First-hand information for everyone
A recent study shows that the simple formal language of basic modal logic (ML) is not merely a toy system but is precisely the set of all first‑order logic (FO) properties that are unchanged under bisimulation— a key notion of equivalence in models of dynamic systems. This means that any property expressible in FO that behaves the same in bisimilar worlds can be written in ML and vice versa.
The result is proved by relating ML to a two‑player back‑and‑forth game, a classic method for comparing structures. The game guarantees that two points in possibly different models are bisimilar exactly when the second player can always mimic the moves of the first player for a given number of rounds. From this game correspondence, the authors derive that ML distinguishes exactly the bisimulation‑invariant FO formulas.
The authors also extend the framework. By adding a global modality— an operator that can refer to any point in the entire structure— they characterize the global bisimulation‑invariant fragment of FO. Thus, ML with the global modality captures every FO property that remains unchanged under global bisimulation.
However, the paper notes a limitation. When counting quantifiers are introduced without a fixed numerical bound, the resulting counting bisimulation equivalence has infinitely many distinct equivalence classes. Consequently, characteristic formulas and expressive‑completeness results become unavailable in that unbounded setting.
These insights clarify the exact expressive power of modal logics and set boundaries for when extensions like counting remain tractable.