Belangrijkste concepten
Addressing the computational challenges of non-flat assumption-based argumentation through theoretical analysis and the development of efficient algorithmic approaches.
Samenvatting
The paper investigates computational aspects of non-flat assumption-based argumentation (ABA), which is a more general case than the commonly studied flat ABA. The authors make the following key contributions:
They establish a theoretical result based on compilability theory, showing that it is impossible to instantiate a non-flat ABA framework into a polynomial-sized argumentation framework or bipolar argumentation framework while preserving the semantics.
To address the exponential size of the instantiated frameworks, the authors identify three notions of redundancy in argument generation - derivation redundancy, expendable arguments, and assumption redundancy. They show that eliminating these redundancies preserves the semantics.
The authors also identify two fragments of non-flat ABA, called atomic and additive ABA, which admit a polynomial-sized instantiation and thus lower computational complexity.
The paper proposes two algorithmic approaches for reasoning in non-flat ABA:
ABABAF: An approach that efficiently instantiates a bipolar argumentation framework using the redundancy notions and then performs SAT-based reasoning on the constructed framework.
ABASP: A direct approach that performs reasoning on the non-flat ABA framework using iterative ASP calls, without constructing the arguments.
The empirical evaluation shows that the ABABAF approach outperforms the ABASP approach on many instances, especially for the problems on the second level of the polynomial hierarchy, in contrast to the current dominance of non-instantiation approaches for other structured argumentation formalisms.