Paraconsistent Belief Revision: A Replacement-Enriched LFI for Epistemic Entrenchment

We further develop the formal foundations of Paraconsistent Belief Revision (PBR) by introducing Logics of Formal Inconsistency (LFIs) specifically designed to support the development of epistemic entrenchment-based models for belief change. The interpretation of formal consistency – and, more broadly, of paraconsistency – in terms of the epistemic attitudes adopted by rational agents and of these agents reasoning with potentially contradictory yet non-trivial epistemic states, respectively, is already well-established within the literature on PBR based on LFIs. However, previous approaches faced a key limitation: the absence of replacement in most LFIs prevented the construction of entrenchment-based operations. We address this gap by first revisiting and systematizing core properties essential for such modeling, formalizing them within Cbr, a previously introduced logic whose foundational properties we now examine and develop in depth. Building on this, we introduce RCbr, a replacement-enriched, self-extensional extension of Cbr, which makes it possible – within an LFI-based framework – to formally define epistemic entrenchment and to construct entrenchment-based belief revision mechanisms. This development enables a fully constructive approach to Belief Revision in paraconsistent settings, further advancing the theoretical treatment of LFIs and paraconsistency within the broader landscape of epistemic states and belief dynamics.

💡 Research Summary

The paper addresses a fundamental limitation in existing paraconsistent belief revision (PBR) frameworks that are based on Logics of Formal Inconsistency (LFIs). While LFIs introduce a consistency operator ◦ that allows contradictions to coexist without trivializing the system, most LFIs lack the replacement property—i.e., the ability to freely substitute logically equivalent formulas in any context without altering inferential outcomes. This property is essential for defining epistemic entrenchment, a ranking of beliefs that underlies the classic AGM belief revision theory.

To overcome this gap, the authors first revisit the logic Cbr, previously introduced but not systematically analyzed. Cbr extends the minimal LFI mbC with additional axioms, notably (Ax 12) ◦α ∨ (α ∧ ¬α) and (Ax 11) ◦α → (α → (¬α → β)). These axioms guarantee two crucial symmetries: (i) ◦α is equivalent to ◦¬α, and (ii) if α≡β and ¬α≡¬β then ◦α≡◦β. The authors provide a three‑valued non‑deterministic matrix (N‑matrix) semantics for Cbr, with designated values {1, ½}, and prove soundness and completeness. The semantics also yields the important result that Cbr validates strong acceptance and strong rejection notions, which are defined via the joint presence of ◦α and α (or ¬α) in a belief set.

However, Cbr still does not satisfy replacement. The authors therefore construct RCbr, a self‑extensional extension of Cbr obtained by adding a global inference rule (E ¬): from α≡β infer ¬α≡¬β. Because of Proposition 3.5(ii), the corresponding rule for the consistency operator (E ◦) becomes derivable, so only (E ¬) needs to be added. RCbr thus fulfills the replacement condition (R): whenever a tuple of formulas is pairwise equivalent, any connective applied to them yields equivalent results.

The algebraic counterpart of RCbr is given by Boolean Algebras with LFI operators (BALFIs). A BALFI expands a Boolean algebra with two unary operators ˜¬ and ˜◦ satisfying x ⊔ ˜¬x = 1 and x ⊓ ˜¬x ⊓ ˜◦x = 0, together with the additional identities required for RCbr (e.g., ˜¬˜¬x = x). The authors prove that RCbr is sound and complete with respect to the class of BALFIs, establishing a robust semantic foundation.

With RCbr’s replacement property in place, the paper defines epistemic entrenchment ≤ as a pre‑order on formulas that respects logical equivalence and the behavior of the consistency operator. Using this entrenchment, the authors reconstruct the four AGM revision operators—expansion, contraction, revision, and recovery—within a paraconsistent setting. The key innovation is that “strongly accepted” formulas (those for which both α and ◦α belong to the belief set) become the most entrenched, and thus are protected during revision unless their consistency is explicitly removed. This yields a fully constructive, AGM‑style belief revision mechanism that works even when the belief base contains contradictions.

The paper’s contributions can be summarized as follows:

1. A systematic axiomatization and N‑matrix semantics for Cbr, including proofs of soundness, completeness, and the two symmetry properties of the consistency operator.
2. The introduction of RCbr, a self‑extensional LFI that satisfies the replacement property, together with its BALFI algebraic semantics.
3. A formal definition of epistemic entrenchment grounded in RCbr, enabling the construction of entrenchment‑based belief revision operators in a paraconsistent environment.
4. Demonstration that the resulting revision framework preserves the essential AGM postulates (including success, inclusion, and recovery) while allowing contradictions to persist without trivialization.

By providing a logic that simultaneously supports paraconsistency and the structural features required for AGM‑style revision, the authors bridge a long‑standing gap between non‑explosive reasoning and dynamic belief change. This work opens the door to applications in multi‑agent systems, legal and ethical reasoning, and knowledge‑graph maintenance where inconsistent information is unavoidable but controlled belief update is essential. Future research directions suggested include algorithmic implementations of RCbr‑based revision, extensions to modal or temporal dimensions, and empirical evaluation on real‑world inconsistent datasets.