Advanced Study Notes – Logical Reasoning
Analogy & Classification
Target audience: MIT‑level undergraduates / graduate students in mathematics, philosophy, or computer science.
Goal:
- Ground analogue & classification reasoning in first‑principle logic.
- Develop a rigorous framework that accommodates edge cases and paradoxes.
- Provide a challenging problem with a detailed solution.
1. Introduction
Logical reasoning in the Science Olympiad (NSO) demands precise use of analogy and classification.
- Analogy: a mapping between two structures that preserves relations.
- Classification: partitioning a domain into mutually exclusive (or overlapping) subsets.
Both involve structures (sets, relations, functions) and can be analyzed using model theory, category theory, and probability when dealing with uncertainty.
2. Formal Foundations
| Symbol | Meaning | Example |
|---|---|---|
| (\mathcal{U}) | Universal set (domain of discourse) | All animals |
| (\subseteq) | Subset relation | ({\text{cats}}\subseteq{\text{mammals}}) |
| (R\subseteq A\times B) | Binary relation | “is a parent of” |
| (f:A\to B) | Function | “age in years” |
| (\forall,\exists) | Quantifiers | ∀x (x∈Animals → …) |
| (\models) | Semantic entailment | Γ ⊨ φ |
2.1. Relations and Functions
- Equivalence relation: reflexive, symmetric, transitive → partitions the domain into equivalence classes.
- Partial order: reflexive, antisymmetric, transitive → induces a Hasse diagram (classification hierarchy).
- Functor (categorical): maps objects and morphisms preserving composition/identity → formalizes analogical mapping.
2.2. First‑Principles Definitions
- Analogy:
A relation-preserving map ( \alpha: A \to B ) such that for every relation (R\subseteq A^n) there exists a corresponding relation (R'\subseteq B^n) with
[ \forall (x_1,\dots,x_n)\in A^n,; (x_1,\dots,x_n)\in R \iff (\alpha(x_1),\dots,\alpha(x_n))\in R'. ] - Classification:
A partition ( {C_i}{i\in I}) of (\mathcal{U}) where
[ C_i \cap C_j =\varnothing \text{ for } i\neq j,\qquad \bigcup{i\in I}C_i = \mathcal{U}. ] Relaxed classification allows overlaps (see §5.3).
3. Analogy
3.1. Types of Analogical Reasoning
| Type | Formalization | Example |
|---|---|---|
| Structural | Isomorphism of graphs | “The solar system ≅ a simple pendulum” |
| Similarity | Approximate homomorphism | “Human brain ≅ neural network” |
| Compositional | Mapping of components | “Engine ≅ heart” |
| Functional | Preservation of function outputs | “Phone ≅ translator” |
3.2. Validity Criteria
| Criterion | Formal Statement | Why it matters |
|---|---|---|
| Contextual relevance | If (P) is true in domain (A), then (P) should be plausible in (B). | Avoids false analogies. |
| Structural alignment | Existence of a bijection preserving relations. | Ensures logical coherence. |
| Non‑triviality | The mapping must be non‑empty and not vacuously true. | Prevents “everybody is a pet” analogies. |
Theorem (Soundness of Structural Analogy)
If (\alpha: A \to B) is an isomorphism of relational structures (R) and (R'), then for any first‑order sentence (\varphi) over (R),
[
A \models \varphi \iff B \models \varphi^\alpha,
]
where (\varphi^\alpha) is (\varphi) with predicates translated via (\alpha).
Proof Sketch: Use induction on the structure of (\varphi), preserving logical connectives and quantifiers under (\alpha).
3.3. Strengthening & Weakening
- Strong analogy: all relations preserved.
- Weak analogy: only a subset of crucial relations preserved.
- Quantitative measure:
[ \text{Strength}(\alpha)= \frac{#\text{preserved relations}}{#\text{total relations}}. ]
3.4. Edge Cases
| Case | Problem | Remedy |
|---|---|---|
| Multiple inheritance | A member belongs to several categories → mapping may be ambiguous. | Use partial isomorphisms |