Simplicial Semantics in Epistemic Modal Logic

• The paper shows that leveraging impure simplicial complexes allows a geometric refinement of standard Kripke semantics in dynamic multi-agent systems.
• It introduces a categorical equivalence between chromatic simplicial complexes and PER frames, clarifying agent survival and knowledge under varying conditions.
• The study applies these methods to distributed computing, proving impossibility results and highlighting consensus challenges amid process failures.

Simplicial semantics for epistemic modal logic interprets the knowledge modalities of multi-agent systems on combinatorial structures called (chromatic) simplicial complexes, providing a geometric refinement of standard Kripke (possible-world) semantics. A key innovation is the use of impure simplicial complexes to model settings in which the ensemble of agents participating in a global state varies dynamically, as in distributed systems with process crashes or failing agents. In this framework, Kripke-model accessibility relations are replaced by higher-dimensional combinatorial configurations, revealing topological invariants and supporting a broader class of epistemic logics, notably including the modal logic $KB4_n$ for "agents that may die" (Goubault et al., 2021). Simplicial semantics supports a categorical equivalence with partial equivalence relation (PER) frames and enables precise reasoning about knowledge in distributed and dynamic environments.

1. Combinatorial and Categorical Structure

A simplicial complex $C = (V,S)$ consists of a finite set of vertices $V$ and a downward-closed nonempty family of nonempty finite subsets $S$ (the simplices), with the property that if $X\in S$ and $\emptyset\neq Y\subseteq X$, then $Y\in S$. A facet is a maximal simplex; its cardinality minus one is the dimension of the facet.

A chromatic simplicial complex additionally specifies a coloring map $\chi: V \to A$ for a finite agent set $A$, such that the coloring is injective on each simplex (no two vertices of the same simplex share a color). The notion of purity is central: $C$ is pure if all facets have equal cardinality (i.e., all agents are always present); otherwise, $C = (V,S)$0 is impure, and some facets omit agents, modeling agent "death".

A simplicial model is a tuple $C = (V,S)$1, where $C = (V,S)$2 labels facets (or, in some variants, vertices) with sets of propositional atoms true in each world. These models can be interpreted as epistemic frames by defining an $C = (V,S)$3-accessibility relation $C = (V,S)$4 on facets $C = (V,S)$5.

The category of (proper) chromatic simplicial complexes is equivalent to the category of PER-frames, where each $C = (V,S)$6 is a partial equivalence relation: symmetric and transitive but not necessarily reflexive. Facets serve as possible worlds, with the agents alive in a world $C = (V,S)$7 identified by $C = (V,S)$8. The categorical equivalence aligns labeling, morphisms, and knowledge indistinguishability relations directly with combinatorial structure (Goubault et al., 2021, Goubault et al., 2023, Bílková et al., 2024).

2. Simplicial Semantics for KB4, S5, and Variants

Simplicial semantics extends naturally to modal logics ranging from S5 (full equivalence relations) to KB4 (partial equivalence relations, allowing non-reflexivity—i.e., dead agents). For a formula $C = (V,S)$9 in the modal language built from agents $V$0 and atomic propositions $V$1:

• Atomic: $V$2 iff $V$3
• Knowledge: $V$4 iff for all facets $V$5 with $V$6, $V$7

In pure complexes (S5), $V$8 interprets knowledge using the equivalence relation structure between facets. In impure complexes, $V$9 uses the PER structure: if agent $S$0 is dead in $S$1 (i.e., $S$2), the $S$3-modality becomes trivial—every formula is vacuously known by $S$4 at such $S$5 (i.e., $S$6 is always true).

Axioms characterizing the logic for impure models (KB4) include:

• K ($S$7)
• B ($S$8) (symmetry)
• 4 ($S$9) (transitivity)
• Absence of T ($X\in S$0)—as dead agents have no reflexive points

Additional axioms (e.g., NE—no empty world, SA$X\in S$1—single-agent worlds) may be imposed to enforce properness (Goubault et al., 2021, Goubault et al., 2023). Soundness and completeness theorems relate these axiomatizations to their respective model classes; for instance, KB4 with NE and (SA$X\in S$2) is sound and conjectured complete for proper impure simplicial models.

3. Semantics of Alive and Dead Agents

In impure complexes, the combinatorial geometry of $X\in S$3 tracks agent lifespans: an agent $X\in S$4 is alive in a facet $X\in S$5 iff $X\in S$6. Modal language is enriched to express alive/dead status with atoms such as $X\in S$7 and $X\in S$8.

Semantics for knowledge must be adapted:

• $X\in S$9 is undefined or vacuously true if $\emptyset\neq Y\subseteq X$0 is dead in the current world.
• Many logics use a three-valued semantics: formulas are true, false, or undefined, where undefined occurs if a formula refers to a dead agent or to knowledge that cannot be evaluated due to absent agents (Ditmarsch et al., 2021, Bílková et al., 2024).

Variants of group knowledge (mutual, distributed, common) are defined in terms of colored intersections or via groupwise indistinguishability:

• $\emptyset\neq Y\subseteq X$1 iff for all $\emptyset\neq Y\subseteq X$2 with $\emptyset\neq Y\subseteq X$3, $\emptyset\neq Y\subseteq X$4.

Such group modalities clarify higher-dimensional, topological "connections" underlying agents' joint knowledge in systems with partial participation (Goubault et al., 6 Feb 2026, Goubault et al., 2023, Goubault et al., 2023).

4. Bisimulation and Expressivity

A robust notion of bisimulation is essential for modal invariance and for analyzing expressivity. For impure complexes, a simplicial bisimulation $\emptyset\neq Y\subseteq X$5 between models $\emptyset\neq Y\subseteq X$6 relates facets such that

• (Atoms) $\emptyset\neq Y\subseteq X$7 and $\emptyset\neq Y\subseteq X$8,
• (Forth/Back) for all agents $\emptyset\neq Y\subseteq X$9, for $Y\in S$0 with $Y\in S$1, there exists $Y\in S$2 with $Y\in S$3 and $Y\in S$4 (and conversely).

Such bisimulations preserve the truth of all formulas—including those mentioning alive/dead status—in the global language (Bílková et al., 2024). The Hennessy–Milner property holds for star-finite complexes: modal equivalence implies bisimulation. Omitting alive/dead atoms from the language prevents a full bisimulation theory: it becomes impossible to distinguish between living and dead agents using formulas alone.

Translation functors $Y\in S$5 and $Y\in S$6 yield faithful correspondences between simplicial models and partial Kripke models (local epistemic frames), preserving both bisimulation and epistemic invariance (Goubault et al., 2021, Goubault et al., 2023, Bílková et al., 2024).

5. Applications: Distributed Computing and Fault Tolerance

Simplicial semantics is intrinsically suited to distributed fault-tolerant computing, where processes (agents) may fail (crash) and system topology dynamically changes. Impure complexes model such scenarios by encoding each global state as a simplex whose colors record the set of currently alive agents. Key constructions:

• Input complexes encode initial knowledge (input assignments to agents).
• Protocol complexes are updated via communication patterns (graphs $Y\in S$7 encoding who communicates with/receives from whom per round).
• The protocol complex after a round with possible crash produces an impure complex, with facets of varying dimensions corresponding to survival patterns (Goubault et al., 2021, Goubault et al., 2023).

For instance, in a 3-agent synchronous system, the protocol complex after one round with arbitrary single crash includes:

• One triangle (no one crashes)
• Three edges (exactly one agent crashes)
• Three vertices (all but one crashed)

Simplicial semantics enables elegant impossibility proofs in distributed computing. Consensus and $Y\in S$8-set agreement tasks can be shown unsolvable in a given model by demonstrating the absence of a simplicial map from the protocol complex (with its epistemic properties) to the task complex, using knowledge gain results and topological arguments (Goubault et al., 2021, Goubault et al., 2023, Goubault et al., 6 Feb 2026).

6. Extensions, Limitations, and Open Problems

Simplicial semantics continues to expand:

• Term-modal and assignment languages eliminate undefined or "dubious" formulas about dead agents by tracking variable assignments and restricting quantification to alive agents (Yang, 27 Nov 2025).
• Three-valued logics rigorously formalize undefinedness, with soundness and strong completeness arguments for adapted axiom systems (Ditmarsch et al., 2021).
• Semi-simplicial sets and chromatic hypergraphs generalize simplicial model structure to describe complex agent-environment interactions, where group knowledge may not be reducible to intersections of individual indistinguishabilities (Goubault et al., 2023, Goubault et al., 2023).
• Dynamic epistemic updates (communication pattern logic) allow characterization of knowledge evolution under information flow and message loss, with dynamic modalities and update product constructions mirroring Kripke-product semantics (Castañeda et al., 2022).

A notable limitation is the requirement to syntactically or semantically exclude or manage knowledge statements about agents that are dead in a given world to avoid conceptual pathologies; this motivates refinements of the language and model structure (Yang, 27 Nov 2025). Open problems include the characterization of completeness for certain fragments (e.g., proper impure complexes), generalizations to temporal and quantum settings, and further applications to distributed computing lower bounds and task solvability (Goubault et al., 2023, Goubault et al., 6 Feb 2026).

7. Comparison with Kripke Semantics and Topological Enrichment

While pure simplicial semantics (S5) is equivalent to classical Kripke semantics, impure simplicial semantics extends this by encoding epistemic structure directly in higher-dimensional geometry:

Aspect	Kripke (S5/PER)	Simplicial (Pure/Impure)
World structure	Set of points, edges	Facets of combinatorial complex
Agent perspective	Equivalence/partial equiv.	Colored vertex intersection
Group knowledge	Relation composition	Higher-dimensional intersection
Agent death	Non-reflexive PER	Missing colors in facet
Topological invariants	Invisible	Central (connectivity, homology)

Topological invariants (connectivity, homology) of the complex correspond to epistemic properties such as common, distributed, and mutual knowledge, and are powerful tools in distributed computing and epistemic game theory (Goubault et al., 6 Feb 2026). In impure models, topological features recover crucial impossibility results and the expressivity to model dynamically varying agent sets in epistemic programs (Goubault et al., 2021, Goubault et al., 2023).

---

Key references:

• "A Simplicial Model for $Y\in S$9: Epistemic Logic with Agents that May Die" (Goubault et al., 2021)
• "Simplicial Models for the Epistemic Logic of Faulty Agents" (Goubault et al., 2023)
• "Bisimulation for Impure Simplicial Complexes" (Bílková et al., 2024)
• "Wanted Dead or Alive: Epistemic logic for impure simplicial complexes" (Ditmarsch et al., 2021)
• "Distributed Knowledge in Simplicial Models" (Goubault et al., 6 Feb 2026)
• "Impure Simplicial Complex and Term-Modal Logic with Assignment Operators" (Yang, 27 Nov 2025)
• "Communication Pattern Logic: Epistemic and Topological Views" (Castañeda et al., 2022)
• "Semi-simplicial Set Models for Distributed Knowledge" (Goubault et al., 2023)
• "A many-sorted epistemic logic for chromatic hypergraphs" (Goubault et al., 2023)