Dynamic Modal Logics of Awareness

• Dynamic Modal Logics of Awareness are formal frameworks that capture agents' knowledge and explicit awareness, defining how these evolve under dynamic updates such as learning, forgetting, and public announcements.
• They employ Kripke-style models enhanced with awareness sets, explicit modalities, and reduction axioms to systematically update epistemic states via event and action models.
• Recent developments extend these logics with probabilistic belief updates and argumentation-based dynamics, offering deeper insights for applications in economics, artificial intelligence, and epistemic planning.

Dynamic modal logics of awareness provide a rigorous framework for studying the interactions between agents' knowledge, their explicit and implicit awareness of formulas, and how these evolve under dynamic changes such as learning, forgetting, and information disclosure. Drawing from a mature tradition in epistemic logic, these frameworks deploy Kripke-style models decorated with awareness sets, explicit awareness modalities, and reduction axioms that govern transformation under epistemic and awareness-changing actions. Recent work extends the tradition with probabilistic beliefs, event-model-based update mechanisms, and precise closure theorems, collectively characterizing both the expressive and operational scope of awareness dynamics (Halpern et al., 2020, Ditmarsch et al., 2013, Proietti et al., 2023, Burrieza et al., 2021).

1. Formal Language and Static Awareness Models

The core languages of dynamic awareness logics extend classical modal syntax with explicit awareness modalities. For a set of agents $Ag$, atoms $At$, and a formula-building grammar,

$$\varphi ::= \top \mid p \mid \neg\varphi \mid (\varphi \wedge \varphi) \mid \Box_i \varphi \mid A_i \varphi$$

where $p \in At$, $i \in Ag$, $\Box_i$ is epistemic possibility (Kripke box), and $A_i$ is explicit awareness. Some variants admit further modalities: $K_i^\mathrm{exp}\varphi$ for explicit knowledge (requiring awareness of all atoms in $\varphi$), $K_i^\mathrm{spec}\varphi$ for speculative knowledge (quantifying over awareness-bisimilar states), as well as probabilistic operators $At$0 for agent $At$1's (conditional) beliefs (Halpern et al., 2020, Ditmarsch et al., 2013).

An awareness model is a tuple $At$2 where $At$3 is a nonempty set of possible worlds, $At$4 encodes epistemic accessibility, $At$5 is atomic valuation, and $At$6 assigns the set of atoms agent $At$7 is aware of at $At$8. Awareness of complex formulas is atomically closed: $At$9 holds iff all atoms in $$\varphi ::= \top \mid p \mid \neg\varphi \mid (\varphi \wedge \varphi) \mid \Box_i \varphi \mid A_i \varphi$$0 belong to $$\varphi ::= \top \mid p \mid \neg\varphi \mid (\varphi \wedge \varphi) \mid \Box_i \varphi \mid A_i \varphi$$1 (Proietti et al., 2023).

Truth clauses for the core connectives are defined as:

• $$\varphi ::= \top \mid p \mid \neg\varphi \mid (\varphi \wedge \varphi) \mid \Box_i \varphi \mid A_i \varphi$$2 iff $$\varphi ::= \top \mid p \mid \neg\varphi \mid (\varphi \wedge \varphi) \mid \Box_i \varphi \mid A_i \varphi$$3 with $$\varphi ::= \top \mid p \mid \neg\varphi \mid (\varphi \wedge \varphi) \mid \Box_i \varphi \mid A_i \varphi$$4
• $$\varphi ::= \top \mid p \mid \neg\varphi \mid (\varphi \wedge \varphi) \mid \Box_i \varphi \mid A_i \varphi$$5 iff $$\varphi ::= \top \mid p \mid \neg\varphi \mid (\varphi \wedge \varphi) \mid \Box_i \varphi \mid A_i \varphi$$6

This foundation supports further structures for implicit, explicit, and speculative knowledge and argumentation-based attitudes, as well as probabilistic measures for belief (Halpern et al., 2020, Burrieza et al., 2021).

2. Dynamic Update Mechanisms: Event and Action Models

Dynamic awareness logics deploy epistemic event models or action models to describe, in a compositional manner, how knowledge and awareness change due to external actions (e.g., public announcement, discovery, forgetting) (Proietti et al., 2023, Ditmarsch et al., 2013).

An event model for awareness updates is a tuple $$\varphi ::= \top \mid p \mid \neg\varphi \mid (\varphi \wedge \varphi) \mid \Box_i \varphi \mid A_i \varphi$$7 where

• $$\varphi ::= \top \mid p \mid \neg\varphi \mid (\varphi \wedge \varphi) \mid \Box_i \varphi \mid A_i \varphi$$8 is a finite set of events,
• $$\varphi ::= \top \mid p \mid \neg\varphi \mid (\varphi \wedge \varphi) \mid \Box_i \varphi \mid A_i \varphi$$9 specifies agents' uncertainty about events,
• $p \in At$0 are preconditions,
• $p \in At$1 are sets of atoms gained or lost in awareness for agent $p \in At$2 (Proietti et al., 2023).

The product-update semantics yields a new model $p \in At$3 where worlds are $p \in At$4 such that $p \in At$5, accessibility is combined from $p \in At$6 and $p \in At$7, and awareness sets are updated as

$p \in At$8

This allows composition of awareness modifications, various agents' perspectives on event occurrence, and public/private update protocols (Ditmarsch et al., 2013).

Alternative frameworks—such as the model-transition approach in "Dynamic Awareness"—define how awareness expansions (and probability distributions over knowledge) are constructed in the successor model, often via shadow symbols and replacement schemes, supporting both explicit and probabilistic belief change (Halpern et al., 2020).

3. Properties, Axiomatization, and Reduction Principles

Dynamic awareness logics satisfy a suite of meta-properties, many inherited from static modal logic but refined for the presence of awareness.

Introspection and Persistence: Awareness modalities often satisfy positive and negative introspection for atoms, i.e., $p \in At$9 and $i \in Ag$0 (Proietti et al., 2023). Persistence and no-forgetting are typically enforced: once $i \in Ag$1 becomes true, it remains so in all epistemically accessible worlds after the update (Halpern et al., 2020).

Closure Theorems: Precise syntactic conditions on event models guarantee preservation of desired properties (e.g., PRES: $i \in Ag$2) under product update. For instance, safety for PRES requires that the positive effects on awareness grow along $i \in Ag$3 and the negative effects contract (Proietti et al., 2023).

Axiomatics and Reduction: The logic is axiomatized by standard modal axioms (K, TAUT, necessitation) plus families of awareness introspection and closure schemas. Dynamic modalities $i \in Ag$4 (for event $i \in Ag$5 in model $i \in Ag$6) admit reduction axioms to rewrite dynamic formulas into the static fragment, ensuring strong completeness via well-founded induction (Proietti et al., 2023, Ditmarsch et al., 2013). Similarly, in "Dynamic Awareness" the dynamic discovery modality $i \in Ag$7 is governed by reduction principles, e.g., $i \in Ag$8 for non-overlapping formulas, and other schemas describe how probabilities and knowledge are updated (Halpern et al., 2020).

4. Variants: Probabilistic and Argumentation-based Awareness

Recent research extends dynamic awareness logics in significant directions:

Probabilistic Awareness: The system in "Dynamic Awareness" augments static models with agent-indexed probability distributions $i \in Ag$9 over accessible worlds, supporting modal operators $\Box_i$0. Model transitions after awareness growth are regulated by commensurability (extended Bayesianism): after expansion, the new measure $\Box_i$1 is a conditional of the prior $\Box_i$2 on the set of worlds whose language was rich enough to encode the discovered formula (Halpern et al., 2020). This yields formal theorems characterizing when belief update is representable as conditioning.

Argumentation and Belief Dynamics: In the Burrieza–Yuste‐Ginel framework, the logic is centered on awareness of structured arguments and the corresponding belief in conclusions. Updates include acquiring/forgetting arguments and rules or receiving public announcements, all with dynamic modalities. The framework connects awareness with ASPIC-style argumentation, tracks explicit and argument-based belief, and supports a complete reduction-style axiomatization (Burrieza et al., 2021).

5. Expressivity, Bisimulation, and Model-Theoretic Characterization

Bisimulation techniques characterize expressivity relationships among logics of awareness. Three principal logics are typically compared: implicit, explicit, and speculative knowledge, all relative to awareness (Ditmarsch et al., 2013). In the purely static setting, implicit knowledge (with $\Box_i$3) is strictly more expressive than either explicit or speculative knowledge, the latter two being equally expressive.

Upon extending each logic with suitable dynamic modalities for awareness action models, the expressivity difference collapses: all three dynamic logics become pairwise equally expressive, as one can systematically translate update modalities between them via reduction axioms and bisimulations (Ditmarsch et al., 2013). This result underscores the robustness of product-update logics in capturing a wide range of dynamic awareness phenomena.

6. Illustrative Examples and Applications

Classic examples include:

• Becoming aware of an atom $\Box_i$4: A singleton event model adding $\Box_i$5 to an agent's awareness, leaving all epistemic information unchanged (Ditmarsch et al., 2013, Proietti et al., 2023).
• Forgetting $\Box_i$6: An event that removes $\Box_i$7 from awareness (Ditmarsch et al., 2013, Proietti et al., 2023).
• Public announcement of a formula $\Box_i$8: All agents gain awareness of atoms in $\Box_i$9, but knowledge and propositional structure are unaffected (Ditmarsch et al., 2013, Proietti et al., 2023).
• Economics/Disclosure scenarios: In "Dynamic Awareness," information disclosure (e.g., ratings) influences both epistemic state and probabilistic assessment, as the update rule composes a model extension (shadow replacement, awareness gain) with Bayesian conditioning (Halpern et al., 2020).
• Argument-based retraction: In the awareness-argumentation framework, learning new rules or arguments can retract previously held beliefs due to updated awareness and argument structure (Burrieza et al., 2021).

7. Connections, Open Problems, and Theoretical Significance

Dynamic modal logics of awareness bridge to multiple areas: justification logics, abstract argumentation theory, epistemic logics with "knowing what/knowing how," and deontic logic, as awareness modalities can effectively encode nuanced epistemic/attitudinal distinctions (Proietti et al., 2023). Preservation and closure theorems provide structural guarantees for principled update protocols.

Current open problems include: direct axiomatization of argument-based belief operators, extension of single-agent frameworks to fully multi-agent interaction (e.g., dialogue games), and integration of plausibility/preference orderings more fine-grained than the strict/defeasible dichotomy (Burrieza et al., 2021). These frameworks continue to inform both the specification of agent epistemics and the meta-theory of dynamic logic broadly.