SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillscynefin

cynefin

General & other

View repository

Updated 10-15 22:39

About this skill

Cynefin Framework · Situational Awareness Framework

One-line positioning: Before choosing "which decision framework to use," first ask "which situation am I in." The same tool is either life-saving or fatal in different situations.

When to Use (TRIGGER)

  • You are already about to use some decision method, but have a vague sense something is off
  • The team is arguing about "whether to follow the SOP or explore"
  • During a retrospective you discover "the method wasn't wrong, it was used in the wrong place"
  • Thefirst stepof any major decision (meta-level: classify first, then handle)

When Not to Use (ANTI-PATTERN)

  • The decision is tiny and already classified (don't fire up Cynefin to decide "what to have for lunch")
  • You are in a true Chaotic quadrant and must act immediately — skip classification, stabilize first
  • The team isn't familiar with Cynefin, and using it actually raises communication costs — use a simplified "known / unknown" dichotomy

5 Domains (DOMAINS)

        ┌──────────────────────┬──────────────────────┐
        │     COMPLEX          │     COMPLICATED      │
        │ Probe → Sense → Resp │ Sense → Analyze → R  │
        │ 涌现的实践            │ 好的实践（专家）       │
        │ 因果只能事后看         │ 因果可分析            │
        │                      │                      │
        ├──────────────────────┼──────────────────────┤
        │     CHAOTIC          │     CLEAR (Obvious)  │
        │  Act → Sense → Resp  │ Sense → Categorize → │
        │ 新颖的实践            │   Best practice      │
        │ 无因果可分析          │ 标准 SOP              │
        └──────────────────────┴──────────────────────┘
                  ↑ CONFUSED (中心点) — 还没分类清楚的状态

1. Clear (Obvious) — Simple/Obvious Domain

  • Characteristics: Cause and effect are obvious; everyone can see them
  • Method: Sense → Categorize → Respond — recognize the pattern, categorize, apply best practice
  • Example: User reports "can't log in" → check for wrong password → reset password
  • Trap: Treating everything as Clear (when it's actually Complex but you didn't notice), leading to using an SOP on a novel problem

2. Complicated — Complicated Domain

  • Characteristics: Cause and effect exist but require expert knowledge to analyze; there may be multiple correct answers
  • Method：Sense → Analyze → Respond — Bring in expertsto do the analysis
  • Example: Database performance tuning / designing a high-availability architecture / legal compliance
  • Trap: Using Clear's fast decision-making (going with your gut), or jumping up to Complex (pointless exploration)

3. Complex — Complex Domain

  • Characteristics: Cause and effectcan only be seen in hindsight; the system reacts to your actions; the same input doesn't necessarily produce the same output
  • Method：Probe → Sense → Respond — Run small experiments, observe the emergent patterns, and amplify what works
  • Example: Product-market fit / team culture building / entering a new market / agent behavior tuning
  • Trap: Trying to "design the perfect solution and then execute it" (that's the Complicated method, used in the wrong place), or switching direction frequently (not giving emergence time)

4. Chaotic — Chaotic Domain

  • Characteristics: Cause and effect are completely absent or broken; the priority is tostabilize
  • Method：Act → Sense → Respond — Act first— do anything that reduces the chaos, then observe
  • Example: Production incident P0 / company financial collapse / disaster scene
  • Trap: Trying to analyze (no time), or moving too early into Complex exploration

5. Confused (Disorder) — Disorder Domain

  • Characteristics: You don't know which quadrant you're in
  • Method: FirstdecomposeProblem: each sub-problem should be classified separately, otherwise you will force familiar frameworks onto unfamiliar problems.
  • Trap: by default, classifying it into the quadrant you are best at (engineers default to Complicated, PMs default to Complex, bosses default to Clear).

Decision Checklist

  • I stated which domain the current problem belongs to and gave reasons.
  • The tool/method I chosematchesthat domain (not "the method I am most familiar with").
  • If the domain is Complex, I designed experiments rather than a plan.
  • If the domain is Chaotic, I am already executing stabilization actions (not sitting in meetings discussing).
  • I acknowledged that the domainmay change during my handling process(halfway through solving a Complex problem it may become Complicated).

Real Cases

Case 1: The 2008 financial crisis — chaos in different quadrants

  • At the time, most CEOs used Complicated tools (hiring experts to analyze) to handle a Chaotic situation (market disorder), and as a result missed the stabilization window.
  • Snowden directly called it out in an HBR article: leaders applying Best Practice to Complex problems is the biggest failure mode in contemporary decision-making.

Case 2: Engineering Team

  • Login failure bug → Clear (check logs and apply the SOP)
  • Designing a payment gateway refactor → Complicated (bring in a senior person + compare multiple options)
  • Whether a new product AI agent will be accepted by users → Complex (cannot be paper-designed; must do small-scale gray release trials + observe actual user behavior)
  • Production database deleted → Chaotic (first stop writes + start backup recovery, do not hold a meeting first)

The cost of mismatch: treating Complex as Complicated → a perfect 6-month plan for a solution goes live, users do not buy in — because the system reacts to your "design," and the assumptions from the planning phase have already been falsified by reality.

Case 3: Early COVID response

  • January 2020: symptoms unknown → Confused → decompose
  • February: shortage of medical resources, unclear transmission path → Chaotic → forced isolation to stabilize (action before analysis)
  • Afterward: vaccine development → Complicated (clear process, requires experts)
  • Afterward: social recovery → Complex (no one knows the outcome, must experiment)

Relationship with Other Frameworks

  • Cynefin is a meta-framework that determines the application scenarios of other frameworks.：
    • Clear → SOP / checklist
    • Complicated → WRAP + expert analysis
    • Complex → OODA Loop + Pre-Mortem + gray release experiments
    • Chaotic → directly use the OODA Loop fast cycle
  • First Principles is especially important in the Complex/Chaotic domains(because there is no ready-made pattern to apply).
  • Pre-Mortem has the highest ROI in the Complicated domain(because it can be analyzed).

Common Traps

  1. Classifying the domain only once: the domain has already changed during decision-making — response: reclassify at every milestone.
  2. Defaulting to Clear: using the most familiar SOP to handle all new problems — response: question "Is this a new type of problem?"
  3. Not daring to say Confused: forcing a choice of domain to appear professional — response: Confused is a valid state, safer than misclassification.
  4. Treating Complex as Complicated: Spending 3 months to create a perfect PRD before development — Response: Can a 1-week prototype test the core hypothesis?

Reference

  • David J. Snowden & Mary E. Boone, "A Leader's Framework for Decision Making", Harvard Business Review, Nov 2007
  • Cognitive Edge official resource: cynefin.io
  • Snowden's Cynefin was updated multiple times in 2014/2021. This document is based on the 2021 version (Clear replaces the original Simple/Obvious).