Analogical Reasoning

Overview

Analogical Reasoning is a cognitive problem-solving method formalized by psychologists Mary Gick and Keith Holyoak (1980, 1983) through landmark research on how people transfer solutions between structurally similar problems from different domains. The method involves three critical steps: noticing an analogical connection exists, mapping corresponding elements between source and target problems, and applying the mapped solution to generate a parallel resolution.

Their classic study used Duncker's radiation problem (destroy tumor without harming healthy tissue) and the fortress story (capture fortress without triggering mines on roads). Only 10% solved it spontaneously; 30% solved it after reading the fortress story; 75% solved it when explicitly told to use the story as a hint. This reveals the core challenge: noticing the analogy is harder than applying it once recognized.

The breakthrough insight: develop problem schemas (abstract structural patterns) that transcend surface features. When you encounter multiple analogous problems, your brain induces a reusable schema—a convergence pattern—that accelerates future problem-solving across domains.

When to Use

• Facing a novel problem with no obvious solution in your domain
• Stuck on a problem that feels unsolvable with conventional approaches
• Need creative solutions by importing ideas from other fields
• Designing new systems based on proven patterns elsewhere (biomimicry, business models)
• Teaching or explaining complex concepts by finding familiar parallels
• Multiple similar problems suggest an underlying pattern worth abstracting
• Innovation through cross-pollination of ideas between industries

The Process

Step 1: Define the Target Problem Structure

Clearly articulate the target problem's abstract structure, not just surface features. Focus on relationships, constraints, and goals—not nouns.

Example (Radiation Problem): Need to apply force to a central point without causing damage along approach paths.

Step 2: Search for Analogous Source Problems

Actively seek solved problems from other domains that share structural similarity. Look beyond surface features (tumor vs. fortress) to functional parallels (converging forces, avoiding path damage).

Example: Military strategy (multi-directional attack), water erosion (multiple small streams), laser cutting (low power beams).

Step 3: Map Corresponding Elements Between Problems

Create explicit mappings: What in the source corresponds to what in the target? This is structural alignment, not literal translation.

Gick & Holyoak Mapping (Fortress → Radiation):

• Fortress → Tumor
• Army → Radiation
• Roads → Healthy tissue
• Mines on roads → Tissue damage from high-intensity ray
• Multiple small groups from different directions → Multiple weak rays from different angles
• Simultaneous arrival → Convergence at tumor site

Step 4: Generate a Parallel Solution

Apply the mapped structure to generate a solution for your target problem. Adapt, don't copy literally—the pattern transfers, not the specifics.

Example (Radiation Problem Solution): Send multiple low-intensity rays from different angles that converge at the tumor. Individually harmless to tissue, collectively destructive to tumor.

Step 5: Develop Problem Schemas from Multiple Analogies

When you encounter 2+ structurally similar problems, induce an abstract schema—the convergence pattern that links them. High-quality schemas dramatically improve transfer performance.

Convergence Schema (Gick & Holyoak):

• Pattern: When direct application of force causes collateral damage, divide and redirect force from multiple sources to converge at the target.
• Instances: Radiation therapy, military siege, distributed computing, load balancing.

Step 6: Test Schema Quality and Refine

Evaluate schema quality: Does it capture the core structural pattern? Can you apply it to new problems? Good schemas recognize the same concept across instances. Gick & Holyoak found 21% of participants created good schemas, and 91% of those solved the radiation problem.

Schema Quality Levels:

• Good (21%): Abstract structural pattern recognized → 91% transfer success
• Medium: Partial pattern → moderate transfer
• Poor: Surface features only → minimal transfer

Step 7: Build a Schema Library Over Time

Catalog effective schemas for future use. As your schema library grows, you recognize analogies faster and solve problems more efficiently—you've trained pattern recognition.

Example Application

Situation (Dropbox Growth 2009): Need rapid user growth with limited marketing budget.

Application:

1. Target Problem Structure: Acquire users at near-zero cost by incentivizing existing users to recruit new users.
2. Source Analogy: PayPal's early growth (1999)—gave $10 to both referrer and referee for signups. Viral coefficient > 1 enabled exponential growth.
3. Mapping: PayPal $10 cash → Dropbox 500MB free storage. Both use two-sided incentives (giver and receiver benefit).
4. Parallel Solution: Offer 500MB bonus storage to both referrer and referee (later optimized to 16GB max).
5. Schema Induced: "Two-sided incentive referral" pattern—reward both parties to maximize viral coefficient. Converts users into distribution channel.
6. Outcome: Dropbox grew from 100K to 4M users in 15 months (3900% growth). Referrals drove 35% of daily signups. The analogical reasoning from PayPal's proven pattern provided the strategic breakthrough.

Schema Application: This same "two-sided incentive" schema has been applied by Airbnb (travel credits), Uber (ride credits), and hundreds of SaaS companies—demonstrating the power of abstracted patterns.

Anti-Patterns

• ❌ Superficial analogies based on surface similarity (both involve computers ≠ structural parallel)
• ❌ Failing to explicitly map elements (vague "this is like that" without correspondence table)
• ❌ Literal copying instead of structural adaptation (context matters; adapt the pattern)
• ❌ Ignoring disanalogies: where the analogy breaks down (know the limits)
• ❌ Not actively searching other domains (analogies require deliberate cross-domain exploration)
• ❌ Single-instance schemas (need 2+ examples to induce robust patterns)
• ❌ Skipping schema induction phase (leaves learning implicit instead of explicit)

Related

• first-principles-thinking (decompose to fundamentals, rebuild—complements analogical reasoning)
• lateral-thinking (provocation and random entry to trigger unexpected analogies)
• inversion (reverse the analogy: what's the opposite pattern?)
• triz-contradiction-matrix (systematized cross-industry analogies for technical problems)
• biomimicry (nature as source domain for engineering analogies)
• mental-models (collection of schemas for fast pattern matching)