Publications

2025

We consider abstract argumentation and explanations for the acceptance of arguments. Based on the notion of serialisability, we introduce sequence explanations as a procedural form of explanation for the acceptance of some argument. Intuitively, these explanations represent the process of accepting (and rejecting) arguments in order to conclude the acceptance of a certain argument. We define several variants of sequence explanations and examine them in detail. In particular, we also incorporate counterarguments into the explanations to make them dialectical. Finally, we relate our explanations to other approaches from the literature via a principle-based analysis. 
@inproceedings{conf/kr/BengelT25,
    title     = {{Sequence Explanations for Acceptance in Abstract Argumentation}},
    author    = {Bengel, Lars and Thimm, Matthias},
    booktitle = {Proceedings of the 22nd International Conference on Principles of Knowledge Representation and Reasoning, ({KR} 2025)},
    pages     = {121--131},
    year      = {2025},
    month     = {11},
    doi       = {10.24963/kr.2025/12},
    url       = {https://doi.org/10.24963/kr.2025/12},
  }

We consider abstract argumentation frameworks and, in particular, the problem of skeptical reasoning wrt. preferred semantics, i.e., deciding whether a given argument is contained in every preferred extension of the argumentation framework. We introduce a novel SAT-based approach, building on recent results from the literature, that searches through complete extensions to efficiently decide this problem. It also employs effective simplification procedures to shorten computation times. As our experimental evaluation shows, our algorithm significantly outperforms state-of-the-art approaches. 
@inproceedings{conf/kr/BengelST25,
    title     = {{A Reduct-based Approach to Skeptical Preferred Reasoning in Abstract Argumentation}},
    author    = {Bengel, Lars and Sander, Julian and Thimm, Matthias},
    booktitle = {Proceedings of the 22nd International Conference on Principles of Knowledge Representation and Reasoning, ({KR} 2025)},
    pages     = {132--136},
    year      = {2025},
    month     = {11},
    doi       = {10.24963/kr.2025/13},
    url       = {https://doi.org/10.24963/kr.2025/13},
  }

We consider abstract argumentation frameworks and, in particular, the problem of skeptical reasoning wrt. preferred semantics, i. e., deciding whether a given argument is contained in every preferred extension of the argumentation framework. We introduce a novel Sat-based approach, building on recent results from the literature, that searches through complete extensions to efficiently decide this problem. It also employs effective simplification procedures to shorten computation times. As our experimental evaluation shows, our algorithm significantly outperforms current state-of-the-art approaches in most instances. 
@inproceedings{DBLP:conf/nmr/BengelST25,
  author       = {Lars Bengel and
                  Julian Sander and
                  Matthias Thimm },
  editor       = {Anna Rapberger and
                  Sebastian Rudolph},
  title        = {An Extension-Based Argument-Ranking Semantics: Social Rankings in
                  Abstract Argumentation},
  booktitle    = {Proceedings of the 23rd International Workshop on Nonmonotonic Reasoning
                  {(NMR} 2025) co-located with 22nd International Conference on Principles
                  of Knowledge Representation and Reasoning {(KR} 2025), Melbourne, Australia,
                  November 11-13, 2025},
  series       = {{CEUR} Workshop Proceedings},
  volume       = {4071},
  pages        = {3--16},
  publisher    = {CEUR-WS.org},
  year         = {2025},
}

In this paper, we introduce a new family of argument-ranking semantics which can be seen as a refinement of the classification of arguments into skeptically accepted, credulously accepted and rejected. To this end we use so-called social ranking functions which have been developed recently to rank individuals based on their performance in groups. We provide necessary and sufficient conditions for a social ranking function to give rise to an argument-ranking semantics satisfying the desired refinement property.

We introduce initial models for abstract dialectical frameworks (ADFs) as a notion of minimal justifiable valuations and based on that, generalise the concept of serialisability of argumentation semantics to ADFs. In particular, we show that the characteristic operator-based semantics for ADFs can be characterised through serialisation sequences, which are, essentially, decompositions of a model into a series of initial models, representing a more fine-grained view into why a model is acceptable wrt. the semantics. We also analyse the computational complexity of tasks related to initial models. 
@inproceedings{DBLP:conf/ijcai/BengelT25,
  author       = {Lars Bengel and
                  Matthias Thimm},
  title        = {Initial Models and Serialisability in Abstract Dialectical Frameworks},
  booktitle    = {Proceedings of the Thirty-Fourth International Joint Conference on
                  Artificial Intelligence, {IJCAI} 2025, Montreal, Canada, August 16-22,
                  2025},
  pages        = {4365--4373},
  publisher    = {ijcai.org},
  year         = {2025},
  url          = {https://doi.org/10.24963/ijcai.2025/486},
  doi          = {10.24963/IJCAI.2025/486},
  timestamp    = {Wed, 24 Sep 2025 17:45:28 +0200},
  biburl       = {https://dblp.org/rec/conf/ijcai/BengelT25.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}

We investigate the computational problem of determining the set of acceptable arguments in abstract argumentation wrt. credulous and skeptical reasoning under grounded, complete, stable, and preferred semantics. In particular, we investigate the computational complexity of that problem and its verification variant, and develop several algorithms for all problem variants, including two baseline approaches based on iterative acceptability queries and extension enumeration, and some optimised versions. We experimentally compare the runtime performance of these algorithms: our results show that our newly optimised algorithms significantly outperform the baseline algorithms in most cases. 
@article{DBLP:journals/ijar/BengelTCV25,
  author       = {Lars Bengel and
                  Matthias Thimm and
                  Federico Cerutti and
                  Mauro Vallati},
  title        = {Algorithms for computing the set of acceptable arguments},
  journal      = {Int. J. Approx. Reason.},
  volume       = {185},
  pages        = {109478},
  year         = {2025},
  url          = {https://doi.org/10.1016/j.ijar.2025.109478},
  doi          = {10.1016/J.IJAR.2025.109478},
  timestamp    = {Fri, 04 Jul 2025 22:12:59 +0200},
  biburl       = {https://dblp.org/rec/journals/ijar/BengelTCV25.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}

In this paper, we introduce a new family of argument-ranking semantics which can be seen as a refinement of the classification of arguments into skeptically accepted, credulously accepted and rejected. To this end we use so-called social ranking functions which have been developed recently to rank individuals based on their performance in groups. We provide necessary and sufficient conditions for a social ranking function to give rise to an argument-ranking semantics satisfying the desired refinement property. 
@inproceedings{DBLP:conf/aaai/BengelB0S25,
  author       = {Lars Bengel and
                  Giovanni Buraglio and
                  Jan Maly and
                  Kenneth Skiba},
  editor       = {Toby Walsh and
                  Julie Shah and
                  Zico Kolter},
  title        = {An Extension-Based Argument-Ranking Semantics: Social Rankings in
                  Abstract Argumentation},
  booktitle    = {AAAI-25, Sponsored by the Association for the Advancement of Artificial
                  Intelligence, February 25 - March 4, 2025, Philadelphia, PA, {USA}},
  pages        = {14790--14797},
  publisher    = {{AAAI} Press},
  year         = {2025},
  url          = {https://doi.org/10.1609/aaai.v39i14.33621},
  doi          = {10.1609/AAAI.V39I14.33621},
  timestamp    = {Fri, 04 Jul 2025 22:04:24 +0200},
  biburl       = {https://dblp.org/rec/conf/aaai/BengelB0S25.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}

2024

We consider abstract argumentation frameworks (AFs) and their semantics. The standard approach in the literature are the extension-based semantics, under which sets of jointly acceptable arguments are computed. While there exist alternatives, like labeling-based semantics, none of these approaches are able to model the procedural aspect of argumentation on the semantical layer. In this work, we present the sequence-based semantics, that compute sequences of minimally acceptable sets (called serialisation sequences), as a step towards a dialectical form of argumentation semantics. 
@incollection{ohaai/Bengel24,
  author        = {Lars Bengel},
  title         = {Towards a Dialectical Approach to Abstract Argumentation Semantics},
  booktitle     = {Online Handbook of Argumentation in AI},
  volume        = {5},
  year          = {2024},
  publisher     = {preprint}
}

In this paper, we introduce a new family of argument-ranking semantics which can be seen as a refinement of the classification of arguments into skeptically accepted, credulously accepted and rejected. To this end we use so-called social ranking functions which have been developed recently to rank individuals based on their performance in groups. We provide necessary and sufficient conditions for a social ranking function to give rise to an argument-ranking semantics satisfying the desired refinement property. Moreover, we analyse the properties of the argument-ranking semantics induced by the most prominent social ranking function that satisfies all of these conditions by investigating the satisfaction of principles known from the argument-ranking literature. 
@inproceedings{DBLP:conf/nmr/BengelB0S24,
  author       = {Lars Bengel and
                  Giovanni Buraglio and
                  Jan Maly and
                  Kenneth Skiba},
  editor       = {Nina Gierasimczuk and
                  Jesse Heyninck},
  title        = {An Extension-Based Argument-Ranking Semantics: Social Rankings in
                  Abstract Argumentation},
  booktitle    = {Proceedings of the 22nd International Workshop on Nonmonotonic Reasoning
                  {(NMR} 2024) co-located with 21st International Conference on Principles
                  of Knowledge Representation and Reasoning {(KR} 2024), Hanoi, Vietnam,
                  November 2-4, 2024},
  series       = {{CEUR} Workshop Proceedings},
  volume       = {3835},
  pages        = {61--71},
  publisher    = {CEUR-WS.org},
  year         = {2024},
  url          = {https://ceur-ws.org/Vol-3835/paper7.pdf},
  timestamp    = {Wed, 04 Dec 2024 17:11:34 +0100},
  biburl       = {https://dblp.org/rec/conf/nmr/BengelB0S24.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}

We introduce the notion of serialisation equivalence, which provides a notion of equivalence that takes the underlying dialectical structure of extensions in an argumentation framework into account. Under this notion, two argumentation frameworks are considered equivalent if they possess not only the same extensions wrt. some semantics but also the same serialisation sequences. A serialisation sequence is a decomposition of an extension into a series of minimal acceptable sets and essentially offers insight into the order in which arguments need to brought forward to resolve the conflicts and to justify a particular position in the argumentation framework. We analyse serialisation equivalence in detail and show that it is generally more strict than standard equivalence and less strict than strong equivalence. Furthermore, we provide a full analysis of the computational complexity of deciding serialisation equivalence. 
@inproceedings{DBLP:conf/ecai/BengelST24,
  author       = {Lars Bengel and
                  Julian Sander and
                  Matthias Thimm},
  editor       = {Ulle Endriss and
                  Francisco S. Melo and
                  Kerstin Bach and
                  Alberto Jos{\'{e}} Bugar{\'{\i}}n Diz and
                  Jose Maria Alonso{-}Moral and
                  Sen{\'{e}}n Barro and
                  Fredrik Heintz},
  title        = {Characterising Serialisation Equivalence for Abstract Argumentation},
  booktitle    = {{ECAI} 2024 - 27th European Conference on Artificial Intelligence,
                  19-24 October 2024, Santiago de Compostela, Spain},
  series       = {Frontiers in Artificial Intelligence and Applications},
  volume       = {392},
  pages        = {3340--3347},
  publisher    = {{IOS} Press},
  year         = {2024},
  url          = {https://doi.org/10.3233/FAIA240883},
  doi          = {10.3233/FAIA240883},
  timestamp    = {Mon, 03 Mar 2025 21:02:32 +0100},
  biburl       = {https://dblp.org/rec/conf/ecai/BengelST24.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}

We consider the problem of learning argumentation frameworks from a given set of labelings such that every input is a σ-labeling of these argumentation frameworks. Our new algorithm takes labelings and computes attack constraints for each argument that represent the restrictions on argumentation frameworks that are consistent with the input labelings. Having constraints on the level of arguments allows for a very effective parallelization of all computations. An important element of this approach is maintaining a representation of all argumentation frameworks that satisfy the input labelings instead of simply finding any suitable argumentation framework. This is especially important, for example, if we receive additional labelings at a later time and want to refine our result without having to start all over again. The developed algorithm is compared to previous works and an evaluation of its performance has been conducted. 
@article{DBLP:journals/argcom/BengelTR24,
  author       = {Lars Bengel and
                  Matthias Thimm and
                  Tjitze Rienstra},
  title        = {Learning argumentation frameworks from labelings},
  journal      = {Argument Comput.},
  volume       = {15},
  number       = {2},
  pages        = {121--159},
  year         = {2024},
  url          = {https://doi.org/10.3233/AAC-220018},
  doi          = {10.3233/AAC-220018},
  timestamp    = {Wed, 24 Jul 2024 21:43:38 +0200},
  biburl       = {https://dblp.org/rec/journals/argcom/BengelTR24.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}

We introduce an argumentation-based approach for conducting probabilistic causal reasoning. For that, we consider Pearl’s causal models where causal relations are modelled via structural equations and a probability distribution over background atoms. The probability that some causal statement holds is then computed by constructing a probabilistic argumentation framework and determining its extensions. This framework can then be used to generate argumentative explanations for the (non-)acceptance of the causal statement. Furthermore, we present an argumentation-based version of the twin network method for dealing with counterfactuals. Finally, we show that our approach yields the same results for causal and counterfactual queries as Pearl’s model. 
@inproceedings{DBLP:conf/ratio/BengelBRT24,
  author       = {Lars Bengel and
                  Lydia Bl{\"{u}}mel and
                  Tjitze Rienstra and
                  Matthias Thimm},
  editor       = {Philipp Cimiano and
                  Anette Frank and
                  Michael Kohlhase and
                  Benno Stein},
  title        = {Argumentation-Based Probabilistic Causal Reasoning},
  booktitle    = {Robust Argumentation Machines - First International Conference, {RATIO}
                  2024, Bielefeld, Germany, June 5-7, 2024, Proceedings},
  series       = {Lecture Notes in Computer Science},
  volume       = {14638},
  pages        = {221--236},
  publisher    = {Springer},
  year         = {2024},
  url          = {https://doi.org/10.1007/978-3-031-63536-6\_13},
  doi          = {10.1007/978-3-031-63536-6\_13},
  timestamp    = {Sun, 19 Jan 2025 13:19:40 +0100},
  biburl       = {https://dblp.org/rec/conf/ratio/BengelBRT24.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}

2023

We consider the notion of serialisability, a non-deterministic construction scheme for admissibility-based semantics in abstract argumentation frameworks. It provides a decomposition of the extension into a sequence of minimal acceptable sets that offers a good foundation for generating argumentative explanations. In this paper, we discuss the goal of generalising serialisability to other types of semantics, to more expressive approaches in formal argumentation, for instance abstract dialectical frameworks, and to other related formalisms. 
@incollection{ohaai/Bengel23,
  author        = {Lars Bengel},
  title         = {Towards Generalising Serialisability to other Argumentation Formalisms},
  editor        = {Elfia Bezou-Vrakatseli and
                    Federico Castagna and
                    Isabelle Kuhlmann and
                    Jack Mumford and
                    Stefan Sarkadi and
                    Daphne Odekerken and
                    Maddie Waller and
                    Andreas Xydis},
  booktitle     = {Online Handbook of Argumentation in AI},
  volume        = {4},
  year          = {2023},
  pages         = {7--12},
  url           = {https://doi.org/10.48550/arXiv.2401.09444},
  doi           = {10.48550/arXiv.2401.09444},
  publisher     = {arXiv}
}

We consider the recently proposed notion of serialisability of semantics for abstract argumentation frameworks. This notion describes a method for the serialised non-deterministic construction of extensions through iterative addition of non-empty minimal admissible sets. Depending on the semantics, the task of enumerating all extensions for an argumentation framework can be computationally complex. Serialisability provides a natural way of parallelising the construction of extensions for most admissible-based semantics. In this work, we investigate the feasibility of using the serialisable construction scheme for a more efficient enumeration of extensions on the example of the recently introduced unchallenged semantics and provide an experimental evaluation. 
@inproceedings{DBLP:conf/kr/BengelT23,
  author       = {Lars Bengel and
                  Matthias Thimm},
  editor       = {Pierre Marquis and
                  Tran Cao Son and
                  Gabriele Kern{-}Isberner},
  title        = {Towards Parallelising Extension Construction for Serialisable Semantics
                  in Abstract Argumentation},
  booktitle    = {Proceedings of the 20th International Conference on Principles of
                  Knowledge Representation and Reasoning, {KR} 2023, Rhodes, Greece,
                  September 2-8, 2023},
  pages        = {732--736},
  year         = {2023},
  url          = {https://doi.org/10.24963/kr.2023/72},
  doi          = {10.24963/KR.2023/72},
  timestamp    = {Tue, 05 Sep 2023 14:50:53 +0200},
  biburl       = {https://dblp.org/rec/conf/kr/BengelT23.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}

2022

We consider the recently proposed notion of serialisability of semantics for abstract argumentation frameworks. This notion describes a method for the serialised non-deterministic construction of extensions through iterative addition of non-empty minimal admissible sets. This paper provides an overview on serialisability and proposes some promising applications of this concept in the context of explainable AI. In particular, we outline how serialisability could be employed to provide more structured explanations. We also discuss how serialisability could be utilised in discussion games. 
@incollection{ohaai/Bengel22,
  author        = {Lars Bengel},
  title         = {On Serialisability for Argumentative Explanations},
  editor        = {Federico Castagna and
                    Jack Mumford and
                    Stefan Sarkadi and
                    Andreas Xydis},
  booktitle     = {Online Handbook of Argumentation in AI},
  volume        = {3},
  year          = {2022},
  pages         = {2--6},
  url           = {https://doi.org/10.48550/arXiv.2212.07996},
  doi           = {10.48550/ARXIV.2212.07996},
  publisher     = {arXiv}
}
                                        

@inproceedings{conf/fatil/Bengel22,
  author = 	{Lars Bengel},
  title  =  {Towards Learning Argumentation Frameworks from Labelings},
  pages  =  {67 -- 69},
  editor = 	{Thimm, Matthias
		and Landes, J{\"u}rgen
		and Skiba, Kenneth},
  booktitle = 	{Proceedings of the First International Conference on Foundations, Applications, and Theory of Inductive Logic, ({FATIL} 2022), Munich, Germany, October 12 - 14, 2022},
  year = 	{2022},
  doi = 	{10.18445/20220817-161753-0},
  url = 	{https://doi.org/10.18445/20220817-161753-0},
}

We investigate the recently proposed notion of serialisability of semantics for abstract argumentation frameworks. This notion describes semantics where the construction of extensions can be serialised through iterative addition of minimal non-empty admissible sets. We investigate general relationships between serialisability and other principles from the literature. We also investigate the novel unchallenged semantics as a new instance of a serialisable semantics and, in particular, analyse it in terms of satisfied principles and computational complexity. 
@inproceedings{DBLP:conf/comma/BengelT22,
  author       = {Lars Bengel and
                  Matthias Thimm},
  editor       = {Francesca Toni and
                  Sylwia Polberg and
                  Richard Booth and
                  Martin Caminada and
                  Hiroyuki Kido},
  title        = {Serialisable Semantics for Abstract Argumentation},
  booktitle    = {Computational Models of Argument - Proceedings of {COMMA} 2022, Cardiff,
                  Wales, UK, 14-16 September 2022},
  series       = {Frontiers in Artificial Intelligence and Applications},
  volume       = {353},
  pages        = {80--91},
  publisher    = {{IOS} Press},
  year         = {2022},
  url          = {https://doi.org/10.3233/FAIA220143},
  doi          = {10.3233/FAIA220143},
  timestamp    = {Mon, 31 Oct 2022 16:49:16 +0100},
  biburl       = {https://dblp.org/rec/conf/comma/BengelT22.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}

In this paper we present a model for argumentative causal and counterfactual reasoning in a logical setting. Causal knowledge is represented in this system using Pearl’s causal model of a set of structural equations and a set of assumptions expressed in propositional logic. Queries concerning observations or actions can be answered by constructing an argumentation framework and determining its extensions. For counterfactual queries we propose an argumentation-based implementation of the twin network method and analyse its expressiveness. 
@inproceedings{DBLP:conf/comma/BengelBRT22,
  author       = {Lars Bengel and
                  Lydia Bl{\"{u}}mel and
                  Tjitze Rienstra and
                  Matthias Thimm},
  editor       = {Kristijonas Cyras and
                  Timotheus Kampik and
                  Oana Cocarascu and
                  Antonio Rago},
  title        = {Argumentation-based Causal and Counterfactual Reasoning},
  booktitle    = {1st International Workshop on Argumentation for eXplainable {AI} co-located
                  with 9th International Conference on Computational Models of Argument
                  {(COMMA} 2022), Cardiff, Wales, September 12, 2022},
  series       = {{CEUR} Workshop Proceedings},
  volume       = {3209},
  publisher    = {CEUR-WS.org},
  year         = {2022},
  url          = {https://ceur-ws.org/Vol-3209/7343.pdf},
  timestamp    = {Fri, 10 Mar 2023 16:22:13 +0100},
  biburl       = {https://dblp.org/rec/conf/comma/BengelBRT22.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}

2021

In abstract argumentation, the admissible semantics can be said to distinguish the preferred semantics in the sense that argumentation frameworks with the same admissible extensions also have the same preferred extensions. In this paper we present an exhaustive study of such distinguishability relationships, including those between sets of semantics. We further examine restricted classes of argumentation frameworks, such as self-attack-free and acyclic frameworks. We discuss the relevance of our results in the context of the argumentation framework elicitation problem. 
@inproceedings{DBLP:conf/kr/KuhlmannRBST21,
  author       = {Isabelle Kuhlmann and
                  Tjitze Rienstra and
                  Lars Bengel and
                  Kenneth Skiba and
                  Matthias Thimm},
  editor       = {Meghyn Bienvenu and
                  Gerhard Lakemeyer and
                  Esra Erdem},
  title        = {Distinguishability in Abstract Argumentation},
  booktitle    = {Proceedings of the 18th International Conference on Principles of
                  Knowledge Representation and Reasoning, {KR} 2021, Online event, November
                  3-12, 2021},
  pages        = {686--690},
  year         = {2021},
  url          = {https://doi.org/10.24963/kr.2021/70},
  doi          = {10.24963/KR.2021/70},
  timestamp    = {Wed, 03 Nov 2021 12:47:32 +0100},
  biburl       = {https://dblp.org/rec/conf/kr/KuhlmannRBST21.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}