Model Thinking

Model Thinking

Response Modes

| Mode | Trigger | Output | |------|---------|--------| | Guided | Ambiguous problem | Diagnostic questions → model recommendations | | Direct | Clear problem or specific model requested | Structured multi-model analysis | | Teaching | Wants to learn models | Model explanation + example + practice |

Workflow

1. Classify: Decision? System? Strategy? Data? Learning?
2. Select mode: Ambiguous → Guided | Clear → Direct | Learning → Teaching
3. Apply 2-3 models: Primary insight + complementary views + blind spot check
4. Deliver: Key insights → Recommendations → Caveats

Reference File Selection

| Problem Pattern | Primary | Also Consider | |-----------------|---------|---------------| | Choosing between options | [decisions.md](references/decisions.md) | economics.md, psychology.md | | Understanding complex behavior | [systems.md](references/systems.md) | networks.md | | Interpreting data, prediction | [statistics.md](references/statistics.md) | algorithms.md, risk.md | | Competition, negotiation | [strategy.md](references/strategy.md) | psychology.md, economics.md | | Human behavior, bias | [psychology.md](references/psychology.md) | economics.md | | Connections, influence, platforms | [networks.md](references/networks.md) | economics.md, systems.md | | Computational problem-solving | [algorithms.md](references/algorithms.md) | statistics.md | | Uncertainty, tail events | [risk.md](references/risk.md) | statistics.md, psychology.md | | Acquiring knowledge, skills | [learning.md](references/learning.md) | psychology.md | | Markets, incentives | [economics.md](references/economics.md) | psychology.md, strategy.md | | Cross-domain synthesis, model pairing | [combinations.md](references/combinations.md) | All domain files as needed |

Guided Mode: Diagnostic Questions

When problem is ambiguous, ask 2-3 from relevant domain:

| Domain | Key Questions | |--------|---------------| | Decisions | Reversibility? (能不能反悔？) Time horizon? (影響多久？) Stakes? (賭注多大？) Stakeholders? (誰會受影響？) | | Systems | Linear/non-linear? (結果跟投入成正比嗎？) Feedback loops? (有沒有自我強化或抑制的循環？) Delays? (行動到看見結果要多久？) Boundary? (問題的邊界畫在哪？) | | Strategy | Players? (有哪些參與者？) Game type? (零和還是共贏？) Info asymmetries? (誰知道得比較多？) Incentives? (各方動機是什麼？) | | Data | Sample size? (資料量夠嗎？) Base rate? (一般情況下機率多少？) Selection bias? (取樣有偏差嗎？) Signal vs noise? (訊號還是雜訊？) | | Risk | Fat tail or thin tail? (極端事件常見嗎？) Reversible? (損害能恢復嗎？) Ruin possible? (有沒有全軍覆沒的可能？) |

Direct Application Template

When applying models directly:

## Analysis: [Problem Summary]

### Model Applied: [Model Name]
**Core Insight**: [One-sentence key takeaway]

**Application**:
[2-4 bullet points applying the model to the specific situation]

### Complementary View: [Second Model]
[Brief application showing different angle]

### Synthesis
- **Recommendation**: [Specific action]
- **Key Risk**: [What could go wrong]
- **Next Step**: [Immediate action to take]

Teaching Mode Template

## [Model Name]
**One-liner**: [Memorable summary]

**Core Concept**: [2-3 sentences]

**Example**: [Concrete scenario]

**When to Use**: [Situations]

**Common Mistake**: [Key pitfall to avoid]

**Practice Prompt**: [A question for the user to apply this model to their own situation]

Multi-Model Synthesis Example

Problem: Should I accept this job offer?

| Model | Insight | |-------|---------| | Regret Minimization | At 80, would I regret not trying this path? | | Opportunity Cost | What salary/growth/learning am I giving up? | | Reversibility | One-way door or can I return to current field? | | Second-Order | How does this affect family, health, skills in 5 years? |

Synthesis: High regret potential + acceptable opportunity cost + reversible → Accept

Use 2-3 models from different domains to triangulate. Agreement = confidence. Disagreement = complexity worth exploring.

Critical Checks

Before finalizing any analysis:

1. Inversion: What would make this analysis wrong?
2. Base Rate: What typically happens in similar situations?
3. Incentives: Who benefits from each outcome?
4. Second-Order Effects: What happens next after the first-order effect?
5. Falsifiability: How would we know if we're wrong?

Quick Reference: 10 Universal Models

> Detailed explanations and application examples for each model are in the reference files listed in the [Reference File Selection](#reference-file-selection) table above.

| Model | One-liner | Apply When | |-------|-----------|------------| | Inversion | Avoid stupidity rather than seek brilliance | Any decision | | Second-Order Thinking | Then what? | Evaluating consequences | | Opportunity Cost | What are you giving up? | Resource allocation | | Base Rates | Prior probability matters | Any prediction | | Feedback Loops | Effects become causes | System analysis | | Margin of Safety | Build in buffers | Risk management | | Incentives | Show me incentive, I show you outcome | Analyzing behavior | | Map vs Territory | The model isn't reality | Any model use | | Sunk Cost | Past costs are irrelevant | Decision-making | | Explore/Exploit | Balance new vs known | Resource allocation |

For all models organized by domain, load reference files above. For multi-model combination strategies and cross-domain examples, see [combinations.md](references/combinations.md).

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: kcchien
• Source: kcchien/model-thinking
• License: MIT

Install and usage instructions live in the source repository linked above.