Augmenting Tractable Fragments of Abstract Argumentation

We present a new and compelling approach to the efficient solution of important computational problems that arise in the context of abstract argumentation. Our approach makes known algorithms defined for restricted fragments generally applicable, at a computational cost that scales with the distance from the fragment. Thus, in a certain sense, we gradually augment tractable fragments. Surprisingly, it turns out that some tractable fragments admit such an augmentation and that others do not. More specifically, we show that the problems of credulous and skeptical acceptance are fixed-parameter tractable when parameterized by the distance from the fragment of acyclic argumentation frameworks. Other tractable fragments such as the fragments of symmetrical and bipartite frameworks seem to prohibit an augmentation: the acceptance problems are already intractable for frameworks at distance 1 from the fragments. For our study we use a broad setting and consider several different semantics. For the algorithmic results we utilize recent advances in fixed-parameter tractability.

💡 Research Summary

This paper investigates how to efficiently solve the fundamental reasoning tasks of abstract argumentation—credulous acceptance (whether an argument belongs to at least one extension) and skeptical acceptance (whether it belongs to all extensions)—by exploiting the proximity of a given argumentation framework (AF) to well‑known tractable fragments. The authors introduce a distance measure, denoted dist_C(F), which is the size of a smallest “backdoor” set B of arguments whose removal places the AF F into a tractable class C (e.g., acyclic, no‑even, symmetric, bipartite). This concept mirrors the backdoor notion from propositional SAT and allows a parameterized‑complexity analysis where the parameter is the distance k = dist_C(F).

The paper’s main contributions are twofold. First, for the fragments of acyclic and no‑even AFs, the authors prove that credulous and skeptical acceptance under the five standard semantics (admissible, complete, preferred, semi‑stable, stable) become fixed‑parameter tractable (FPT) or, at worst, solvable in non‑uniform polynomial time when parameterized by the distance to the fragment. For acyclic AFs, backdoor detection reduces to the Directed Feedback Vertex Set problem, which is known to be FPT; for no‑even AFs, a simple exhaustive search yields a non‑uniform polynomial‑time algorithm, though the FPT status of backdoor detection remains open.

Second, the authors show a stark contrast for the symmetric and bipartite fragments: even a distance of one (i.e., removal of a single argument) already makes credulous and skeptical acceptance (co‑)NP‑hard. Consequently, these fragments do not admit any meaningful augmentation via the distance‑parameter approach.

Algorithmically, the approach consists of two phases. In the backdoor detection phase, the algorithm searches for a minimal set B of size at most k such that F − B belongs to the target fragment C. For acyclic, symmetric, and bipartite fragments, this detection is FPT because it corresponds respectively to Directed Feedback Vertex Set, Vertex Cover, and Odd Cycle Transversal, each of which has known FPT algorithms. For no‑even, detection is performed by brute‑force enumeration, yielding a running time of |X|^O(k) but still polynomial for fixed k.

In the backdoor evaluation phase, the algorithm enumerates all possible labelings of the backdoor arguments (each can be labeled “in”, “out”, or “undecided”). Using three propagation rules (label out if attacked by an “in”, label in if all attackers are “out”, label undecided if all attackers are either “out” or “undecided” and at least one is undecided), the partial labeling is extended to a total labeling λ*. The algorithm then constructs candidate complete extensions compatible with λ* by solving the acceptance problem on the reduced, tractable sub‑framework F − def(λ*). Because the reduced framework belongs to a tractable fragment, existing polynomial‑time solvers can be applied directly. The overall runtime is bounded by 3^k · poly(|X|), which together with the FPT backdoor detection yields an overall FPT algorithm for acyclic AFs and a non‑uniform polynomial‑time algorithm for no‑even AFs.

The authors also provide a thorough complexity landscape. Table 1 (referenced in the paper) lists the known complexities of credulous and skeptical acceptance for each semantics across the four fragments. The positive results demonstrate that the distance‑parameter strictly refines the known tractability borders: many AFs that are hard for general frameworks become efficiently solvable when they are “close” to an acyclic or no‑even fragment. The negative results for symmetric and bipartite fragments underline that not all tractable fragments can be augmented; their structural properties are such that even a minimal deviation destroys tractability.

In the discussion, the paper relates its distance measure to earlier parameters such as cycle‑rank and even‑cycle‑rank, showing that the new parameter can be incomparable and sometimes strictly more powerful. It also suggests future research directions, including (i) investigating whether backdoor detection for no‑even AFs can be made FPT, (ii) exploring other tractable fragments that might admit augmentation, and (iii) combining the distance‑parameter with other structural parameters (e.g., treewidth, clique‑width) to obtain even finer-grained algorithms.

Overall, the work makes a significant methodological contribution by importing the backdoor paradigm into abstract argumentation, providing both positive algorithmic results for certain fragments and clear limitations for others, and opening a promising line of research in parameterized reasoning for argumentation frameworks.