Product Logic

Origin. Hájek, Godo, Esteva developed Product Logic (1990s). Many-valued logic with multiplication as conjunction. One of three basic fuzzy logics (with Łukasiewicz, Gödel). Algebraically: product algebras. Part of monoidal t-norm logic family.

Models. Conjunction as multiplication. Łukasiewicz: bounded addition. Gödel: minimum. Product: multiplication on [0,1]. Models probabilistic combination: P(A∧B) = P(A)·P(B) when independent. Different properties from other fuzzy logics.

Formalism.

Truth values: [0,1] real interval.

Connectives:

• A ⊗ B = A · B (product conjunction)
• A → B = 1 if A ≤ B, else B/A (residuum)
• A ∧ B = min(A, B) (lattice meet)
• A ∨ B = max(A, B) (lattice join)
• ¬A = A → 0 = 0 if A > 0, else 1 (Gödel negation)

Key properties:

• A ⊗ A ≠ A (not idempotent unless A ∈ {0,1})
• Continuity: multiplication is continuous
• 0 is annihilator: 0 ⊗ A = 0
• No proper negation: ¬A = 0 for all A > 0

Product vs Łukasiewicz:

• Łuk: A ⊕ B = min(1, A+B), cancellation law fails
• Product: A ⊗ B = A·B, cancellation law holds (if A>0)

Axioms: BL (basic fuzzy logic) + Π1: ¬¬C → ((A⊗C → B⊗C) → (A → B)) + Π2: A ∧ ¬A → 0

Symbols.

Metatheory. Hájek, Godo, and Esteva (1996) introduced the system and proved it sound and complete both for the variety of product algebras — BL-algebras satisfying cancellativity — and, in the sharper standard form, for the single algebra [0,1] under ordinary multiplication. The structure behind that result is elementary: on (0,1] the map x ↦ −log x carries multiplication to addition and reverses the order, so the non-zero part of the standard product chain is the non-negative cone of the additive reals, and the residuum B/A becomes truncated subtraction. Product logic is one of the three principal schematic extensions of BL that Mostert and Shields's ordinal-sum decomposition of the continuous t-norms singles out, alongside Łukasiewicz and Gödel. The propositional logic is decidable and its tautology problem is coNP-complete. Its negation is the pseudocomplement, so Π proves the weak excluded middle ¬φ ∨ ¬¬φ while refuting φ ∨ ¬φ. The first-order extension is undecidable, and the set of formulas valid in the standard product algebra is not arithmetical.

Applies to. Probabilistic reasoning (independent events). Fuzzy control (product inference). Uncertainty combination. Decision-making under vagueness. Economic modeling. Signal processing.

Limitations. Negation retains only whether a degree is zero: ¬A = 0 for every A of positive degree and 1 otherwise, so all gradation is destroyed by a single negation, and ¬¬A → A fails. Adding involution does not extend the logic but collapses it — in the standard chain ¬¬A → A takes the value of A, so imposing it as an axiom forces the two-element algebra and yields classical logic. Zero annihilates, and iterated strong conjunction decays: Aⁿ tends to 0 for any A < 1, so a long chain of nearly-true premises conjoins to something arbitrarily close to false. There are no truth constants for intermediate degrees and no crispness operator, so neither "A holds to degree ½" nor "A is fully true" is expressible in the language. The probabilistic reading is formal only: product conjunction matches P(A ∧ B) = P(A)·P(B) for independent events, but the logic is truth-functional and probability is not, so no dependent case is modelled. First-order product validity over [0,1] is not arithmetical, hence not axiomatizable by any means.

© 2026 Lingenic LLC