without remainder, into owned pieces:

• Verifying a match IS the causal-loop-diagrams skill. An archetype is a small signed CLD. To assert “this is Shifting the Burden” honestly, you must identify the symptomatic-fix balancing loop, the delayed fundamental-fix balancing loop, and the side-effect reinforcing loop in the actual situation - which is precisely the shipped skill’s steps (list variables, close the loops, sign the links, label R/B). The field’s own critique (Lane and Smart; the misuse warnings throughout the practitioner literature) says a match asserted without that verification is the failure mode. So the trustworthy version of the move shares its working mechanism with causal-loop-diagrams far above the roughly 20 percent overlap ceiling - the mechanism share is most of the method.
• Reading the behavior signature is the shipped skill’s step 6. “Limits to Growth produces growth-then-plateau” is reading dynamics off which loop dominates - the CLD skill’s existing behavior read-out (spiral, goal-seeking, oscillation).
• The prescription half is already owned by the iceberg-model. “Find the leverage point for this structure” is the Meadows leverage move, and the registry already folded leverage-points into iceberg-model (the leverage-ladder mode). A second skill importing canonical leverage prescriptions would re-ship that fold under a new name.
• The residue is a reference table, not a move. What no shipped skill has is the static library itself: ten named templates with their signatures and prescriptions. That is reference material - exactly the kind of content that lives in a skill’s references/ directory. The natural home is a mode of think-causal-loop-diagrams: after closing and signing the loops, compare the structure against the named archetype table; on a verified match, note the archetype name (for communication) and its canonical intervention principle as a labeled hypothesis. A reference file plus one comparison step is a fold, not a skill.

The steelman for Build - recognition from a story is a different cognitive act than de novo construction, the way differential diagnosis differs from physiology modeling - fails on the field’s own terms: every methodological treatment, from Senge’s appendix to Wolstenholme’s condensation, insists the label is only as good as the loop structure behind it, and the recognition shortcut without construction is the documented misuse. A standalone matcher skill would institutionalize that misuse, and the one randomized trial in the vicinity (Green, Molloy and Duggan 2022) gives no support that the pattern-name framing itself carries the benefit.

Hard walls against the rest of this research batch (the rival-hypothesis / configurational causal cluster), stated for the record because the collision that decides this entry is with the shipped pair, not the cluster:

• vs process-tracing (cand): process tracing adjudicates rival causal explanations of one case by weighing each evidence item’s diagnosticity (hoop, smoking-gun, straw-in-the-wind, doubly-decisive). Archetypes hold no rival hypotheses and weigh no evidence items; they classify a dynamic into a feedback-structure template. No shared mechanism beyond both being causal talk.
• vs qualitative-comparative-analysis (cand/reject): QCA codes multiple cases into a truth table and minimizes to necessary or sufficient configurations. Archetypes match one situation against a pre-built template; there is no cross-case coding and no set logic.
• vs analysis-of-competing-hypotheses (cand/reject, X): ACH is an evidence-by-hypothesis disconfirmation matrix. Archetypes have no matrix and no disconfirmation step, and the randomized ACH record (Dhami, Belton and Mandel 2019, and the later null results) concerns the matrix procedure - it transfers to archetypes in neither direction and plays no part in this entry’s grade.

Why fold rather than recipe or reject: it is not a chain of separable existing moves (recipe), it is one comparison step plus a table inside one shipped skill’s territory; and reject would be less informative than fold, because the vocabulary is genuinely useful and worth locating - the honest service is to point the reader to where the move already lives, exactly as the library did for inversion -> premortem and leverage-points -> iceberg-model. The learning value of the NO: a famous pattern library is not a skill; the skill is building and signing the loops, and the names are an index into structures you still have to verify.

The fold target resolves: causal-loop-diagrams is status: shipped.

Lineage and who to read

The patterns were observed informally by the first generation of system dynamicists - Jay Forrester, Dennis Meadows, Donella Meadows - in the 1960s and 1970s. In the 1980s, Michael Goodman, Charles Kiefer, Jenny Kemeny, and Peter Senge, drawing in part on notes by John Sterman, catalogued them as generic “systems templates”; Peter Senge coined the term “archetype” and made the catalog famous in The Fifth Discipline (Doubleday, 1990), where the archetypes appear as causal loop diagrams with management morals. Daniel Kim built the largest teaching corpus through Pegasus Communications and The Systems Thinker (with Virginia Anderson, Systems Archetype Basics: From Story to Structure, 1998); William Braun’s “The System Archetypes” (2002) is the widely circulated one-document field reference. The academically rigorous treatments are in System Dynamics Review: Lane and Smart (1996) for the critical unbundling of “generic structure,” and Eric Wolstenholme (2003) for the four-core condensation and the problem/solution archetype pairing. Donella Meadows’ leverage-points work is the prescription side’s deeper source (already cited by the shipped iceberg-model). “System archetypes” is the field’s generic descriptive term - no trademark, not branded; attribution to Senge and the system-dynamics tradition suffices.

Named sources

• Peter M. Senge, The Fifth Discipline: The Art and Practice of the Learning Organization (Doubleday, 1990). Coined “archetype”; the popularizing catalog, drawn as causal loop diagrams. Foundational, practitioner. (P)
• Daniel H. Kim and Virginia Anderson, Systems Archetype Basics: From Story to Structure (Pegasus Communications, 1998). The teaching workbook; defines each archetype by its R/B loop template. Practitioner. (P)
• William Braun, “The System Archetypes” (2002, widely circulated reference document). Each archetype as a CLD plus behavior signature plus management principle. Practitioner reference. (P)
• David C. Lane and Chris Smart, “Reinterpreting ‘generic structure’: evolution, application and limitations of a concept,” System Dynamics Review 12(2) (1996): 87-120. Identifies three incompatible views of what a generic structure is and what transferability means; argues the concept’s unity is superficial and needs unbundling. Counter-evidence on the core move’s conceptual validity. (peer-reviewed critique)
• Eric F. Wolstenholme, “Towards the definition and use of a core set of archetypal structures in system dynamics,” System Dynamics Review 19(1) (2003): 7-26. Condenses the archetypes to four generic problem/solution pairs across organizational boundaries; the most rigorous formalization, motivated by the confusion of the proliferating set. (peer-reviewed method)
• James K. Doyle, “The cognitive psychology of systems thinking,” System Dynamics Review 13(3) (1997): 253-265. Argues claims that systems-thinking tools improve thinking were untested and prescribes controlled experimental research. Supports the under-tested grade. (peer-reviewed critique)
• Cathal Green, Owen Molloy and Jim Duggan, “An Empirical Study of the Impact of Systems Thinking and Simulation on Sustainability Education,” Sustainability 14(1): 394 (2022). RCT, n = 106, systems thinking and simulation as separate factors, two problems sharing a common archetype to test transfer; simulation, not the archetype-framed systems-thinking factor, drove the significant gains. The nearest controlled evidence; does not credit archetype recognition. (M, for what it measured - not transferable to the archetype move)
• Vladimír Bureš and Filip Racz, “Application of system archetypes in practice: an underutilised pathway to better managerial performance,” Journal of Business Economics and Management 17(6) (2016): 1081-1096. Qualitative study of 54 senior managers: archetypes largely unknown and unused despite the dynamics being recognized. Adoption evidence, thinner than the method’s fame. (field survey)

Excluded under the evidence rule: no effect size for archetype training or archetype recognition is quoted anywhere in this dossier, because none traces to a primary source measuring the archetype move itself. Senge’s “elegant simplicity … leverage” claims are practitioner assertion, counted as P-corpus, not as outcome evidence.