6.12.2024

Our work on "Dynamic Programming Algorithms for Probabilistic Bipolar Argumentation Frameworks" was accepted to SAC (Symposium On Applied Computing) in the Track on Knowledge Representation and Reasoning (KRR). We extend our previous works on probabilistic argumentation frameworks to the case of having supporting relations, instead of only attack relations, between arguments. We develop dynamic programming algorithms utilizing tree decompositions for challenging computational tasks that arise in such frameworks.

29.10.2024

Our group organizes the next edition of the International Competitions on Computational Models of Argumentation (ICCMA) in 2025. Call for solvers and benchmarks can be found here. The aim of the competition is to further development of solvers for (hard) reasoning tasks in computational argumentation.

5.9.2024

Our work "A Semantical Approach to Abstraction in Answer Set Programming and Assumption-based Argumentation" was accepted to this year's edition of LPNMR. Following the theme of the conference, non-monotonic reasoning and logic programming, we investigate forms of abstractions on logic programs under the answer set semantics and on the related assumption-based argumentation formalism.

11.07.2024

In this year's edition of the International Conference on Principles of Knowledge Representation and Reasoning (KR'24), three of our works got accepted:

• Complexity Results and Algorithms for Preferential Argumentative Reasoning in ASPIC+
• Abstraction in Assumption-based Argumentation
• Advancing Algorithmic Approaches to Probabilistic Argumentation under the Constellation Approach

In the first work, we extend computational approaches to the prominent structured argumentation formalism of ASPIC+, in particular incorporating also preferential argumentative reasoning. In the second work we deal with abstracting information to attain a high-level view aimed at supporting explainability. In the third work we studied alorithmic approaches to the highly complex field of probabilistic reasoning in formal argumentation.

9.7.2024

Wallner gave an invited talk at the Workshop on Recent Trends in Formal Argumentation at TU Wien on the topic "Instantiations and Computational Aspects of Assumption-based Argumentation". This workshop discussed recent advances in the field of computational argumentation. 

14.06.2024

Three of our works are accepted to COMMA 2024, a conference dedicated to computational argumentation. In the first work "Value-based Reasoning in ASPIC+" we investigate how value-based reasoning can be integrated in the argumentation formalism ASPIC+. In "On Computing Admissibility in ABA" we look at algorithmic approaches that compute compact representations of admissible sets of arguments in non-flat assumption-based argumentation. In the third work, "Complexity of Semi-Stable Semantics in Abstract Dialectical Frameworks", we extend the current state-of-the-art complexity analysis on abstract dialectical frameworks, by considering the recently introduced semi-stable semantics.  

2.5.2024

Our works on "Ranking Transition-based Medical Recommendations using Assumption-based Argumentation" and "Instantiations and Computational Aspects of Non-Flat Assumption-based Argumentation" have been accepted to RATIO and IJCAI 2024, respectively. RATIO is a new conference on advances in argumentation machines and IJCAI is a premier conference in Artificial Intelligence. 

In the first work we look at using a ranking-based semantics for assumption-based argumentation in a medical scenario. In the second work we look at issues arising when creating argument structures from a possibly conflicting knowledge base. We consider algorithmic approaches to address the issues. 

22.02.2024

We are co-organizing this year's edition of the Doctoral Consortium (DC) of the prominent International Conference on Principles of Knowledge Representation and Reasoning (KR'24). The DC is a student mentoring program bringing together PhD students and senior researchers from the area of KR. 

The call for applications is available at the KR'24 webpage. 

16.02.2024

Our work on "Abstracting Assumptions in Structured Argumentation" will be presented at this year's edition of AAMAS, and an extended abstract of our work will be published in the proceedings of this conference. In this work we look at possibilities of simplifying - abstracting - argumentation scenarios in the prominent structured argumentation formalisms of assumption-based argumentation (ABA).

1.9.2023

Our work on "Ranking-based Semantics for Assumption-based Argumentation" was accepted to the 9th Workshop on Formal and Cognitive Reasoning (FCR'23), a workshop at the German KI conference in 2023. 

In this work we look at ways of defining ranking-based semantics on the knowledge representation formalism assumption-based argumentation (ABA). Our ranking-based semantics relate assumptions, and we present ways of obtaining such semantics from existing ranking-based semantics on abstract argumentation. We define principles, i.e., properties of interest, ranking-based semantics on ABA may satisfy. 

15.07.2023

Our work on "Reasoning in Assumption-based Argumentation using Tree-decompositions" was accepted to JELIA, the european conference on logics in Artificial Intelligence. 

In this work we look at methods that decompose instances of complex reasoning tasks arising in formal argumentation. We decompose the structures into tree-decompositions and apply a recently developed declarative framework for specifying dynamic programming algorithms on such tree-decompositions.

23.5.2023

Our works on "Argumentation Frameworks induced by Assumption-based Argumentation: Relating Size and Complexity" and "Argumentative Reasoning in ASPIC+ under Incomplete Information" were accepted to KR'23, a leading conference on knowledge representation and reasoning.

In the first paper we study reasoning in the formal approach of assumption-based argumentation (ABA). After observing that argumentative reasoning in this instance has two sources of complexity, namely construction of arguments and complexity of reasoning based on such arguments, we identify and study fragments of ABA that can be utilized to confine complexity to either source. In an experimental evaluation one of the fragments, that of of so-called atomic ABAs, shows promise for utilization in systems that solve ABA reasoning.

In the second work we look at ASPIC+, an argumentation formalisms related to ABA, and investigate reasoning under incomplete information. In particular, we look at notions of stability and relevance, with stability intuitively signaling when a justification status of a conclusion cannot be altered with new information and relevance points towards useful information to achieve such a stable status. We study the complexity of these problems, provide algorithmic approaches, and an empirical evaluation of our algorithms.

15.12.2022

In the regular seminar (Oberseminar) of the Artificial Intelligence Group (AIG) at FernUnversität Hagen, Johannes Wallner gave an invited talk on structured argumentation titled "Computation in Structured Argumentation: to Instantiate or not to Instantiate?" on December 15th. In this talk we outlined advantages and disadvantages of argument instantiation, i.e., of the process of generating explicit arguments from given knowledge bases.

We are offering a new course "Logic-based Knowledge Representation" for the first time this winter term. 

In this course we look at prominent logic-based approaches in knowledge representation and reasoning, with a focus on modeling, expressivity, and complexity of reasoning in these approaches. We look at (fragments of) classical logic, non-classical logic, and non-monotonic reasoning.

This course is part of several elective catalogues.

Registration and information can be found in TU Graz Online.

Our works "Strongly Accepting Subframeworks: Connecting Abstract and Structured Argumentation" and "Algorithms for Reasoning in a Default Logic Instantiation of Assumption-Based Argumentation" were accepted to COMMA'22, a conference specialized to computational argumentation in AI. Moreover, a system demo associated to our recent paper to be presented at LPNMR'22 was accepted.

In the first work we look at possiblities of extending and generalizing recent research on strongly accepting subframeworks to more general structured argumentation frameworks, in particular assumption-based argumentation. We investigate conditions under which a generalization is possible.

In the second work we consider algorithmic approaches to a default logic instantiation of assumption-based argumentation, paving the way in particular for more expressive frameworks than were focused on before in terms of system implementations.  

Our work on "Representing Abstract Dialectical Frameworks with Binary Decision Diagrams" was accepted to LPNMR'22 and our work on "Argumentation Frameworks induced by Assumption-based Argumentation: Relating Size and Complexity" was accepted to the NMR'22 workshop. 

In the first work we investigate possibilities to utilize binary decision diagrams (BDDs) to reasoning in abstract dialectical frameworks (ADFs), a formal approach to argumentative reasoning. We show that complexity of reasoning can be milder, but not necessarily in polynomial-time, when using BDDs for ADFs. Based on the computational advantage of BDDs we apply a recently proposed framework of faceted navigation to develop heuristics for reasoning in ADFs. Preliminary experiments show potential advantage of this approach.

In the second work we look at the formal argument generation process that is prominent in computational argumentation. We show complexity results of argument generation, number of resulting argument structures, and that there is an apparent trade-off between number and complexity of resulting argument structures.

We will give a talk about our recent work on "Existential Abstraction on Argumentation Frameworks via Clustering" in the upcoming NAVAS workshop 2022 in Vienna. This workshop is organized by research groups at TU Wien and TU Dresden. 

Our paper "Computing Stable Conclusions under the Weakest-Link Principle in the ASPIC+ Argumentation Formalism" was accepted to KR 2022. In this work, we look at argumentative reasoning including preferential information in the formalisms called ASPIC+. We derive novel complexity results, algorithms, and an empirically evaluated prototype for this formalism.

Our paper "Advanced Algorithms for Abstract Dialectical Frameworks based on Complexity Analysis of Subclasses and SAT Solving" was accepted to the Artificial Intelligence journal. In this article we investigate computational complexity, develop algorithms, and present a prototypical implementation, for Abstract Dialectical Frameworks (ADFs), which are a formalization of arguments and their relationships within the area of Computational Argumentation. We study several fragments of the formal language of ADFs and show complexity results for these fragments. Based on these insights, we develop algorithmic approaches to solve reasoning tasks on ADFs, and present an empirical evaluation of resulting prototypical implementations. More information in pure. 

We have two openings for PhD students on the topic of knowledge representation and reasoning. See Jobs.

Our paper "An Axiomatic Approach to Revising Preferences" was accepted to AAAI 2022. In this paper we study a general approach to revision of preferences, along the lines of the well-known AGM approach to (belief) revision in AI, which bases revision operations on rationality postulates (formal properties stating desirable or interesting behavior). More information in pure.

In the paper "Harnessing Incremental ASP Solving for Reasoning in Assumption-Based Argumentation" we develop novel algorithms based on counterexample-guided abstraction refinement and incremental answer set programming (ASP) to address challenging computational problems in assumption-based argumentation. More information in pure.

In the paper "Existential Abstraction on Argumentation Frameworks via Clustering" we investigate abstraction techniques for simplification of argumentation scenarios, with the inherent aim to balance simplification and spuriousness (wrong conclusions drawn solely due to the abstraction), which is an inherent issue in abstraction. More information in pure.

The article "Declarative Algorithms and Complexity Results for Assumption-Based Argumentation" establishes novel complexity-theoretic and algorithmic advances for the formalism of assumption-based argumentation within the wider field of computational argumentation in AI. More information in pure.