Analogy Search (Cross-Domain Transfer)

Overview

Many problems have been solved before - just in different domains. Velcro from burrs, sonar from bats, assembly lines from meatpacking. This procedure systematizes the search for useful analogies.

Goal

Find solutions by searching for analogous problems in other domains and transferring their solutions. Systematic domain list reduces the intelligence needed.

Steps

Step 1: Describe the Problem

State the problem concretely. Include: What’s the situation? What’s wrong? What do you need?

Output: Concrete problem description

Step 2: Extract Problem Structure

Abstract the problem to its structural elements. Remove domain-specific details, keep the pattern.

Ask:

• What is the core challenge? (optimization, matching, search, etc.)
• What are the inputs and outputs?
• What are the constraints?
• What makes it hard?

Output: Abstract problem structure

Step 3: Search Domains Systematically

For each domain in the list, ask: “What problem in [domain] has similar structure?” “How do they solve it?”

Don’t skip domains - even unlikely ones may have insights.

Output: Analogies from each domain

Step 4: Evaluate Analogies

For each analogy found, assess:

• Structural similarity (how well does it map?)
• Solution quality (does it work well in source domain?)
• Transferability (can it be adapted to our domain?)

Output: Evaluated analogies

Step 5: Transfer Solutions

For promising analogies, work out how to transfer:

• What maps directly?
• What needs adaptation?
• What doesn’t transfer?

Output: Transferred solutions

Step 6: Evaluate Transferred Solutions

Assess each transferred solution:

• Does it actually solve the problem?
• What are the limitations?
• What would need to change?

Output: Solution evaluation

When to Use

• Stuck on a problem
• Looking for innovative solutions
• Want to avoid reinventing the wheel
• Need fresh perspective
• Problem seems unique (probably isn’t)

Verification

• Problem structure was abstracted (not just surface features)
• Multiple domains were searched systematically
• Structural similarity was assessed, not just surface similarity
• Transfer was explicit about what maps and what doesn’t
• Transferred solution addresses the actual problem