🔢 mathematics

A Semantics for Belief in Simplicial Complexes

This paper introduces a novel semantics for belief using simplicial complexes that successfully distinguishes belief from knowledge under standard KD45 conditions, establishes a truth-preserving correspondence with relational models, and provides a simple axiomatization.

Original authors: Adam Bjorndahl (Carnegie Mellon University), Philip Sink (Caregie Mellon University)

Published 2026-07-01

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Adam Bjorndahl (Carnegie Mellon University), Philip Sink (Caregie Mellon University)

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

The Big Picture: Mapping Minds with Shapes

Imagine you are trying to map out what a group of people knows and what they believe. In computer science and logic, we usually do this with "maps" made of dots and lines (relational models).

The authors of this paper are trying to do the same thing, but using shapes instead of dots and lines. Specifically, they are using simplicial complexes.

The Shape Analogy: Think of a simplex as a multi-dimensional triangle.
- 2 points make a line (a 1D triangle).
- 3 points make a standard triangle (a 2D triangle).
- 4 points make a tetrahedron (a 3D triangle).
The Rules: In their system, every "biggest" shape (called a facet) must have exactly one corner (vertex) for every person (agent) in the group. If there are three people (Alice, Bob, and Charlie), every shape in the map must have exactly three corners: one for Alice, one for Bob, and one for Charlie.

The Problem: Knowing vs. Believing

The authors wanted to use these shapes to represent two different things: Knowledge and Belief.

Knowledge (The "Truth" Rule): If you know something, it must be true. In their shape system, this works perfectly. If a shape represents a possible world, and Alice "knows" a fact, that fact is true in all the shapes she can see.
Belief (The "Maybe" Rule): If you believe something, it might be false. You can believe the sun is green, even if it's not.

The Conflict:
The authors discovered a major snag. In their shape system, the rules for "Knowledge" are so strict that they accidentally force "Belief" to be true as well.

The Analogy: Imagine a room where everyone is holding a flashlight. If you are standing in a room where everyone agrees on the facts, you can't really "believe" something false without breaking the rules of the room.
If they tried to simply shrink the map to represent beliefs (like removing some shapes), they found that the agents would still be forced to know everything they believed. This makes "belief" useless because it becomes identical to "knowledge."

The Solution: The "Belief Sub-Map"

To fix this, the authors came up with a clever trick. Instead of having just one big map of shapes, they gave each person their own private "belief sub-map."

The Main Map (Knowledge): This is the big, shared map of all possible worlds. Everyone agrees on the rules here.
The Private Maps (Belief): Each agent (Alice, Bob, etc.) has their own smaller map inside the big one.
- When Alice thinks about what she believes, she only looks at the shapes in her private map.
- When Bob thinks about what he believes, he only looks at his private map.

Why this works:
Because Alice's private map is different from Bob's, Alice can believe something that is false in the main map (and in Bob's map). This allows for "false beliefs" without breaking the rules of the system. It solves the problem where belief was accidentally forced to be true.

The Translation Challenge: The "Properness" Hurdle

The authors wanted to prove that their new shape system is just as good as the old dot-and-line system. They tried to translate a standard "dot-and-line" belief model into their "shape" model.

The Problem:
They found that some standard belief models couldn't be translated directly.

The Analogy: Imagine a standard map where three people all stand in the exact same spot and look at the exact same three houses. In the shape world, you can't have three people standing in the exact same spot; the rules say every shape must have distinct corners for everyone.
If you try to force the standard map into the shape rules, you lose information. The shape system seemed to demand that every person has a unique "perspective" that no one else shares, which isn't always true in standard logic.

The Fix (The "Copy-Paste" Trick):
The authors proved a surprising result: You can always fix this.
If a standard map doesn't fit the shape rules, you can simply copy and paste the whole world multiple times to create a "proper" version.

Imagine you have a small group of friends. To make them fit the shape rules, you create a duplicate group, then another, and another.
In this new, larger world, every person now has a unique "copy" of themselves in different groups.
This new, larger world behaves exactly like the original one (logically speaking), but it fits the shape rules perfectly.

This is a huge deal because it means their shape system can represent any standard belief scenario, provided you are willing to use a slightly larger map.

The Result: A New Logic System

Because they solved the translation problem, the authors were able to create a simple set of rules (an axiom system) for this new shape-based logic.

They proved that their system is Sound (it never produces nonsense) and Complete (it can prove everything that is logically true).
They showed that their system handles the relationship between knowledge and belief correctly: "If you know it, you believe it," but "You can believe things you don't know."

What About Other Approaches?

The paper briefly mentions that other researchers have tried to solve this before, but they had to break the rules (like allowing multiple corners for the same person in one shape) or ended up making belief identical to knowledge. The authors argue their method is the cleanest way to keep the rules strict while still allowing for human-like, fallible belief.

Summary

The Goal: Create a logic system for belief using geometric shapes (simplicial complexes) instead of dots and lines.
The Obstacle: The shape rules naturally force "belief" to be "true," which defeats the purpose of belief.
The Innovation: Give each agent their own private "belief sub-map" inside the main shape structure. This allows for false beliefs.
The Breakthrough: They proved that even if a standard belief model doesn't fit the shape rules, you can "copy and paste" the world to make it fit without changing the logic.
The Outcome: A robust, mathematically proven system where belief and knowledge are distinct, and the shape system works perfectly with standard logic.

Technical Summary: A Semantics for Belief in Simplicial Complexes

Problem Statement
The paper addresses the challenge of extending simplicial complex semantics, a thriving area in multi-agent epistemic logic, to accommodate belief alongside knowledge. While simplicial semantics successfully models knowledge (where facets represent possible worlds and vertices represent agent perspectives), it inherently hardcodes factivity ( $K\phi \to \phi$ ) because an agent's knowledge is defined by quantifying over facets sharing a vertex with the current facet. Belief, however, must be non-factive ( $B\phi \not\to \phi$ ) to allow for false beliefs.

The authors identify a critical obstacle: standard attempts to model belief in simplicial complexes by simply restricting the set of quantified facets (e.g., using a subcomplex $S' \subseteq S$ ) fail to satisfy the condition that knowledge implies belief ( $K\phi \to B\phi$ ) without trivializing belief to coincide with knowledge. Specifically, if a single subcomplex is used for all agents, agents cannot consider it possible that other agents hold false beliefs, violating the intended epistemic structure. Furthermore, existing approaches in the literature either violate the "Uniquely Colored Facets" (UCF) condition (where each facet contains exactly one vertex per agent) or fail to validate $K\phi \to B\phi$ .

Methodology
The authors propose a novel framework where belief is modeled by enriching the simplicial model with multiple belief subcomplexes, one for each agent.

Simplicial Belief Models: A model is defined as a tuple $(N, V, S, \{S_a\}_{a \in Ag}, L)$ , where $(N, V, S, L)$ is a standard simplicial model for knowledge, and each $S_a$ is a non-empty simplicial subcomplex of $S$ satisfying the UCF condition.
- Knowledge ( $K_a$ ): Quantifies over all facets in $S$ sharing an $a$ -colored vertex.
- Belief ( $B_a$ ): Quantifies only over facets in the specific subcomplex $S_a$ sharing an $a$ -colored vertex.
- This separation allows $B_a$ to be non-factive (since $S_a$ may not contain the current facet) while ensuring $K_a \to B_a$ holds (since $S_a \subseteq S$ ).
Translation to Relational Semantics: The paper establishes a truth-preserving correspondence between these simplicial models and standard relational models (Kripke frames) for introspective knowledge and belief.
- A major hurdle is the "properness" condition required for the standard translation from relational to simplicial models. A relational model is "proper" if no two distinct worlds are indistinguishable by all agents.
- The authors demonstrate that many natural belief models (where agents have constant belief sets) are not proper and thus cannot be directly translated.
Simulation via Proper Models: To overcome the properness restriction, the authors prove a general result (Theorem 4.5) that any relational model is bisimilar to a proper relational model.
- They construct a new model $\tilde{N}$ by creating multiple copies of the original world set $W$ (specifically $W \times W$ ).
- Accessibility relations are defined such that they "skew" across these copies for one distinguished agent, ensuring that no two distinct points in $\tilde{N}$ are indistinguishable by all agents.
- This construction preserves all relevant logical properties (reflexivity, symmetry, transitivity, seriality, Euclidean property) and validates the same formulas as the original model via a surjective bounded morphism.
Axiomatization: Using the translation results, the authors provide a sound and complete axiomatization for their framework. The system FULL combines S5 axioms for knowledge and KD45 axioms for belief, augmented with the interaction axiom $K_a\phi \to B_a\phi$ and strong introspection axioms ( $B_a\phi \to K_a B_a\phi$ and $\neg B_a\phi \to K_a \neg B_a\phi$ ).

Key Contributions and Results

Novel Semantics: The paper introduces a simplicial semantics for belief that satisfies three critical conditions simultaneously: (1) adherence to the UCF condition, (2) validity of $K\phi \to B\phi$ , and (3) non-triviality (belief does not collapse into knowledge).
Independence Result: The proof that any relational model can be simulated by a proper relational model (Theorem 4.5) is presented as a result of independent interest, effectively dissolving the "properness" restriction that has historically limited the translation between relational and simplicial frameworks.
Completeness: The paper establishes the soundness and completeness of the FULL axiom system with respect to the class of simplicial belief models.
Comparative Analysis: The authors rigorously contrast their approach with two existing frameworks ("Knowledge and Simplicial Complexes" and "Simplicial Belief"), demonstrating that those approaches fail to satisfy the combination of UCF and $K\phi \to B\phi$ without trivialization.

Significance
The paper claims to provide a "simple axiomatization" and a robust semantic foundation for belief in simplicial complexes that resolves long-standing subtleties in the field. By proving that the properness restriction is not a fundamental barrier but a technical one that can be circumvented via simulation, the authors enable the direct application of simplicial methods to standard epistemic belief problems.

The authors modestly note that while their current approach requires proliferating perspectives (worlds) to achieve properness—potentially leading to structural redundancy—they view simplicial sets (which allow for multiple facets on the same set of vertices) as a promising future direction to represent non-proper models more naturally without artificial proliferation.