Analogical & Causal Reasoning: Models, Inference & Fallacies

Core Concepts

1. Analogical Reasoning: Drawing Logical Bridges

Analogical reasoning compares two domains to infer properties of one from the other. It follows this structure:

If A has properties P₁, P₂, ..., Pₙ and B has properties P₁, P₂, ..., Pₙ then B likely has property Q (if A has Q).

How It Works

• Source Domain (A): The known system (e.g., "a computer’s CPU").
• Target Domain (B): The unknown system (e.g., "the human brain").
• Mapping: Identify shared features (e.g., both process information via parallel pathways).
• Inference: Extend conclusions from A to B (e.g., "If CPUs use caching, maybe the brain does too").

Strengths & Weaknesses

Real-World Example: eSewa’s Fraud Detection

Analogy Used: eSewa’s fraud team models transaction patterns like bank fraud detection systems (source domain). Shared features:

• Both use anomaly detection (e.g., sudden large transfers).
• Both rely on user behavior baselines (e.g., typical spending limits). Inference: If banks flag unusual ATM withdrawals, eSewa flags unusual e-wallet top-ups. Result: Reduced fraud by 30% in 2023 (Nepal Rastra Bank report).

2. Causal Reasoning: Unraveling Cause and Effect

Causal reasoning identifies why an event occurs by tracing cause-effect chains. Key terms:

• Cause (C): The event that triggers another (e.g., "heavy rain").
• Effect (E): The resulting event (e.g., "Kathmandu traffic jams").
• Spurious Correlation: Two events linked by coincidence (e.g., "Ice cream sales ↑ → Drownings ↑" because both rise in summer).

Types of Causal Relationships

graph LR
    A["Direct Cause"] --> B["Effect"]
    C["Indirect Cause"] --> D["Intermediate Event"] --> B
    E["Common Cause"] --> F["Two Effects"]
    G["Spurious"] -->|"No real link"| H["Coincidence"]

Example: Why do Daraz orders spike on Sundays?

• Direct Cause: Nepalis shop more on weekends (behavior).
• Indirect Cause: Salary days → more disposable income.
• Spurious: "Daraz orders ↑ because the moon is full" (no evidence).

Causal Fallacies to Avoid

3. Analogical vs. Causal Reasoning: Key Differences

4. Worked Example: NEPSE Stock Prices

Scenario: "NEPSE’s stock prices drop when global oil prices rise. Does oil cause NEPSE’s decline?" Step-by-Step Analysis:

1. Identify the Link:
   • Oil price ↑ → Fuel costs ↑ → Transportation/logistics expenses ↑ for Nepali companies.
2. Check for Confounding Variables:
   • Global recession (2020) also hurt NEPSE. Is oil the only cause? No.
3. Test the Analogy:
   • Compare to other markets: "When Brent crude rises, Asian stock indices (e.g., Nifty 50) also fall."
   • Inference: Oil prices are a partial cause, but not the sole factor.
4. Causal Model:

   flowchart TD
     A["Global Oil Price ↑"] --> B["Nepali Fuel Costs ↑"]
     B --> C["Company Profits ↓"]
     C --> D["Investor Confidence ↓"]
     D --> E["NEPSE Index ↓"]

Real-World Tie-In: In 2022, when Brent crude hit $120/barrel, NEPSE’s Micro Cap Index fell 18% (NEPSE data). The analogy: "NEPSE reacts to oil shocks like the S&P 500 to Fed rate hikes."

5. Applications in Nepal’s Tech Industry

In the Real World

1. Khalti’s Loan Approval System

   • Analogy: Khalti’s "instant loan" feature uses credit scoring models (source domain: traditional banks).
   • How: Maps user data (income, repayment history) to bank loan approval rules.
   • Result: 70% of applicants get decisions in <2 minutes (Khalti 2023 report).

2. Ncell’s Network Outage Predictions

   • Causal Reasoning: Ncell’s AI predicts outages by analyzing:
     • Cause: Heavy rain → tower signal loss.
     • Effect: Call drops in Chitwan.
   • Analogy: "Our system works like weather forecasts predicting storms."

3. NEPSE’s ETF Investments

   • Analogy: NEPSE’s ETF (Exchange-Traded Fund) mimics global indices (e.g., "Nepal’s S&P-like fund").
   • Causal Link: "When global markets rise, Nepali ETFs follow (but with a 3-month lag)."

Exam Tip

1. For Analogies:

   • Always state the source and target domains explicitly.
   • Example Answer:

     "The analogy between Pathao’s surge pricing and Uber’s algorithm is valid because both (1) adjust fares based on real-time demand data, (2) use dynamic pricing models, and (3) aim to balance driver incentives with user costs. However, the analogy fails if we assume both systems handle peak-hour surges identically—Pathao’s smaller fleet size in Nepal may lead to faster price spikes."

2. For Causal Reasoning:

   • Draw a causal diagram (even in text) to show chains.
   • Watch for fallacies: If the question asks "Why does X happen?", reject answers like "Because Y happened earlier" without testing for confounding variables.

3. Common Pitfalls:

   • False Analogies: Never assume two systems are identical. Highlight differences too.
   • Overgeneralizing: "All Nepali apps use cloud storage" → Counter: eSewa uses on-premise servers for security.