🤖 AI

An Abstract Worlds Semantic Framework for Belief Change Operators

This paper introduces "Abstract Worlds Semantics," a syntax-free, set-theoretic framework that treats worlds as primitive elements to unify and generalize classical and non-prioritized belief change models, including AGM, KM, and multiple change operators.

Original authors: Daniel Grimaldi, M. Vanina Martinez, Ricardo O. Rodriguez

Published 2026-06-02

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Daniel Grimaldi, M. Vanina Martinez, Ricardo O. Rodriguez

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine your mind as a library filled with books representing everything you believe to be true. Sometimes, you get new information (a new book) that changes how you organize your library. You might need to throw out old books, add new ones, or rearrange the shelves. In the world of Artificial Intelligence, this process is called Belief Change.

For decades, scientists have built complex mathematical rules to describe how this library should be updated. However, most of these rules were built specifically for "Classical Logic"—a very rigid, black-and-white way of thinking. If you tried to use these rules for more flexible or fuzzy types of logic, they often broke down or didn't fit.

This paper introduces a new, simpler way to look at belief change called Abstract Worlds Semantics (AWS). Here is how it works, using simple analogies:

1. The "World" as a Lego Brick

Instead of starting with complex sentences or logic formulas, the authors start with the most basic building blocks: "Worlds."

Think of a "World" as a single, complete Lego brick. A "World" represents one specific way reality could be.

Your Knowledge is a pile of Lego bricks you have selected.
New Information is a new pile of bricks someone hands you.

In this framework, you don't need to know the "language" of the bricks (the logic). You just look at the piles. If your pile fits inside the new pile, you believe the new info. If they clash, you have to change your pile.

2. The Five Ways to Move the Bricks

The paper identifies five basic "movements" you can make when you get new information. Imagine you have a pile of bricks (your knowledge) and someone hands you a new brick (the information):

No Change: You ignore the new brick. Your pile stays exactly the same.
Expansion: You just dump the new brick on top of your pile, even if it doesn't fit or creates a mess. You believe everything, even if it's contradictory.
Revision: You want to keep your pile, but you also want to keep the new brick. You rearrange your pile, throwing away some old bricks so the new one fits perfectly.
Contraction: You want to keep the new brick, but you don't want to believe the opposite of it. You throw away the bricks in your pile that contradict the new one, but you don't necessarily add the new brick itself.
Moderation: This is the "Goldilocks" move. You don't fully accept the new brick, nor do you fully reject it. You keep some of your old bricks, keep some of the new ones, and throw away the rest. You end up with a mix.

3. The "Versatile" Operator: The Swiss Army Knife

Before this paper, scientists had different tools for different jobs. One tool for Revision, another for Contraction, another for "Credibility" (when you aren't sure if the new info is trustworthy).

The authors created a new tool called the Versatile Operator. Think of this as a Swiss Army Knife.

It is a single, flexible mechanism that can perform any of the five movements above.
Depending on the situation, it can act like a Revision tool, a Contraction tool, or a Moderation tool.
This unifies all the different theories into one big, coherent family.

4. The Bridge to Different Logics

The biggest achievement of this paper is building a bridge.

The Old Way: You had to rebuild the whole library system every time you switched from "Classical Logic" (rigid) to "Non-Classical Logic" (flexible).
The New Way (AWS): Because they started with the basic "Lego bricks" (Worlds) instead of the complex language, they can apply the same rules to any type of logic.

They show that if you have a logic system that follows certain basic rules (called Tarskian logics), you can translate your "World" rules directly into "Library" rules. This means the same versatile Swiss Army Knife works for:

AGM: The classic, strict library rules.
KM: Rules for updating beliefs about a changing world.
Multiple Change: Rules for handling many updates at once.

Summary

The paper argues that by stripping away the complex language of logic and focusing on the simple concept of "Worlds" (possible scenarios), we can create a universal framework for belief change. This framework:

Treats all types of belief updates (from strict to flexible) as variations of the same process.
Introduces "Versatile" operators that can handle any situation.
Proves that this simple, semantic approach works for almost any logical system, unifying decades of separate research into one clear picture.

In short, they took the complicated machinery of belief change, took it apart, and rebuilt it using simple Lego bricks so that it works for everyone, everywhere.

Technical Summary: An Abstract Worlds Semantic Framework for Belief Change Operators

Problem Statement
The theory of belief change, anchored by the seminal AGM framework (Alchourrón et al., 1985), has historically been tied to specific logical syntaxes, particularly Classical Propositional Logic (CPL). While subsequent frameworks like Katsuno-Mendelzon (KM) and Multiple Change have expanded the scope, most operators remain structurally bound to their underlying logic. This dependency hinders cross-framework comparison and the adaptation of operators to non-classical logics. Furthermore, existing studies on non-classical logics often rely solely on classical AGM operators, failing to account for the diversity of non-prioritized operators (e.g., shielded contraction, credibility-limited revision, filtered revision, and moderated revision). There is a need for a unified, syntax-independent semantic framework that can systematize these diverse operators and generalize belief change theory beyond AGM-compliant logics.

Methodology
The authors introduce Abstract Worlds Semantics (AWS), a set-theoretic framework where "worlds" are treated as primitive elements of an arbitrary index set $I$ , independent of any specific logical syntax. This approach generalizes Grove's possible world semantics (1988) by utilizing the algebra of sets $(\mathcal{P}(I), \cap, \cup, ^c, \subseteq)$ as the foundational structure.

The methodology proceeds in three main stages:

Semantic Definition of Movements: The framework defines belief change not as a single operator but as a set of "movements" (Expansion, Contraction, Revision, No Change, and Moderation) based on the intersection and inclusion of sets representing knowledge ( $A$ ) and new information ( $B$ ).
Operator Construction via Selection Functions: The authors define elementary world operators (contraction $-_w$ and revision $*_w$ ) using world-selection functions $\sigma$ . They introduce a new family of versatile operators ( $\circ_w$ ) that unify these elementary operators. Versatile operators are defined by two selection functions, $\sigma$ and $\rho$ , allowing them to cover the full range of movements, including the "Moderation" movement which lies between contraction and revision.
Translation to Tarskian Logics: To connect the abstract semantic framework to concrete logical systems, the authors apply Suszko's Reduction and Galois Connections. They map the AWS framework onto Tarskian logics (logics defined by a consequence operator satisfying inclusion, monotony, and idempotence). This involves establishing a bijection between belief sets (closed sets of formulas) and closed sets of valuations, allowing the translation of world-based postulates into belief set postulates.

Key Contributions

The Versatile Operator Family: The paper defines elementary world versatile operators as a unifying class that encompasses classical contraction, revision, and non-prioritized operators. These operators satisfy a range of postulates allowing them to behave as expansions, revisions, contractions, or moderated changes depending on the input.
Systematization of Non-Prioritized Operators: The framework provides a homogeneous account of non-prioritized operators (shielded contraction, credibility-limited revision, filtered revision, and moderated revision) by characterizing them as subfamilies of versatile operators. This is achieved by defining credible and questionable relations ( $\prec_w, \succ_w$ ) derived from the operator's behavior, rather than relying on external sets of formulas.
Generalization to Tarskian Logics: The authors demonstrate that AWS is not limited to CPL. By utilizing Tarskian representations, they show how belief change operators can be defined for any Tarskian logic, including those that are not AGM-compliant (i.e., where the recovery postulate may fail).
Unified Representation Theorems: The paper establishes representation theorems showing that:
- Elementary world contraction and revision are duals constructed by the same selection function.
- Versatile operators can be constructed from pairs of selection functions or pairs of associated contraction/revision operators.
- In T1 Tarskian logics (which include CPL), classical constructions (choice remainders and package openings) are isomorphic to constructions based on classical world-selection functions.

Results

Unification of Frameworks: The AWS framework successfully recovers the AGM, KM, and Multiple Change frameworks when restricted to belief sets in Classical Propositional Logic.
Duality and Levi/Harper Identities: The paper re-examines the Levi and Harper identities within the versatile context. It shows that while elementary contraction and revision are extreme positions, versatile operators occupy the intermediate space, effectively generalizing the duality between contraction and revision.
Characterization of Non-Prioritized Operators:
- Shielded Contraction is shown to be a subfamily of versatile operators where the "questionable" relation is trivial.
- Credibility-Limited (CL) Revision and Filtered Revision are characterized by specific entanglement properties between the credible and questionable relations.
- Moderated Revision is defined as a versatile operator satisfying specific consistency and vacuity postulates.
AGM-Compliance Analysis: The paper identifies T1 Tarskian logics (those with topological and complete representations) as the class of logics that are AGM-compliant. It proves that in these logics, the standard AGM constructions (remainders) correspond directly to world-selection functions, while in non-AGM-compliant logics, versatile operators can still be defined axiomatically without requiring the recovery postulate.

Significance and Claims
The authors claim that AWS provides a purely semantic foundation for belief change that decouples the theory from specific logical syntax. By treating worlds as primitive elements, the framework simplifies and generalizes belief change theory over belief sets.

The significance of the work lies in its ability to:

Systematize disparate belief change frameworks and operators under a single abstract construction.
Generalize belief change theory to non-AGM-compliant logics, offering a formal connection between semantic structures and Tarskian logics.
Reinterpret the duality between contraction and revision, showing how non-prioritized operators fit into the broader landscape of belief change.
Standardize concepts like credibility and questionability as relations inherent to the operator's behavior, rather than external constraints.

The paper concludes that AWS establishes a foundation for a research program rather than a single formal result, offering a systematic translation mechanism that can be applied to future analyses of non-classical logics (e.g., modal, paraconsistent, description logics) and other operators (e.g., update, erase), though these specific applications are left for future work.