gh-lyndonkl-claude/skills/synthesis-and-analogy/resources/methodology.md

Advanced Synthesis & Analogy Methodology

Workflow

Advanced Synthesis & Analogy Progress:
- [ ] Step 1: Advanced synthesis techniques for complex sources
- [ ] Step 2: Structural mapping theory for deep analogies
- [ ] Step 3: Cross-domain problem-solving and creative ideation
- [ ] Step 4: Validity testing and refinement
- [ ] Step 5: Integration and advanced applications

Step 1: Advanced synthesis - meta-synthesis, mixed methods, temporal synthesis. See 1. Advanced Synthesis.

Step 2: Structural mapping - systematic correspondence, relational structure. See 2. Structural Mapping Theory.

Step 3: Creative problem-solving - analogical transfer, forced connections. See 3. Analogical Problem-Solving.

Step 4: Validity testing - systematicity, productivity, scope limits. See 4. Testing Analogy Validity.

Step 5: Integration - combining synthesis + analogy, avoiding pitfalls. See 5. Common Pitfalls.

---

1. Advanced Synthesis Techniques

Meta-Synthesis (Synthesis of Syntheses)

When to use: Combining multiple literature reviews or systematic reviews into higher-level synthesis.

Process:

1. Gather existing syntheses: Find 3-10 literature reviews, systematic reviews, or meta-analyses
2. Extract findings: For each synthesis, note main conclusions, themes, effect sizes
3. Compare: Where do syntheses agree? Disagree? What explains differences (population, methodology, timeframe)?
4. Higher-level patterns: What emerges across all syntheses that wasn't obvious in any single one?
5. Synthesize: Create narrative connecting all synthesis-level findings

Example: Meta-synthesis of 5 systematic reviews on "effective onboarding" → Review A (tech): automation reduces friction. Review B (retail): personalization increases engagement. Review C (healthcare): compliance training critical. Meta-synthesis: Effective onboarding = (1) reduce friction (automation/simplification), (2) personalize to role/individual, (3) ensure compliance for regulated industries. All three necessary, priority varies by domain.

Mixed-Methods Integration (Qualitative + Quantitative)

Challenge: Combining narrative/thematic data (interviews, observations) with numerical data (surveys, metrics).

Approaches:

Approach	Description	When to Use	Example
Convergent	Collect qual + quant simultaneously, merge during analysis	Validate findings across methods	Survey shows 70% churn at Step 3 (quant). Interviews reveal "Step 3 too complex" (qual). Integration: Step 3 complexity causes 70% churn.
Explanatory	Quant first (identify pattern), qual second (explain why)	Unexpected quant results need explanation	Metric: Feature X has low usage (quant). Interviews: "didn't know it existed" (qual). Explanation: Low usage due to discoverability, not value.
Exploratory	Qual first (generate hypotheses), quant second (test at scale)	New domain, need to develop measures	Interviews identify 4 user personas (qual). Survey confirms distribution: 40%/30%/20%/10% (quant). Integration: Validated persona model with prevalence data.
Embedded	One method primary, other secondary/supporting	One method dominant, other adds context	Experiment shows Feature A outperforms B (quant primary). User comments explain why (qual secondary).

Temporal Synthesis (Longitudinal Integration)

When to use: Synthesizing data points across time to understand trends, cycles, evolution.

Techniques:

• Trend analysis: Identify directional changes (increasing, decreasing, stable)
• Cycle detection: Look for repeating patterns (seasonal, periodic)
• Event correlation: Link events to outcome changes (did X cause shift in Y?)
• Phase transitions: Identify inflection points where system shifts regimes

Example: Synthesizing 24 months of customer satisfaction data → Month 1-8: declining (65→55%), trigger: pricing change in Month 2. Month 9-16: stable ~55%. Month 17-24: improving (55→70%), trigger: new onboarding in Month 17. Synthesis: Pricing hurt satisfaction short-term (-10pp), plateaued, new onboarding recovered and exceeded baseline (+5pp net).

Cross-Cultural Synthesis

Challenge: Synthesizing findings across different cultural contexts where same phenomenon manifests differently.

Approach:

1. Etic analysis: Identify universal patterns (what's same across all cultures?)
2. Emic analysis: Identify culture-specific manifestations (what's unique per culture?)
3. Synthesis: "Universal pattern X manifests as A in Culture 1, B in Culture 2, C in Culture 3"

Example: Synthesizing user research across US, Japan, Brazil on "collaboration tools" → Etic (universal): All cultures value real-time communication and file sharing. Emic (specific): US prefers async text (Slack), Japan prefers video (face-to-face culture), Brazil prefers voice (WhatsApp voice notes). Synthesis: Build core real-time communication + file sharing (universal), with mode preferences (cultural adaptation).

---

2. Structural Mapping Theory Deep Dive

Gentner's Structure-Mapping Theory

Core principle: Good analogies map relational structure, not object attributes.

Components:

• Entities/Objects: Elements in domains (can differ)
• Attributes: Properties of entities (can differ)
• Relations: Connections between entities (must correspond)
• Higher-order relations: Relations between relations (strongest mapping)

Example:

Source: Solar System	Target: Atom
Entities: Sun, planets	Entities: Nucleus, electrons (DIFFERENT objects)
Attributes: Sun is hot, large	Attributes: Nucleus is positive charge (DIFFERENT properties)
Relations: Planets orbit sun (mass attracts)	Relations: Electrons orbit nucleus (charge attracts) (SAME structure)
Higher-order: More massive planet → elliptical orbit	Higher-order: Different energy electrons → different orbital shells (SAME pattern)

Verdict: Strong analogy because relational structure maps despite different entities/attributes.

Systematicity Principle

Definition: Analogies are stronger when they map interconnected systems of relations, not isolated facts.

Test: Does analogy map single relation or network of relations?

Weak (single relation): "Brain is like computer: both process information" (1 relation, vague)

Strong (systematic): "Brain is like computer:

• Input devices (sensors) → Processing (neurons/CPU) → Output (motor/display) [information flow]
• Storage (synaptic weights/RAM) ↔ Processing [interaction]
• Feedback loops (learning/software updates) [adaptation]
• Systematic: Maps entire input-processing-output-storage-feedback network"

Pragmatic Centrality

Principle: Focus mapping on aspects most relevant to current goal, even if other mappings exist.

Example: "Cell is like factory"

• Goal: Explain protein synthesis → Map nucleus (blueprint storage) to office, ribosomes (assembly) to assembly line, mitochondria (power) to power plant. Central: Manufacturing process.
• Goal: Explain energy management → Map mitochondria (ATP production) to power plant, glucose transport to fuel delivery. Central: Energy system.

Same analogy, different pragmatic focus based on goal.

---

3. Analogical Problem-Solving

Retrieve-Map-Transfer Process

Step 1: Retrieve - Find source domain with similar structural problem

• Spontaneous retrieval: Reminded of similar problem from past experience
• Deliberate search: Systematically search analogous domains
• Trick: Look for structural similarity (relationships), not surface (objects)

Step 2: Map - Align elements between source and target

• Map entities (what corresponds to what?)
• Map relations (A→B in source corresponds to X→Y in target?)
• Test mapping validity

Step 3: Transfer - Bring solution from source to target

• Identify solution element in source
• Map solution to target domain
• Adapt as needed (exact transfer rare)

Example:

Target problem: Users abandoning mid-signup (structural issue: high friction in multi-step process)

Step 1: Retrieve source domain: Restaurant ordering process (similar structure: multi-step, high abandonment at payment)

Step 2: Map:

• Source: Menu → Order → Customize → Pay (steps)
• Target: Email → Password → Profile → Verify (steps)
• Both: Multi-step funnel with dropoff at customization/profile (high cognitive load)

Step 3: Transfer solution:

• Source solution: Restaurants use "defaults" (chef's recommendation), allow skip customization, pay at end
• Target application: Use defaults (auto-generate username), allow skip profile (fill later), verify at end (not middle)
• Result: Reduced abandonment 40% by reordering steps and adding defaults

Forced Connections (Systematic Variation)

When to use: Creative ideation, generating novel solutions.

Technique: Systematically pair target problem with random domains, force analogy, see what emerges.

Process:

1. State problem: [What are you trying to solve?]
2. Random domain: Pick unrelated domain (use random generator: biology, music, sports, architecture, cooking, weather...)
3. Force mapping: "How is [problem] like [random domain]?" - Map structure even if feels forced
4. Extract: What insights emerge from forced mapping?
5. Repeat: Try 5-10 random domains

Example:

Problem: Improve code review process

Random domain 1: Restaurant food critique

• Map: Code → Dish, Reviewer → Food critic, Review → Critique
• Insight: Critics use multi-criteria rubrics (presentation, taste, technique). Transfer: Create multi-criteria code review rubric (readability, correctness, architecture).

Random domain 2: Airport security

• Map: Code submission → Passenger, Review → Security check, Approval → Boarding
• Insight: Tiered security (pre-check for trusted). Transfer: Trusted developers get lighter review, new contributors get thorough review.

Random domain 3: Jazz improvisation

• Map: Code changes → Musical variations, Codebase → Jazz standard, Review → Band synchronization
• Insight: Jazz uses "call and response", tight feedback loops. Transfer: Pair programming as real-time "call and response" review.

Result: 3 novel ideas from forced analogies (rubric-based review, tiered trust, pair programming emphasis).

---

4. Testing Analogy Validity

Systematicity Test

Question: Does analogy map interconnected system of relations or just one fact?

Strong analogy: Maps multiple interconnected relations Weak analogy: Maps single isolated fact

Test procedure:

1. List all mapped relations
2. Check if relations interconnect (does Relation A affect Relation B?)
3. Count: 3+ interconnected relations = systematic

Productivity Test

Question: Does analogy generate new predictions or insights about target domain?

Productive analogy: Leads to testable predictions in target Unproductive analogy: Just restates what we already know

Test procedure:

1. What does analogy predict about target that we didn't already know?
2. Is prediction testable?
3. If tested, does it hold?

Example:

• Analogy: "Immune system is like cybersecurity defense"
• Source domain fact: Immune system has "memory" (antibodies persist)
• Prediction for target: Cybersecurity should have "threat memory" (remember past attack signatures)
• Test: Implement threat signature database. Does it improve detection? (Yes → productive analogy)

Scope Limitation Test

Question: Where does analogy break down? What doesn't transfer?

Critical: Every analogy has limits. Acknowledging them prevents overgeneralization.

Test procedure:

1. List source domain facts
2. For each fact, does it map to target? (Yes/No)
3. Explicitly state "No" cases as limitations

Example:

• Analogy: "Startup growth is like rocket launch"
• Maps: Escape velocity (critical early momentum), fuel burn (runway), trajectory (growth curve)
• Breaks down: Rockets don't pivot mid-flight, can't refuel in space, one-shot (not iterative)
• Limitation: "Analogy useful for early momentum needs, but breaks down for pivot/iteration/long-term sustainability aspects. Startups ARE iterative and can refuel (fundraise), unlike rockets."

---

5. Common Pitfalls & How to Avoid

Pitfall	Description	How to Avoid
Surface Analogy	Maps superficial features, not structure. "Both are blue" doesn't help.	Explicitly map relations, not attributes. Test: Does analogy work if you change objects?
False Synthesis	Forcing agreement where genuine conflict exists.	When sources disagree, state it clearly. Don't paper over conflicts. Provide meta-framework or state uncertainty.
Cherry-Picking	Selecting only sources/data that support pre-existing view.	Systematic source inclusion. Explicitly address disconfirming evidence.
Analogy as Proof	Treating analogy as evidence rather than illustration.	State clearly: "Analogies illustrate, don't prove." Use analogy for explanation, not argumentation.
Overgeneralization	One data point or source → sweeping pattern claim.	Require 3+ sources for pattern claim. Acknowledge n=1 as anecdote, not trend.
Overextending Analogy	Pushing analogy beyond where it holds.	Explicitly test scope limits. State where analogy breaks down. Know when to stop.
Mixing Levels	Confusing data (facts) with interpretation (analysis).	Clearly label what's observation vs interpretation. "Data shows X (fact). This suggests Y (interpretation)."
Ignoring Disconfirming	Dismissing sources that contradict synthesis.	Actively seek disconfirming evidence. Explain why disconfirming source doesn't invalidate (scope, quality) OR revise synthesis.
Vague Mapping	Unclear what corresponds to what in analogy.	Create explicit mapping table. "In source, X. In target, Y. X↔Y because [relation]."
Single-Source Pattern	Claiming pattern from one source.	Patterns require repetition across sources. One source = theme, not pattern.

Synthesis Quality Checklist

• All relevant sources included and documented
• Source quality assessed (not all sources equal)
• Themes identified with frequency counts (X/Y sources)
• Agreements clearly stated
• Conflicts explicitly addressed and resolved (or uncertainty stated)
• Patterns identified with evidence from 3+ sources
• New insights generated (synthesis adds value beyond summarizing)
• Evidence cited for claims
• Gaps and uncertainties acknowledged
• Conclusions proportional to evidence (not over-claiming)

Analogy Quality Checklist

• Source domain familiar to audience
• Target domain clearly defined
• Structural mapping explicit (entities and relations mapped)
• Multiple relations map (systematicity)
• Generates new predictions (productivity)
• Limitations explicitly stated (where it breaks down)
• Not used as proof (illustrative only)
• Appropriate for goal (pragmatic centrality)
• Deep (relational) not surface (attribute-based)
• Doesn't overextend beyond valid scope

---

6. Advanced Integration Techniques

Synthesis → Analogy Pipeline

Use case: Synthesize complex findings, then use analogy to make accessible.

Process:

1. Synthesize first: Identify patterns across sources
2. Find familiar domain: What domain does audience know that has similar structure?
3. Map pattern: Transfer synthesized pattern to familiar domain
4. Explain: Use analogy to make synthesis accessible

Example:

• Synthesis: Analyzed 20 incident postmortems → Pattern: 80% involve config change + missing rollback.
• Familiar domain: Home renovation
• Map: Config change = renovation, Rollback = "undo" plan, Missing rollback = no plan B if renovation fails
• Analogy: "System configs are like home renovations. You can change layout (config), but without 'undo' plan (rollback), failed renovation leaves you homeless (broken system). Always have rollback plan."
• Result: Technical pattern accessible to non-technical stakeholders via home renovation analogy.

Multiple Analogies (Triangulation)

Use case: Complex target needs multiple analogies, each illuminating different aspect.

Technique: Use 2-3 analogies for same target, each mapping different facet.

Example - Explaining "Technical Debt":

Analogy 1: Financial debt

• Maps: Borrowing (quick hacks) → Interest (maintenance cost) → Compounding (code harder to change)
• Illuminates: Cost over time, eventually must pay back

Analogy 2: Tooth decay

• Maps: Skipping brushing (skipping quality) → Cavity (bug) → Root canal (rewrite)
• Illuminates: Gradual degradation, early prevention cheaper than late fix

Analogy 3: Garden weeds

• Maps: Weeds (bad code) → Spread (contagion) → Harder to remove when established
• Illuminates: Contagion aspect, importance of early removal

Together: Three analogies illuminate cost accumulation (financial), gradual degradation (tooth), and contagion (weeds). More complete understanding than any single analogy.

Key Takeaway: Advanced synthesis combines information rigorously (thematic analysis, conflict resolution, pattern identification). Advanced analogy maps relational structure systematically (structural mapping theory, systematicity, productivity testing). Together, they enable understanding complex systems, transferring knowledge across domains, and generating creative solutions.