Propositional Logic¶
Propositional logic is the simplest formal logical system, consisting of propositions (true/false statements) and logical connectives. While less expressive than first-order-logic, it illustrates all fundamental concepts and provides a complete inference system.
Syntax¶
Propositions¶
- Atomic: Single propositions (e.g., P, Q, Raining)
- Complex: Built from atomic propositions using connectives
Logical Connectives¶
| Connective | Symbol | Meaning |
|---|---|---|
| Negation | ¬ | NOT |
| Conjunction | ∧ | AND |
| Disjunction | ∨ | OR |
| Implication | → | IF...THEN |
| Biconditional | ↔ | IF AND ONLY IF |
Well-formed formulas (WFFs): - Atomic propositions are WFFs - If α and β are WFFs, so are ¬α, (α ∧ β), (α ∨ β), (α → β), (α ↔ β)
Semantics¶
Truth Tables¶
- Determine truth value of complex formulas given truth assignments
- Complete enumeration of all 2^n possible assignments for n propositions
Key Concepts¶
Model: Assignment of true/false to all propositions Satisfiable: Formula is true under at least one model Valid/Tautology: Formula is true under all models Entailment (⊨): α ⊨ β iff whenever α is true, β is true - β is a logical consequence of α
Inference Problem¶
Given KB and query q: - Does KB ⊨ q? (Does the KB entail the query?) - In other words: Is q true in all models where KB is true?
Inference Methods¶
Model Checking (TT-ENTAILS)¶
- Enumerate all possible models
- Check KB true in each model
- Check if q is true in all those models
- Complete but exponential: O(2^n)
Theorem Proving¶
Derive new formulas using inference rules:
Modus Ponens: (α → β), α ⊨ β And-Elimination: (α ∧ β) ⊨ α Unit Resolution: (α ∨ β), ¬β ⊨ α
Resolution¶
Most important inference rule: - (α ∨ β), (¬β ∨ γ) ⊨ (α ∨ γ) (Resolution rule) - Convert KB to Conjunctive Normal Form (CNF) - Repeated application of resolution generates all entailed clauses - Refutation: Prove KB ∧ ¬q is unsatisfiable (contains empty clause)
CNF: Conjunction of disjunctions of literals - (A ∨ B) ∧ (¬A ∨ C) ∧ (B ∨ ¬C)
Horn Clauses: Disjunction with at most one positive literal - Enables efficient forward/backward chaining - P ∨ ¬Q ∨ ¬R (same as: Q ∧ R → P)
Forward & Backward Chaining¶
Efficient for definite clauses (Horn clauses with exactly one positive literal):
Forward Chaining (PL-FC-ENTAILS): - Start from known facts (premises) - Apply Modus Ponens forward - Derive new facts until query found or no progress - Linear time in KB size
Backward Chaining: - Start from query - Work backward to premises - Recursively prove subgoals - Depth-first search approach
SAT Solving¶
Boolean Satisfiability Problem: Is formula satisfiable?
DPLL Algorithm: - Backtracking search with pruning - Early termination, unit propagation, pure literal elimination - Exponential worst-case but practical for many problems
WALKSAT: - Local search approach - Random walk to escape local maxima - Effective on large random SAT instances
Important Properties¶
- Soundness: Only derive entailed formulas
- Completeness: Can derive all entailed formulas (given enough time)
- Decidability: Problem is decidable (solution always found eventually)
Limitations¶
- No explicit representation of objects or relations
- Cannot express "for all X" or "there exists X"
- Cumbersome for domains with many propositions
- Leads to combinatorial explosion
Solution: First-Order Logic¶
See first-order-logic for more expressive power.
Related Concepts¶
- First-Order-Logic — More expressive extension
- Inference — General inference mechanisms
- Logic-Programming — Programming paradigm based on logic
- Knowledge-Representation — Systematic knowledge encoding
References¶
Russell & Norvig (2010): Chapter 7 - Propositional Logic