「‍」 Lingenic

Abstract Dialectical Frameworks

(⤓.md ≡ ◇.md); γ ≜ [2026-07-17T135416.643, 2026-10-22T054143.488] ∧ |γ| = 3

Abstract Dialectical Frameworks

Origin. Brewka, Woltran (2010). Generalized argumentation. Acceptance conditions. Beyond attack. Flexible argumentative structure.

Models. Arguments with conditions. Generalized links. Acceptance functions. Boolean conditions.

Formalism.

ADF: D = ⟨S, L, C⟩ where: S: statements (arguments). L ⊆ S × S: links. C = {Cₛ}_{s∈S}: acceptance conditions.

Acceptance condition: Cₛ: 2^{par(s)} → {in, out}. par(s) = {r : (r,s) ∈ L}: parents. Boolean function. When s accepted.

Link types (derived): (r,s) supporting if Cₛ monotone increasing in r. (r,s) attacking if Cₛ monotone decreasing in r. Can be both/neither. Generalized.

Interpretation: v: S → {t, f, u}. Three-valued. t: true/in. f: false/out. u: undecided.

Operator: Γ_D(v)(s) based on Cₛ and v of parents. t if Cₛ true for all completions of v. f if Cₛ false for all completions. u otherwise.

Semantics: Admissible: v ≤_i Γ_D(v). Complete: v = Γ_D(v). Grounded: least fixpoint. Preferred, stable: adapted.

Relation to Dung: Dung AFs: special case. Cₛ = ¬r₁ ∧ ... ∧ ¬rₙ. Only attacks. ADFs generalize.

Bipolar ADFs: Links only support or attack. No mixed links. Common subclass.

Symbols.

SymbolUnicodeNameMeaning
⟨U+27E8Mathematical left angle bracketOpens the framework triple
⟩U+27E9Mathematical right angle bracketCloses the framework triple
⊆U+2286SubsetLinks lie inside S × S
×U+00D7Multiplication sign / Cartesian productLinks are ordered pairs of statements
∈U+2208MembershipStatement in S, or link in L
ₛU+209BLatin subscript small letter sIndexes an acceptance condition to its statement
₁U+2081Subscript oneIndex on the first parent
ₙU+2099Latin subscript small letter nIndex on the last parent
→U+2192Rightwards arrow / FunctionType of an acceptance condition or an interpretation
ΓU+0393Greek capital gamma / Characteristic operatorRevises an interpretation from the acceptance conditions
≤U+2264Less-than or equal to / Information orderv ≤_i w : w decides at least as much as v
¬U+00ACNegationAttack, written into an acceptance condition
∧U+2227ConjunctionJoins the conjuncts of an acceptance condition
2U+0032Digit two / Power set2^{par(s)} is the set of subsets of the parents
tU+0074Latin small letter t / TrueStatement accepted
fU+0066Latin small letter f / FalseStatement rejected
uU+0075Latin small letter u / UndecidedStatement given neither value
S—StatementsNodes of the framework
L—LinksDependencies between statements
C—Acceptance conditionsOne Boolean condition per statement
Cₛ—Acceptance conditionWhen s is accepted, as a function of its parents
par(s)—ParentsStatements that s depends on
v—InterpretationThree-valued assignment to statements
D—FrameworkThe triple ⟨S, L, C⟩
in—ValueAcceptance condition satisfied
out—ValueAcceptance condition not satisfied
ADF—Abstract dialectical frameworkThe formalism as a whole

Metatheory. Brewka and Woltran (2010) introduced ADFs and Brewka, Ellmauthaler, Strass, Wallner and Woltran (2013) revised the semantics, defining Γ_D by consensus over the two-valued completions of a three-valued interpretation. That operator is monotone with respect to the information ordering, so by Knaster–Tarski it has a least fixpoint — the grounded interpretation — while complete interpretations are its fixpoints and preferred ones the maximally informative among them, mirroring Dung exactly. Dung frameworks are the instance in which every acceptance condition is a conjunction of negated parents, and the ADF semantics restrict to Dung's on that class. The generality raises the complexity. Strass (2013) placed the semantics inside approximation fixpoint theory and gave translations showing ADFs, normal logic programs and default logic to be mutually expressible, and Strass and Wallner (2015) showed that acceptance sits one level higher in the polynomial hierarchy than for Dung frameworks — credulous acceptance Σ₂ᵖ-complete, skeptical acceptance under preferred semantics Π₃ᵖ-complete — because evaluating an acceptance condition is itself a hard step. Restricting to bipolar ADFs returns the complexity to the Dung levels.

Applies to. Argumentation. Non-monotonic reasoning. Flexible dependencies. Mixed support/attack.

Limitations. An ADF is only as compact as its acceptance conditions: a statement with n parents carries a condition over n variables, so the framework can be exponentially larger than the graph it is drawn on. Because evaluating a condition is an NP-hard step in its own right, the general acceptance problems sit a level higher in the polynomial hierarchy than the Dung problems they generalize, and the polynomial grounded semantics of Dung frameworks has no polynomial counterpart here. Link types are derived from the conditions rather than declared, so whether a link supports or attacks can depend on the rest of the condition, and a link may be neither; the bipolar restriction that recovers the Dung complexity is a restriction on admissible conditions, not a property of the graph. Statements are Boolean, leaving degrees, weights and preferences outside the ontology, and as with Dung frameworks the structure is static.

© 2026 Lingenic LLC