hypotheticals-counterfactuals（假设反事实）

Hypotheticals and Counterfactuals

Table of Contents

• Workflow
• Common Patterns
• Guardrails
• Quick Reference

Workflow

Copy this checklist and track your progress:

Hypotheticals & Counterfactuals Progress:
- [ ] Step 1: Define the focal question
- [ ] Step 2: Generate counterfactuals or scenarios
- [ ] Step 3: Develop each scenario
- [ ] Step 4: Identify implications and insights
- [ ] Step 5: Extract actions or decisions
- [ ] Step 6: Monitor and update

Step 1: Define the focal question

What are you exploring? Past decision (counterfactual)? Future possibility (hypothetical)? Assumption to test? See resources/template.md.

Step 2: Generate counterfactuals or scenarios

Counterfactual: Change one key factor, ask "what would have happened?" Hypothetical: Imagine future scenarios (2-4 plausible alternatives). See resources/template.md and resources/methodology.md.

Step 3: Develop each scenario

Describe what's different, trace implications, identify key assumptions. Make it vivid and concrete. See resources/template.md and resources/methodology.md.

Step 4: Identify implications and insights

What does each scenario teach? What assumptions are tested? What risks revealed? See resources/methodology.md.

Step 5: Extract actions or decisions

What should we do differently based on these scenarios? Hedge against downside? Prepare for upside? See resources/template.md.

Step 6: Monitor and update

Track which scenario is unfolding. Update plans as reality diverges from expectations. See resources/methodology.md.

Validate using resources/evaluators/rubric_hypotheticals_counterfactuals.json. Minimum standard: Average score ≥ 3.5.

Common Patterns

Pattern 1: Pre-Mortem (Prospective Hindsight)

• Format: Imagine it's future date, project failed. List reasons why.
• Best for: Project planning, risk identification before launch
• Process: (1) Set future date, (2) Assume failure, (3) List causes, (4) Prioritize top 3-5 risks, (5) Mitigate now
• When: Before major launch, strategic decision, resource commitment
• Output: Risk list with mitigations

Pattern 2: Counterfactual Causal Analysis

• Format: "What would have happened if we had done X instead of Y?"
• Best for: Learning from past decisions, understanding what mattered
• Process: (1) Identify decision, (2) Imagine alternative, (3) Trace different outcome, (4) Identify causal factor
• When: Post-mortem, retrospective, learning from success/failure
• Output: Causal insight (X caused Y because...)

Pattern 3: Three Scenarios (Optimistic, Baseline, Pessimistic)

• Format: Describe best case, expected case, worst case futures
• Best for: Strategic planning, forecasting, resource allocation
• Process: (1) Define time horizon, (2) Describe three futures, (3) Assign probabilities, (4) Plan for each
• When: Annual planning, market uncertainty, investment decisions
• Output: Three detailed scenarios with implications

Pattern 4: 2×2 Scenario Matrix

• Format: Two key uncertainties create four quadrants (scenarios)
• Best for: Exploring interaction of two critical unknowns
• Process: (1) Identify two key uncertainties, (2) Define extremes, (3) Develop four scenarios, (4) Name each world
• When: Strategic planning with multiple drivers of uncertainty
• Output: Four distinct future worlds with narratives

Pattern 5: Assumption Reversal

• Format: "What if our key assumption is backwards?"
• Best for: Challenging mental models, unlocking innovation
• Process: (1) List key assumptions, (2) Reverse each, (3) Explore implications, (4) Identify if reversal plausible
• When: Stuck in conventional thinking, need breakthrough
• Output: New perspectives, potential pivots

Pattern 6: Stress Test (Extreme Scenarios)

• Format: Push key variables to extremes, test if decision holds
• Best for: Risk management, decision robustness testing
• Process: (1) Identify decision, (2) List key variables, (3) Set to extremes, (4) Check if decision still valid
• When: High-stakes decisions, need to ensure resilience
• Output: Decision validation or hedges needed

Guardrails

Key requirements:

1. Plausibility constraint: Scenarios must be possible, not just imaginable. "What if gravity reversed?" is not useful counterfactual. Stay within bounds of plausibility given current knowledge.
2. Minimal rewrite principle (counterfactuals): Change as little as possible. "What if we had chosen Y instead of X?" not "What if we had chosen Y and market doubled and competitor failed?" Isolate causal factor.
3. Avoid hindsight bias: Pre-mortem assumes failure, but don't just list things that went wrong in similar past failures. Generate new failure modes specific to this context.
4. Specify mechanism: Don't just state outcome ("sales would be higher"), explain HOW ("sales would be higher because lower price → higher conversion → more customers despite lower margin").
5. Assign probabilities (scenarios): Don't treat all scenarios as equally likely. Estimate rough probabilities (e.g., 60% baseline, 25% pessimistic, 15% optimistic). Avoids equal-weight fallacy.
6. Time horizon clarity: Specify WHEN in future. "Product fails" is vague. "In 6 months, adoption <1000 users" is concrete. Enables tracking.
7. Extract actions, not just stories: Scenarios are useless without implications. Always end with "so what should we do?" Prepare, hedge, pivot, or double-down.
8. Update scenarios: Reality evolves. Quarterly review: which scenario is unfolding? Update probabilities and plans accordingly.

Common pitfalls:

• ❌ Confusing counterfactual with fantasy: "What if we had $100M funding from start?" vs. realistic "What if we had raised $2M seed instead of $1M?"
• ❌ Too many scenarios: 10 scenarios = analysis paralysis. Stick to 2-4 meaningful, distinct futures.
• ❌ Scenarios too similar: Three scenarios that differ only in magnitude (10% growth, 15% growth, 20% growth). Need qualitatively different worlds.
• ❌ No causal mechanism: "Sales would be 2× higher" without explaining why. Must specify how change leads to outcome.
• ❌ Hindsight bias in pre-mortem: Just listing past failures. Need to imagine new, context-specific risks.
• ❌ Ignoring low-probability, high-impact: "Black swan won't happen" until it does. Include tail risks.

Quick Reference

Counterfactual vs. Hypothetical:

Scenario types:

Pre-mortem process (6 steps):

1. Set future date: "It's 6 months from now..."
2. Assume failure: "...the project has failed completely."
3. Individual brainstorm: Each person writes 3-5 reasons (5 min, silent)
4. Share and consolidate: Round-robin sharing, group similar items
5. Vote on top risks: Dot voting or force-rank top 5 causes
6. Mitigate now: For each top risk, assign owner and mitigation action

2×2 Scenario Matrix (example):

Uncertainties: (1) Market adoption rate, (2) Regulatory environment

Assumption reversal questions:

• "What if our biggest advantage is actually a liability?"
• "What if the problem we're solving isn't the real problem?"
• "What if our target customer is wrong?"
• "What if cheaper/slower is better than premium/fast?"
• "What if we're too early/too late, not right on time?"

Inputs required:

• Focal decision or event: What are you analyzing?
• Key uncertainties: What factors most shape outcomes?
• Time horizon: How far into future/past?
• Constraints: What must remain fixed vs. what can vary?
• Stakeholders: Who should contribute scenarios?

Outputs produced:

• counterfactual-analysis.md: Alternative history analysis with causal insights
• pre-mortem-risks.md: List of potential failure modes and mitigations
• scenarios.md: 2-4 future scenarios with narratives and implications
• action-plan.md: Decisions and preparations based on scenario insights