Overview
Deontic Defeasible Description Logic, abbreviated here as 3DL, denotes the framework of Casini, Meyer, Peñaloza and van der Torre (DL 2026). It extends defeasible \mathcal{ALC} to reason about conditional obligations and permissions and combines typicality (expectations) with normative preference. The paper uses ranked interpretations and rational-closure-inspired minimal entailment; the abbreviation 3DL is a convenient catalogue label, not a claim that all defeasible deontic DLs share this semantics.
Knowledge Bases and Syntax
The framework considers a knowledge base
\mathcal K=\langle\mathcal T,\mathcal D,\mathcal N,\mathcal P\rangle,
where \mathcal T contains strict \mathcal{ALC} concept inclusions C\sqsubseteq D, \mathcal D contains defeasible expectations, \mathcal N contains conditional obligations, and \mathcal P contains conditional permissions:
C\mathrel{\sqsubseteq\mkern-12mu\sim_E}D,\qquad C\mathrel{\sqsubseteq\mkern-12mu\sim_N}D,\qquad C\mathrel{\sqsubseteq\mkern-12mu\sim_P}D.
The decorated inclusion symbols are explanatory notation for the paper's three distinct kinds of defeasible inclusion. Unlike modal DDL, the arguments are description-logic concepts, not arbitrary propositional formulas about worlds.
Two Ranked Orders
A combined DD-interpretation has the shape
\mathcal R=\langle\Delta^{\mathcal R},\cdot^{\mathcal R}, \prec_E^{\mathcal R},\prec_N^{\mathcal R}\rangle.
The ranked order \prec_E measures expectedness / typicality; the independent ranked order \prec_N measures normative preference. For a concept C, write \min_E(C^{\mathcal R}) and \min_N(C^{\mathcal R}) for the best members under the respective rankings. The three original inclusion types are interpreted as
\min_E(C^{\mathcal R})\subseteq D^{\mathcal R}\quad\text{(expectation)},
\min_N(C^{\mathcal R})\subseteq D^{\mathcal R}\quad\text{(obligation)},
\min_N(C^{\mathcal R})\cap D^{\mathcal R}\ne\varnothing \quad\text{or}\quad C^{\mathcal R}=\varnothing \quad\text{(permission)}.
The empty-premise disjunct reflects the paper's particular convention for permission, dual to negated obligation; it should not be silently replaced by a different permission semantics.
Presumptive Norms
When expectations and norms interact, the paper additionally defines presumptive obligations and presumptive permissions. The idea is to select typical C-individuals first, then assess ideality among that selection. Schematically, the presumptive obligation tests
\min_N(\min_E(C^{\mathcal R}))\subseteq D^{\mathcal R},
rather than simply \min_N(C^{\mathcal R})\subseteq D^{\mathcal R}. The formal treatment of presumptive permissions also accounts for empty antecedents and non-emptiness of normatively preferred typical cases.
Example
A knowledge base can express that typical graduate students work, while normatively students should not take on excessive commitments, with explicit exceptions for employed students. The two rankings separate what normally happens from what ought to happen. For example, a conflict between an expectation \mathit{Student}\mathrel{\sqsubseteq\mkern-12mu\sim_E}\mathit{Works} and an obligation \mathit{Student}\mathrel{\sqsubseteq\mkern-12mu\sim_N}\neg\mathit{Overworked} need not be collapsed into one preorder.
Reasoning and Complexity
The paper constructs a big ranked/DD-model and defines minimal entailment via rational-closure-style rankings and combined ranking procedures, reducing relevant checks to classical \mathcal{ALC} reasoning. Minimal entailment for the studied deontic defeasible knowledge bases is EXPTIME-complete (Proposition 5). This does not establish the same bound for arbitrary ABox-based conjunctive-query answering or for extensions to SROIQ; those are separate research questions.
Relationships
ALC provides the underlying description-logic constructors. Preference-based DDL motivates the optimality treatment, but its worlds and formulas are not interchangeable with DL individuals and concepts. DLP and SROIQ represent different fragments or extensions of description-logical reasoning.
References
- Giovanni Casini, Thomas Meyer, Rafael Peñaloza, and Leendert van der Torre (2026). Deontic Defeasible Description Logic. Proceedings of DL 2026, CEUR-WS Vol. 4230. Full paper.
- Franz Baader, Ian Horrocks, Carsten Lutz, and Uli Sattler (2017). An Introduction to Description Logic. Cambridge University Press.
- Xavier Parent et al. (2024). Conditional Normative Reasoning as a Fragment of HOL. DOI.