thinking-in-systems

name	thinking-in-systems
description	Apply Donella Meadows' systems thinking framework to map, diagnose, and redesign any system — organizational, technical, ecological, or policy. Covers stocks/flows, feedback loops, system archetypes, leverage points, and concrete intervention recommendations.
license	MIT
tags	["systems-thinking","analysis","design","leverage-points","archetypes"]
sources	["Meadows, Donella H. Thinking in Systems: A Primer. Chelsea Green Publishing, 2008."]

Thinking in Systems

Apply Donella Meadows' systems thinking framework to any system — organizational, technical, ecological, or policy. The skill runs four phases: map the system's structure, diagnose which system traps are active, rank leverage points by impact, and produce concrete intervention recommendations.

Use this skill when:

• Analyzing why a system keeps producing the same bad outcomes despite repeated fixes
• Designing a new system and wanting to avoid classic failure traps from the start
• Looking for the highest-leverage intervention in a complex, interconnected problem
• Helping a team stop treating symptoms and start addressing root structure

---

Parse Arguments

Extract from $ARGUMENTS:

Argument	Effect
(none)	Ask for system description, then run all four phases
--system "<text>"	Inline system description — skip the prompt
--focus map	Phase 1 only (stocks/flows map)
--focus archetypes	Phases 1–2 (map + archetype diagnosis)
--focus leverage	Phases 1 + 3 (map + leverage points, abbreviated archetype pass)
--design	Design mode: user is building a new system, not analyzing an existing one

---

Phase 1: Map the System

Gather Context

If no system was provided via --system, ask:

Describe the system you want to analyze. Include: what it's trying to accomplish, who the key actors are, what resources or quantities flow through it, and what problem or behavior pattern you're concerned about. Example: "A fishing economy where fleets keep collapsing despite regulations," or "A software team that ships faster but accumulates more bugs with each release."

Produce the System Map

Work through each layer in order. Every layer must have at least one entry or an explicit "none identified."

---

Elements

The visible, tangible parts — people, organizations, physical components, infrastructure. List them in a simple table:

Element	Type	Role in the system
[name]	Person / Org / Infrastructure / Resource	[what it does]

---

Stocks and Flows

Stocks are accumulations measurable at a point in time. Flows are the rates that change them.

Stock	What increases it (Inflows)	What decreases it (Outflows)	Current state
[name]	[inflow 1], [inflow 2]	[outflow 1]	Growing / Shrinking / Stable / Unknown

Delays: Note any significant time lag between a flow and its effect. Delays are a primary source of oscillation, overshoot, and policy failure.

Delay	Between what and what	Estimated lag	Risk
[name]	[cause] → [effect]	[duration]	Oscillation / Overshoot / Both

---

Feedback Loops

Identify every feedback loop and label it:

• Reinforcing (R): A change in the stock causes more change in the same direction. Drives exponential growth or collapse. Label: R — [name]
• Balancing (B): A change triggers corrective action returning the stock toward a target. Drives stability or oscillation. Label: B — [name]

Format each loop as a causal chain:

R — Compound Growth
  Cash in bank → interest earned → more cash in bank → ...
  Direction: self-amplifying (growth or collapse depending on starting sign)

B — Thermostat Control
  Room temperature → gap from setpoint → heater on/off → room temperature
  Direction: stabilizing toward target (with overshoot if delay is large)

---

System Purpose

State what the system actually does — inferred from its behavior pattern over time, not from its mission statement or designers' intent.

The system's real purpose is revealed by what it consistently produces, not by what stakeholders say it's for. A hiring process that consistently selects people from elite universities has "prestige filtering" as its actual purpose, regardless of stated diversity goals.

Write: "This system's actual purpose appears to be: [one sentence]"

---

Worked Example (abbreviated)

System: A fishing economy

Stock	Inflows	Outflows
Fish population	Natural reproduction	Fishing harvest
Fleet capacity	New boat purchases	Boat retirement / bankruptcy
Fishing profit	Revenue from catch	Operating costs

Feedback loops:

R — Fleet Expansion
  Profit → buy more boats → larger harvest → more profit → ...
  Self-amplifying: drives rapid fleet growth in good years

B — Population Recovery
  Fish population → reproduction rate → population grows back
  Stabilizing: self-limiting when fish are abundant

B — Depletion Brake (weak, delayed)
  Fewer fish → harder to catch → less profit → fewer boats
  Stabilizing, BUT: long delay between depletion and fleet reduction;
  by the time profit falls, fish stock may already be below recovery threshold

System purpose (actual): Maximize short-term harvest returns, with population sustainability as a secondary constraint that gets sacrificed whenever it conflicts with profit.

---

Phase 2: Diagnose System Archetypes

If --focus leverage is active: Do a single-pass only — mark each archetype Present/Absent with one line of evidence, skip expanded analysis, then proceed directly to Phase 3.

Check the mapped system against all six archetypes. State: Present / Absent / Suspected. For "Present" entries, identify which elements play which role.

Archetype	How to recognize it	Present?	Evidence
Policy Resistance	Multiple actors pulling the same stock toward different goals, actions cancel each other out, effort is wasted
Tragedy of the Commons	Shared, depletable resource overused because personal benefit > shared cost; no actor has incentive to hold back unilaterally
Drift to Low Performance	Bad past performance lowers the target, which allows more bad performance — standards erode over time in a vicious cycle
Escalation	Two actors each respond to the other's moves with more of the same, creating a reinforcing arms-race loop
Success to the Successful	Winners gain resources that let them win bigger next time; losers are starved of resources; outcome monopolizes over time
Shifting the Burden / Addiction	A quick symptom-fix reduces pressure to solve the root cause; over time the system becomes dependent on the fix and the underlying problem worsens

Additional Traps

Non-linearities: Are there thresholds in the system where a small additional nudge triggers a sudden, large, hard-to-reverse shift? (Stock-recruitment collapse in fisheries; bank runs; ecological tipping points.) Identify any suspected threshold stocks.

Bounded rationality: Are actors making locally rational decisions that produce globally irrational outcomes? Name the actor, the information they lack, and the collective result.

---

Phase 3: Leverage Point Analysis

Meadows' 12 leverage points, ordered least to most powerful. For each point that applies to the mapped system, fill in the intervention and feasibility.

The Leverage Point Paradox: The most commonly targeted interventions (numbers, subsidies, tax rates) are the least effective. The most powerful interventions (paradigm shifts, system goals) are systematically ignored by policy. Flag this explicitly when you see it.

Low Leverage — Constants and Structure (12–9)

Point	What it is	Intervention	Feasibility	Notes
12 — Numbers	Tax rates, subsidies, standards, quotas		High/Med/Low	Most lobbied, least impactful
11 — Buffer size	Size of a stock relative to its flows (large buffers are stable; small buffers are fragile)
10 — Material flows	Physical infrastructure — pipes, roads, factories (expensive and slow to change)
9 — Delays	Time lags between action and feedback			Reducing delays often has high impact despite its "low" ranking

Medium Leverage — Information and Feedback (8–6)

Point	What it is	Intervention	Feasibility	Notes
8 — Balancing loop strength	Strength of negative feedback (e.g., how aggressively a regulator responds to deviations)
7 — Reinforcing loop gain	Speed of a runaway loop (faster = more dangerous; slower = more controllable)
6 — Information flows	Who gets what information when; missing feedback is a major system failure mode			Often high impact in practice — easy to overlook

High Leverage — Rules, Goals, Paradigms (5–1)

Point	What it is	Intervention	Feasibility	Notes
5 — Rules	Incentives, constraints, enforcement mechanisms — who can do what
4 — Self-organization	The system's power to change its own structure, learn, and evolve
3 — System goals	What the system is actually optimizing for			Changing this changes everything downstream
2 — Paradigm	The shared mental model, beliefs, and assumptions from which the system arose			Hard to change; enormously powerful when shifted
1 — Transcend paradigms	Ability to hold all paradigms lightly; not being locked into any one worldview			Rarely actionable, but worth naming

Top Recommendations

After filling the table, highlight the top 2–3 interventions — the highest-leverage points that are also feasible to act on. Format:

RECOMMENDED INTERVENTIONS
─────────────────────────
1. [Point #N — Name]: [one-sentence description]
   Why: [what changes in the system if this lever moves]
   Feasibility: [who can do this and how hard]

2. ...

3. ...

---

Phase 4: Intervention Recommendations

For each recommended intervention from Phase 3, produce a concrete implementation plan.

Intervention Template

## Intervention: [Name]
**Leverage point:** #N — [category]
**System target:** [which stock, flow, loop, or goal this changes]

### What to change
[Specific, actionable description — not "improve communication" but
"create a public dashboard showing fish stock levels updated weekly,
visible to all fleet operators before booking trips"]

### Who holds this lever
[Actor, institution, or role with the authority or capability to make this change]

### Expected behavior change
[What system behavior shifts if the lever is moved — use stock/flow language]
"If [intervention], then [stock] should [increase/decrease/stabilize] because
[causal chain], reducing [problem behavior] within [timeframe]."

### Watch for
- [Likely resistance: who benefits from the current structure]
- [Rebound effects: ways the system might route around the fix]
- [Unintended consequences: second-order effects to monitor]

### Monitoring signal
[The single most important metric or observable that shows this intervention is working.
Must be directly tied to the target stock or feedback loop.]

---

Living in the System Checklist

Meadows closes the book with principles for navigating systems we cannot fully control. After presenting interventions, run this checklist and flag any gaps:

LIVING IN THE SYSTEM
────────────────────
[ ] Get the beat first — watch actual system behavior over time before
    intervening. Data before action.

[ ] Surface mental models — all system maps are simplifications. Write down
    the assumptions in this analysis; they are wrong in some way.

[ ] Distribute information — who in the system lacks feedback they need?
    Hoarded or distorted information is a leading cause of system failure.

[ ] Locate responsibility in structure — avoid blaming individual actors.
    Ask: what incentive structure produced this behavior?

[ ] Design for resilience — redundant feedback loops over pure efficiency.
    What fails if a single connection breaks?

[ ] Stay adaptive — the system will surprise you. Set a review date to
    reassess this analysis after the first intervention has run.

Present the full analysis as a structured report. Offer to drill deeper into any single phase or produce a one-page executive summary on request.

---

Design Mode (--design)

When --design is active, all phases shift from analysis to construction:

Phase 1 (Design): Help the user define the system they want to build:

• What stocks should exist, and at what levels?
• What inflows and outflows will govern each stock?
• Which reinforcing loops are desired (growth engines)?
• Which balancing loops are essential (governors, safety valves)?
• Where should delays be minimized? Where is delay acceptable?

Phase 2 (Design): Proactively flag which archetypes the proposed design is structurally vulnerable to, before it's built. Ask: does this design create a commons? Does it create a success-to-the-successful dynamic? Does it offer a symptom-fix escape hatch?

Phase 3 (Design): Guide toward high-leverage design choices:

Note on coverage: Points 12–7 (numbers, buffers, material flows, delays, loop gains, loop strengths) are shaped by your stock/flow and loop design choices in Phase 1. Phase 3 focuses on the structural and paradigmatic levers (6–1) that are fixed at design time and hardest to change later.

• What information should every key actor receive? (Point 6)
• What rules will govern the system? Are they enforceable? (Point 5)
• What is the explicit goal the system will optimize for? (Point 3)
• Can the system be designed to learn and evolve its own structure over time? (Point 4 — self-organization)
• What paradigm does this system embody, and is that paradigm correct? (Point 2)

Phase 4 (Design): Produce a design spec with explicit resilience mechanisms:

• Named backup feedback loops for each critical balancing function
• Explicit threshold stocks with defined intervention triggers
• Monitoring dashboard specification (what gets measured and by whom)
• Review cadence for reassessing whether the system's actual purpose has drifted from intent

---

Anti-Patterns to Call Out

Flag these explicitly whenever you spot them — whether in an existing system being analyzed or in a proposed design:

Anti-pattern	What it looks like	Correct response
Element fixation	Analysis focuses on visible parts (people, buildings, budgets) while ignoring interconnections	Redirect to flows and feedback loops
Goal displacement	Stated purpose and actual behavior have diverged; metrics being gamed	Name the real purpose; redesign the metric
Delay blindness	Assuming cause and effect are close in time and space	Map every delay explicitly; estimate lag
Single-variable fix	Proposing to change one number while the rest of the structure stays the same	Check for compensating feedback loops that will neutralize the fix
Resilience sacrifice	Optimizing for efficiency by eliminating redundancy	Name what breaks when the single path fails
Paradigm lock	Proposing only technical fixes for a problem rooted in a shared mental model	Escalate to leverage points 2–3
Short-term symptom addiction	Repeated application of a fix that works briefly but worsens the root cause	Name the reinforcing loop being created; identify structural alternative

Flag any anti-pattern you spot throughout the live analysis — not just during this final review pass.