Belief contraction, inquisitive action, and quantum bases: AiML 2026 pulls modal logic toward update problems

Modal logic's newest pressure point is not possible worlds by themselves. It is what happens when agents revise, contract, coordinate, determine, or justify within those worlds.

AiML 2026 makes that pressure unusually visible. The open-access EPTCS 447 volume lists 43 full papers accepted from 87 regular submissions, plus 15 short presentations, and the editors describe the accepted work as spanning dynamic epistemic logic, belief revision and belief change, team and inquisitive logics, constructive and intuitionistic modal logics, probability logic, evidence logic, topology, bundled modalities, causal modalities, and related proof-theoretic questions.1 The arXiv record for the volume was submitted on 28 June 2026 and identifies AiML 2026 as the sixteenth conference in the series, held in Amsterdam from 29 June to 3 July 2026.2

That is the week’s real signal: not one breakout paper, and not yet a citation wave. It is a concentrated batch of papers treating modal logic as a technology for update, agency, and evidence under non-classical constraints.

The belief-change papers are the cleanest sign. Belardinelli and Zhang push dynamic epistemic logic toward contraction, including cases where a public announcement says that a proposition might be false rather than simply removing possibilities.3 Camargo's preprint is adjacent rather than part of AiML, but it asks a parallel question: how much of AGM belief revision survives when the epistemic state includes quantitative degrees of conviction?8

The knowledge papers shift the update problem from single-agent revision to coordination under partial awareness. "Resolving Asynchronous Distributed Knowledge" keeps the distributed-knowledge machinery but removes the common global clock; truth is evaluated relative to a history of previous resolutions, and each agent observes only part of that history.4 That is a stronger fit for distributed computing and multi-agent settings than synchronous public update models.

The agency and inquisitive-logic thread changes what the modal operator is tracking. InqAL distinguishes forcing an outcome from determining an aspect of the outcome, and treats determination claims as modal claims involving questions.5 This connects modal logic to formal semantics more directly than a standard coalition-ability reading, because questions and issue resolution become part of the logical object.

The non-classical base papers are doing a different job. Aguilera and Massas ask how far constructive and quantum logics can be combined without losing algebraic control.6 Sedlar asks whether topological evidence logic depends on a classical propositional base, and answers by rebuilding coherent justification in intuitionistic and relevant settings.7 Together, they make the same methodological point from opposite sides: if modal logic is going to model agency, evidence, or quantum structure, the background propositional logic cannot be treated as a harmless default.

The citation signal is still too thin for a strong trend claim. Semantic Scholar currently records the AiML proceedings volume with 0 citations and 0 influential citations, with publication date 28 June 2026.9 It records "Belief Contraction in Dynamic Epistemic Logic" with 31 references, 0 citations, and 0 influential citations, with publication date 29 June 2026.10

That does not weaken the selection. It changes what the selection means. This is an early publication-clustering issue, not a citation-momentum issue. The thing to watch over the next few months is whether the belief-contraction and asynchronous-knowledge papers start linking to the same DEL and AGM reference base, and whether the inquisitive-action paper pulls formal-semantics readers toward the AiML proceedings.

AiML 2026 shows modal logic becoming more operational without becoming merely computational. The papers above still ask recognizably philosophical questions: what it is to give up a belief, what an agent can determine, what counts as coherent evidence, and what happens when a logic's base is constructive, quantum, relevant, or topological.

The technical work is doing the sorting. Some formalisms can handle revision but not contraction. Some can represent shared knowledge only when everyone knows the update clock. Some can talk about forcing outcomes but not determining questions. The interesting movement this week is that those limits are being stated as theorems and model conditions, not as programmatic slogans.