Live web research (subagent, multiple searches + fetches), not a downloaded document. Collected 2026-08-08 for the “Karl” strategy-stack coaching-agent design (Schicht 2/3: complexity theory anchor). Every claim cited inline; inferences marked ⟨Inference: …⟩.
Cynefin (Welsh, roughly “the place of your multiple belongings”) is described by its authors as “a decision support framework, not a method or model,” grounded in bounded applicability: “there are few if any context-free solutions, but many valid context-specific ones” (cynefin.io/wiki/Cynefin). Its core move for questioning is diagnostic-before-prescriptive: you first establish what kind of system you are in, because the type of system dictates what kinds of questions and actions are even legitimate. Getting the domain wrong is the primary error.
1. The five domains, constraints, and decision logics
The framework rests on three fundamental system types — ordered, complex, chaotic — distinguished not by subject matter but by the type of constraint governing cause and effect (cynefin.io/wiki/Cynefin_Domains; chriscorrigan.com). These expand into five named domains:
| Domain | Constraint type | Causality | Decision logic | Practice |
|---|---|---|---|---|
| Clear (formerly Obvious/Simple) | Fixed / rigid constraints (can break catastrophically) | Self-evident, “clear to any reasonable person” | Sense → Categorise → Respond | Best practice |
| Complicated | Governing constraints (laws, procedures; bound the solution space) | Knowable via expertise/analysis | Sense → Analyse → Respond | Good practice |
| Complex | Enabling constraints (allow emergence, self-organization; constraints and system co-modify) | Only knowable in hindsight; dispositional, no linear causality | Probe → Sense → Respond | Emergent / exaptive practice |
| Chaotic | No effective / absent constraints | None discernible | Act → Sense → Respond | Novel practice |
| Confused / Aporetic (center) | Constraint type unknown | — | Break down to lowest coherence, then move | — |
Sources: cynefin.io/wiki/Cynefin_Domains; chriscorrigan.com tour; infoq.com; en.wikipedia.org/wiki/Cynefin_framework.
The question logic per domain is the practical payload. Each verb sequence is literally a different order of inquiry:
- Ordered domains ask first, act last — sense, then categorise (Clear) or analyse (Complicated), then respond. “What does the data mean?” has a right answer discoverable before acting.
- Complex inverts this: you cannot know before you act, so you probe first — run a small intervention — and only then sense what emerged and respond. The governing question shifts from “what is the right answer?” to “what is a safe way to find out?”
- Chaotic collapses the gap: act first to impose any constraint and stabilize, sense the reaction, respond. The only question is “what will restore enough order to make sense possible?”
⟨Inference: the deepest “how to ask questions” implication is that in complex/chaotic domains, analytical, answer-seeking questions are actively harmful because they presume a knowable cause-effect structure that does not exist; the correct questions are generative and experimental, not diagnostic and convergent.⟩
2. The 2020+ refinements: liminality and the aporetic
From March 2020, the central “disorder/confused” domain was reconceived; the most-cited recent development is liminality (from limen, threshold) (cynefin.io/wiki/Cynefin_Domains; agileleanhouse.com).
Liminal zones are transitional bands between domains — states neither one domain nor the other but in suspended transition (agileleanhouse.com). Named liminal areas:
- COCO — Complex→Complicated: still uncertain but transiting toward analysability; you tighten/increase constraints on a narrowing challenge until it becomes properly analysable.
- CHCO / Chaos-Complex — the deliberate, temporary removal of effective constraints; you reduce/relax constraints until “old thought patterns can be left” and fresh sensemaking becomes possible (used for innovation).
- A Complicated liminal where the choice of expertise or analytical method is itself in question.
- Aporia within Confusion itself.
Sources: chriscorrigan.com; natzir.com.
The central domain split — Aporetic vs. Confused:
- Aporetic (Greek aporia, “at a loss”) is a productive confusion — an acknowledged paradox or unresolved tension held deliberately, suspending judgment. It functions as a “triple point” allowing an issue to be broken down “to its lowest level of coherence” and then moved out into one of the four domains (cynefin.io/wiki/Cynefin_Domains; nomotherships.com).
- Confused (the “C” that is unproductive) is the state of “not getting it” — failing to see past one’s own biases, habits, and entrained patterns (medium.com/workmatters).
The movement rule: “If you need to learn something, or discover new things, loosen constraints. If you need to stabilize a situation, tighten constraints” (chriscorrigan.com). ⟨Inference: the aporetic move is itself a questioning discipline — instead of forcing a premature classification, you deliberately dwell in “we don’t yet know which system this is” and decompose the problem until each fragment reveals its own domain.⟩
3. The “cliff” between Clear and Chaotic (complacency collapse)
The boundary between Clear and Chaotic is not a normal border but a “catastrophic fold, or cliff” (cynefin.io/wiki/Cynefin_Domains). Because Clear-domain systems rely on fixed/rigid constraints, over-confidence in those constraints hides the warning signs: “One falls off of the Clear domain into chaos and it is difficult – if not impossible – to recover” (chriscorrigan.com).
The mechanism is complacency collapse: treating something as simple/best-practice-governed breeds rigidity; the liminal warning signals that would normally flag an approaching transition remain invisible, so the fall is a sudden plunge rather than a gradient (cynefin.io/wiki/Cynefin_Domains). ⟨Inference: for question-asking, the cliff is an argument for continually asking “is this really still simple, or have we just stopped looking?” — best-practice regimes suppress the very questions that would detect impending failure.⟩
4. Snowden’s methods and their question logic
Snowden’s methods share a common signature: push interpretation down to the people with direct experience, and strip out the expert/analyst as the meaning-maker.
Anecdote circles — facilitated group sessions eliciting many short, real, lived stories (not opinions or generalisations) around a theme (netikx.org).
SenseMaker (self-signification) — the flagship. Respondents share a brief micro-narrative and immediately interpret it themselves by positioning it on triads, dyads, and “stones” (visual signifier scales). Crucially it “removes ethnographic coding and expert re-interpretation, as participants assign meaning to their own micro-narratives,” making each person “the owner of their own micro-narratives” (cynefin.io/wiki/SenseMaker). Snowden frames this as “both a methodological and ethical advance” (thecynefin.co/ma-3-12). ⟨Inference: this is the most radical “how one asks questions” claim — the expert does not ask about people and code the answers; the instrument asks people to become their own coders, so bias moves from the analyst to a distributed, transparent, auditable signification grid.⟩
Ritual Dissent — a proposal is presented to a group that receives it in silence; the presenter then turns their chair away or dons a mask and listens, taking notes, unable to rebut (cynefin.io/index.php/Ritual_dissent; mspguide.org). Two modes: Dissent (attack the idea hard) and Assent (offer a better alternative). Depersonalisation via the mask/turn prevents “the bland tyranny of premature consensus.”
Safe-to-fail probes — in the complex domain you “construct a safe-to-fail micro experiment around each coherent hypothesis… run them in parallel” (jimruttshow transcript). Canonical illustration: Meals-on-Wheels work near Broken Hill — 13 projects funded at ~A$6,000 each; some flopped, some merged, some combined synergistically (netikx.org).
“We can only understand a system by changing it” — in complex systems there is no vantage point from which to analyse before acting; knowledge only arrives through intervention (jimruttshow).
Rejection of idealised future-state goal-setting — Snowden’s sharpest methodological stance: “The single most fundamental error of the last three decades is to try and design an idealised future state rather than working the evolutionary potential of the here and now, the adjacent possibles” (thecynefin.co/start-a-journey). Instead: “in complexity work we start journeys with a sense of direction; we don’t over-commit to a specific goal.” His slogan contrast: “in systems thinking you define the ideal future state and try to close the gap; in complexity you describe the present and see what you can change.” Calls his own approach “sidecasting” — neither forecasting nor backcasting (jimruttshow). On goals vs. direction: “If you indicate goals, you’ll achieve what you ask for but that might not be what you needed. If you need achievement and you indicate direction, you might discover valuable alternatives you could not have planned for” (infoq.com). Formalised in Estuarine Mapping, which maps constraints/connections of the present without a target end-state (cynefin.io/wiki/Estuarine_framework).
Rejection of best-practice transfer into complex domains — best practice legitimate only in Clear; good practice in Complicated. “Business process re-engineering was universal, learning organization was universal. None of these were universal. They all work within a specific context” (jimruttshow). The Ferrari-vs-rainforest metaphor: a Ferrari (complicated) can be disassembled and reassembled unchanged by an expert; a rainforest (complex) cannot, because it changes when you touch it (medium.com/topology).
5. Snowden’s explicit criticisms of other schools
Systems thinking (Forrester-Senge lineage): the ideal-future-state / gap-closing logic; also an engineering model of the human — “adopts an engineering model of the human brain, which fails to recognize that human sentiment can be radically different” (redefineschool.com). Counter-slogan: “we’re defined by our interactions, not by what we are… changing interactions is a much more successful strategy than trying to change people” (thecynefin.co/learning).
Senge (The Fifth Discipline): “happy clappy” — “part of the new age fluffy bunny thinking” arising because “some things can’t be engineered” (redefineschool.com; thecynefin.co/learning).
Theory U / Scharmer: shares the idealised-future move — “learning from a future that has not yet happened” / “presencing” — for Snowden mistaking an unplannable emergent future for a discoverable destination (leadtogether.org; en.wikipedia.org/wiki/Theory_U). Academic reviews echo that Theory U “falls short of adequately capturing real world complexity” (link.springer.com).
Appreciative Inquiry: “happy-clappy,” “fluffy-bunny” denial of the negative, “no engagement with the real problems of real life — at best distracting and at worst delusional, magical thinking” (stangarfield.medium.com). Procedural objection: consensus around an idealised positive future is easy to fake and easy to game (thecynefin.co/start-a-journey).
⟨Inference: all four criticisms reduce to one target — methods that begin from a desired future and drive consensus toward it — opposed on epistemological grounds (the future is unknowable/emergent) and political grounds (consensus is coerced and gameable). His alternative is to start from a richly-described, dissent-tested present.⟩
6. Serious critiques of Cynefin
(a) Critical Systems Thinking — Michael C. Jackson. Jackson argues Snowden’s “naturalising” approach commits reductionism. Human systems are constituted by “shared appreciations, values, and intentions,” which demand interpretive and emancipatory (power-attending) epistemologies Cynefin’s natural-science stance neglects. Faults Snowden’s insistence that valid social science meet natural-science criteria. Net charge: Cynefin lacks CST’s pluralism and explicit treatment of power, boundaries, and coercion (linkedin.com/pulse — Jackson). ⟨Inference: this is the deepest critique because it attacks the “transparency/relatability” claim from the opposite side — Jackson says pushing interpretation to naturalistic signifiers hides the political and value-laden dimension of whose questions get asked.⟩
(b) Misuse of “chaos” (Murray Robinson; Tom Graves) — Cynefin’s “chaotic” is crisis / no-effective-constraint, not mathematical chaos (ev0lve.medium.com).
(c) The “hidden 2×2” / axis problem — lacks explicit axes; most viewers perceive only two axes and miss the five-domain + liminal structure (complexified.wordpress.com; apppm.man.dtu.dk).
(d) Subjectivity of classification — placing a situation in a domain is itself an interpretive judgment with no objective test, undermining the “transparent” claim (ev0lve.medium.com).
(e) Accessibility / increasing complexity — the ever-growing apparatus criticised as jargon-heavy (complexified.wordpress.com).
(f) Scope dispute — Graves: Cynefin is really a sensemaking tool for complex contexts leaning on narrative enquiry rather than actionable decision logic across all domains (complexified.wordpress.com).
Bottom line for “how one asks questions”
Cynefin’s central discipline: match the type of question to the type of system before asking anything substantive. In ordered domains, converge on the right answer. In complex domains, stop asking for the answer and start asking “what safe experiment reveals it?” — because you can only know the system by changing it. Snowden’s methods operationalise transparency by relocating interpretation from experts to participants (SenseMaker), structuring dissent so it must be heard (Ritual Dissent), and replacing idealised-future consensus with present-state mapping and adjacent-possible steps (safe-to-fail probes, Estuarine Mapping). His unifying polemic — against Senge, Scharmer/Theory U, Appreciative Inquiry, and “systems thinking” — is that all of them start from where we would like to be; his counter-rule is to start from where we are. The serious critiques (Jackson’s CST above all) argue this naturalistic, present-first stance under-theorises human meaning and power, and that classification itself remains an unexamined subjective act.