Dynamics of Argumentation, Rules and Conditionals
by: Tjitze Rienstra, University of Luxembourg
The two-day workshop on the Dynamics of Argumentation, Rules and Conditionals (DARC),(http://icr.uni.lu/darc ) was held at the University of Luxembourg on April 2-3. It was organized by Richard Booth, Emil Weydert and Tjitze Rienstra of the Individual and Collective Reasoning (ICR) research group (http://icr.uni.lu/ ), and was associated with the DYNARG project (http://icr.uni.lu/dynarg/ ). The event was supported by the European Network for Social intelligence (SINTELNET) and BNVKI. Its aim was to bring together people interested in the dynamics of formal argumentation in a broad sense, and to exchange ideas, techniques and results.
The first invited speaker, Joao Leite, opened the workshop with an overview of the problem of updating logic programs. Applying the Katzuno and Mendelzon postulates for knowledge base update to logic program update is problematic, so a different set of postulates is required. Several of these postulates were presented and discussed with respect to a number of approaches to updating logic programs.
Ringo Baumann addressed the problem of revising a Dung style argumentation framework, with the aim of enforcing a set E of arguments to be accepted. Pierre Bisquert looked at the duality between the operations of removing and adding arguments in an abstract argumentation framework. Dov Gabbay addressed the problem of evaluating odd cycles in argumentation frameworks. After briefly introducing two known approaches to this problem, namely the CF2 and the equational approach, he presented a novel interpretation of odd cycles, which he calls the “Shkop approach”.
Tjitze Rienstra described a generalization of the concept of an acceptance function. In addition to a framework, it takes as input a set of labelings. This generalization allows evaluation of argumentation frameworks given a set of constraints, which opens up the possibility of applying Dung-style argumentation in a more dynamic way. In his talk, Martin Caminada argued for natural interpretations of dialog games for the various argumentation semantics. He presented an example demonstrating that the dialog game for preferred semantics corresponds to socratic discussion.
Johannes Wallner dealt with the problem of updating Abstract Dialectical Frameworks in the light of new information. Abstract Dialectical Frameworks generalize Dung-style argumentation frameworks, allowing relations between arguments such as support and collective attack. Two change operations, namely addition and removal of an argument, were discussed.
In the second invited talk, Henry Prakken first introduced the ASPIC framework, which combines assumption-based and defeasible rule based reasoning, and is an instantiation of Dung-style abstract argumentation. Then he pointed out the problem of representing preferences on the abstract level. When preferences are represented on the abstract level, inconsistencies arise at the instantiated level. Instead, preferences should be represented on the instantiated level, i.e. over the defeasible rules. This argument applies more broadly. That is, when information is represented on the abstract level, care must be taken that this does not lead to problems on the instantiated level.
In his invited presentation, Alexander Bochman gave an overview of the early motivation of AGM revision, and pointed out that rules (conditionals) constitute the core of epistemic states, which is the concept that underlies the AGM model. He discussed the possibility of representing evolutionary rule change via expansions of (prioritized) argumentation systems.
Jan Sefranek considered the problem of updating assumption based argumentation frameworks. The presentation focused on the problem of removing rules that cause conflict, and properties that should be satisfied by the update operation. Martin Balaz presented a number of principles for MDLP, such as causal rejection, inertia, reinstatement and immunity to tautological and cyclic updates. Several existing semantics for MDLP were characterized in terms of those principles, and the effect of adding explicit negation was discussed.
Brian Logan presented a method of belief contraction for rule-based agents that satisfies the AGM postulates, respects preferences over facts and can be computed in polynomial time. Beishui Liao proposed the concept of a layered argumentation framework with subargument relations, allowing the structure of arguments to be expressed at the abstract level. Some preliminary considerations regarding dynamics (i.e. adding and removing arguments) were presented.
In sum, the workshop proved to be a successful event. The quality of the contributions was high and the discussions were lively. The discussions centered around topics such as the connection between dynamics of argumentation, dynamics of logic programming and belief change, and the issue of abstract vs. instantiated or structured argumentation.
