Paraconsistent Logic

Classical logic — like the sentential and first-order systems developed here — is explosive: from a contradiction, everything follows. The principle

$φ, \neg φ ⊢ ψ (ex contradictione quodlibet, ECQ)$

means a single inconsistency trivializes a theory, making every statement provable. Paraconsistent logic is any logic in which this fails — where the consequence relation tolerates contradictions without exploding. Such logics let one reason non-trivially in the presence of inconsistent information, isolating a contradiction instead of letting it contaminate everything.

Paraconsistency is a property of a consequence relation, not a single system: there are many paraconsistent logics, built on different diagnoses of why classical logic explodes. This page assumes the syntax of sentential and first-order logic; what changes is which inferences are valid.

Why classical logic explodes

The standard derivation of ECQ uses disjunctive syllogism together with disjunction introduction:

$φ$ — assumption;
$\neg φ$ — assumption;
$φ \lor ψ$ — from (1) by disjunction introduction ( $\lor$ I);
$ψ$ — from (2) and (3) by disjunctive syllogism (DS).

A paraconsistent logic must reject at least one step. The two main families differ on which:

Relevance (relevant) logics reject disjunction introduction's use here, or more precisely the irrelevant implication that lets an arbitrary $ψ$ ride in. They demand that premises be genuinely relevant to the conclusion — that antecedent and consequent share propositional content.
Many-valued / "Logics of Formal Inconsistency" reject disjunctive syllogism: when $φ$ is both true and false, knowing $\neg φ$ does not license discharging the $φ$ disjunct.

Semantic strategies

Many-valued and gappy/glutty truth

Classical bivalence forces every sentence to be exactly true or exactly false. Paraconsistent semantics typically admits truth-value gluts — sentences that are both true and false:

LP (Priest's Logic of Paradox). Three values ${T, B, F}$ , where $B$ ("both") is a designated (truth-preserving) value alongside $T$ . The connectives use the strong Kleene tables, but because $B$ is designated, a contradiction $φ \land \neg φ$ can take a designated value without every formula doing so. ECQ fails; disjunctive syllogism fails. LP validates all classical theorems (it has the same tautologies) but a weaker consequence relation.
Belnap–Dunn four-valued logic (FDE, "First-Degree Entailment"). Four values — true, false, both, neither — designed for reasoning over possibly inconsistent and incomplete databases (the values are subsets of ${T, F}$ ). FDE is the common paraconsistent and paracomplete core: it drops both ECQ and excluded middle.

Contrast with intuitionistic logic, which is paracomplete (gives up excluded middle, allowing gaps) but remains explosive. Paraconsistency is in this sense the formal dual of intuitionistic paracompleteness: gluts versus gaps.

Relevant logics

Relevant logics (Anderson–Belnap R, E, the basic system B, …) re-engineer the conditional so that $φ \to ψ$ holds only when $φ$ is relevant to $ψ$ . The standard semantics is Routley–Meyer relational semantics, which uses a ternary accessibility relation $R (a, b, c)$ on worlds plus the Routley star $a^{*}$ (an involution on worlds interpreting negation):

$a ⊩ φ \to ψ ⟺ \forall b, c (R (a, b, c) \land b ⊩ φ \Rightarrow c ⊩ ψ), a ⊩ \neg φ ⟺ a^{*} \neq ⊩ φ .$

The star semantics permits worlds where $φ$ and $\neg φ$ both hold, blocking explosion. Relevant logics reject classical paradoxes of material implication such as $φ \to (ψ \to φ)$ and $⊥ \to φ$ .

Logics of Formal Inconsistency (LFIs)

The Brazilian school (da Costa's hierarchy $C_{n}$ , and the modern LFI framework) keeps explosion as a controlled principle. A connective $\circ φ$ (" $φ$ is consistent / well-behaved") is added, and a gentle explosion is retained:

$\circ φ, φ, \neg φ ⊢ ψ .$

Contradictions explode only for formulas explicitly marked consistent; unmarked formulas may be inconsistent harmlessly. This localizes classical reasoning to the "consistent fragment" while tolerating contradictions elsewhere.

Dialetheism — a philosophical aside

Paraconsistency is a thesis about logic (some contradictions don't entail everything); dialetheism is the stronger metaphysical thesis that some contradictions are actually true — there exist true contradictions (dialetheia). Every dialetheist needs a paraconsistent logic, but one can adopt paraconsistency without dialetheism, e.g. purely to reason about inconsistent databases or inconsistent scientific theories without committing to their truth. Priest defends dialetheism using the Liar ("this sentence is false") and the set-theoretic / semantic paradoxes (Russell, Curry), arguing that an LP-style logic lets one accept the paradoxical sentences as gluts without triviality. Curry's paradox is the sharpest pressure test: it derives triviality using only contraction and modus ponens, so paraconsistent logics must also restrict contraction ( $φ \to (φ \to ψ) ⊢ φ \to ψ$ ) to avoid collapse.

Paraconsistent set theory

The most striking payoff of paraconsistency is that it lets one revive naive set theory. Classical ZFC responds to the paradoxes by abandoning unrestricted comprehension: Separation only lets you carve subsets out of sets you already possess. Paraconsistent set theory makes the opposite trade — keep the full naive (unrestricted) comprehension schema

$\exists y \forall x (x \in y \leftrightarrow φ (x)) for every formula φ,$

and simply tolerate the contradictions it generates, because explosion no longer turns a local inconsistency into global triviality.

Russell's set as a glut

With full comprehension one may form $R = {x : x \in / x}$ . The usual reasoning still goes through:

$R \in R \leftrightarrow R \in / R, hence R \in R \land R \in / R .$

Classically this is catastrophic — it proves everything. Paraconsistently it is a quarantined glut: $R$ is a genuinely inconsistent object, both a member and a non-member of itself, but $0 = 1$ does not follow. Russell's set exists; it is simply contradictory.

The universal set and friends

Taking $φ (x) := (x = x)$ yields the universal set $V = {x : x = x}$ with $V \in V$ : naive paraconsistent set theory is a genuinely universal set theory, lifting the cumulative-hierarchy ban on a set of all sets. The set of all ordinals (Burali-Forti), Cantor's largest cardinal, and the Russell class of the universe all return as legitimate — if locally inconsistent — objects, rather than as proofs that something went wrong.

The real technical content

The interesting results are not that these sets exist (that is easy) but that one can have them without trivializing:

Non-triviality. The hard theorem is that naive comprehension over a suitable relevant logic does not prove every sentence. Ross Brady (1989, 2006) gave a model-theoretic non-triviality proof for naive set theory built on a depth-relevant logic (close to DJ/DK), using infinite-valued fixed-point models. This is the paraconsistent analogue of a consistency proof: it shows the gluts stay isolated.
Contraction, not negation, is the culprit. What actually forces collapse is structural contraction $φ \to (φ \to ψ) ⊢ φ \to ψ$ , weaponized by Curry's paradox (form $C = {x : x \in x \to ψ}$ for arbitrary $ψ$ ). Every viable naive set theory must therefore restrict contraction — the deepest lesson of the area is that the set-theoretic paradoxes are about contraction, not about $\neg$ .
How much mathematics survives? This is the active frontier. Zach Weber (2010, 2012) pushes hardest: he recovers a paraconsistent theory of ordinals and cardinals, a paraconsistent well-ordering theorem and a version of Cantor's theorem, with Burali-Forti turned from a paradox-to-avoid into a theorem-to-accept. The price is that inconsistent objects pervade the arithmetic and classical recapture — making consistent objects behave classically — is delicate and partly open.
Extensionality is the pressure point. Combining naive comprehension with full extensionality $(\forall z (z \in x \leftrightarrow z \in y) \to x = y)$ is where systems tend to break: extensionality interacts with contraction-like principles to re-import triviality (the same extensional mechanism that powers Diaconescu's theorem on the constructive side). Brady's and Weber's systems admit extensionality only with care.

The honest verdict

Paraconsistent set theory is mathematically real and non-trivial, but it is not a drop-in replacement for ZFC. Recapturing ordinary "consistent" mathematics is laborious, proofs are harder once contraction and disjunctive syllogism are gone, and most mathematicians regard it as a philosophically motivated research program (Priest, Weber, Brady) rather than a working foundation. Its principal pull is philosophical: for the dialetheist it vindicates the claim that the paradoxical sets genuinely exist and are genuinely contradictory, with the logic guaranteeing the contradiction cannot spread.

Applications

Paraconsistency is motivated by settings where inconsistency is unavoidable yet reasoning must continue:

Inconsistent databases and knowledge bases — querying data that contains conflicting records without returning every answer trivially (the original motivation for Belnap's four-valued logic).
Belief revision and merging — combining testimony or sources that disagree.
Inconsistent but useful scientific theories — e.g. the early Bohr model, or naive calculus with infinitesimals, which were productive despite formal contradictions.
Automated reasoning and AI — agents that must act on contradictory inputs without their entire reasoning collapsing.
Naive set theory and semantics — retaining unrestricted comprehension or a transparent truth predicate by tolerating the resulting paradoxes paraconsistently.

A worked example: infinitesimals

The canonical "inconsistent but useful" theory is the early calculus. Leibniz and the 18th-century practitioners treated an infinitesimal $d x$ inconsistently — in the same derivation it is both nonzero (so you may divide by it) and zero (so you may discard it):

$\frac{( x + d x ) ^{2} - x ^{2}}{d x} = \frac{2 x d x + d x ^{2}}{d x} = 2 x + d x = 2 x .$

The first equality requires $d x \neq = 0$ ; the final step requires $d x = 0$ . Berkeley's The Analyst (1734) mocked these "ghosts of departed quantities" for being $0$ and not- $0$ at once — yet the calculus never produced false theorems about real functions. It was a formally inconsistent theory that was nonetheless non-trivial and fruitful for roughly 150 years, exactly the phenomenon paraconsistency models: contradictions get quarantined rather than exploded.

The twist is how the contradiction was eventually tamed. The rigorous modern homes of infinitesimals are mostly not paraconsistent:

Framework	Status of $d x$	Logic
Historical (Leibniz/Newton)	$d x = 0$ and $d x \neq = 0$ (informal)	inconsistent, unformalized
Non-standard analysis (Robinson)	nonzero hyperreal	classical
Smooth infinitesimal analysis (Lawvere–Kock)	nilpotent, $d x = 0$ undecided	intuitionistic (gap)
Mortensen's inconsistent calculus	$d x = 0$ and $d x \neq = 0$ (true glut)	paraconsistent

Non-standard analysis dissolves the contradiction classically (a genuine nonzero infinitesimal); smooth infinitesimal analysis dissolves it intuitionistically, refusing to decide $d x = 0 \lor d x \neq = 0$ so the infinitesimal sits in a gap rather than a glut. The genuinely paraconsistent reconstruction is Chris Mortensen's inconsistent mathematics (1995), which takes $d x = 0 \land d x \neq = 0$ as a true glut and runs differentiation over a paraconsistent logic so the contradiction stays local. It is a legitimate rational reconstruction of Leibniz's reasoning — though a niche program, not how working analysts compute.

More inconsistent-but-useful examples

The infinitesimal is one of a family. They split into two kinds.

Inconsistent, then rigorized. Like $d x$ , these were formally contradictory as first used, worked anyway, and were later given a consistent foundation:

The Dirac delta "function." The closest twin to the infinitesimal: $δ (x) = 0$ for all $x \neq = 0$ yet $\int_{- \infty}^{\infty} δ (x) d x = 1$ — impossible for an ordinary function, which if zero almost everywhere integrates to $0$ . Physicists computed with it for two decades before Schwartz's theory of distributions (1945) made it rigorous, exactly as non-standard analysis later rationalized $d x$ .
Heaviside's operational calculus. Heaviside treated the differentiation operator $p = d / d t$ as an algebraic quantity — dividing by it, taking $p$ , expanding $1/ (p - a)$ as a series — with no justification, yet derived correct circuit solutions. The Laplace transform and operator theory later vindicated it.
Divergent series (Euler). Euler manipulated divergent series as if convergent, e.g. $1 - 1 + 1 - 1 + \dots = \frac{1}{2}$ and $1 + 2 + 4 + \dots = - 1$ . Formally inconsistent, but the values were meaningful; summation methods (Abel, Cesàro, Borel) later legitimized them.
The Bohr model. An electron in a stationary orbit is an accelerating charge, so Maxwell's electrodynamics demands it radiate and spiral inward — yet Bohr stipulated it does not. Logically inconsistent with the theory it borrowed, but it predicted the hydrogen spectrum; quantum mechanics superseded it.

Genuine gluts. Here there is no clean consistent fix; the contradiction is faced head-on, which is paraconsistency's real home turf:

The Liar and semantic paradoxes. "This sentence is false" is true iff false. The dialetheist accepts it as a true contradiction and uses LP so it does not explode — the motivating case for the whole program.
Inconsistent databases. One record says Smith earns 50k, another 60k. You still want "Is Smith's salary above 40k?" answered yes, not every query returning everything — the original motivation for Belnap's FDE.
Conflicting legal and normative codes. A legal system can contain statutes that jointly permit and forbid the same act, yet courts keep functioning rather than concluding "everything is permitted."
The preface paradox. An author rationally believes each claim in her book and believes the preface's admission that it contains some error — a jointly inconsistent yet rational belief set.
Motion and the instant of change. Priest argues that at the instant a body begins to move it is momentarily both at rest and in motion — a glut about the continuum, defended in In Contradiction.

The pattern: the first group shows reasoners quarantining contradictions temporarily until rigor arrives (a descriptive motivation for paraconsistency); the second shows contradictions that arguably cannot be eliminated, where a paraconsistent logic is the standing tool.

Summary

	Classical / intuitionistic	Paraconsistent
Explosion ( $φ, \neg φ ⊢ ψ$ )	valid	rejected (non-trivial under contradiction)
Truth values	bivalent (classical); gaps (intuitionistic)	gluts (both), often gaps too
Disjunctive syllogism	valid	fails (LP, FDE) or restricted
Conditional	material / intuitionistic	relevant (R, E) in relevance logics
Excluded middle	classical: yes; intuitionistic: no	varies (FDE: no; LP: yes)
Representative systems	classical, intuitionistic	LP, FDE, R/E, da Costa $C_{n}$ , LFIs

Paraconsistent logic is what remains when one refuses to let a single contradiction prove everything. By admitting truth-value gluts (LP, FDE), demanding relevance (R, E), or controlling explosion with a consistency operator (LFIs), these systems support non-trivial reasoning under inconsistency — whether the inconsistency is treated as a temporary artifact of imperfect information or, for the dialetheist, as a genuine feature of the world.

Keyboard shortcuts

youyuanwu