Change in quantitative bipolar argumentation: Sufficient, necessary, and counterfactual explanations

Bipolar Argumentation Frameworks (BAFs) are an important class of argumentation frameworks useful for capturing, reasoning with, and deriving conclusions from debates. They have the potential to make solid contributions to real-world multi-agent systems and human-agent interaction in domains such as legal reasoning, healthcare and politics. Despite this fact, practical systems implementing BAFs are largely lacking. In this demonstration, we provide a software system implementing novel algorithms for calculating extensions (winning sets of arguments) of BAFs. Participants in the demonstration will be able to input their own debates into our system, and watch a graphical representation of the algorithms as they process information and decide which sets of arguments are winners of the debate.

We introduce Incomplete Bipolar Argumentation Frameworks (iBAFs), the extension of Dung’s Abstract Argumentation Frameworks (AAFs) allowing the simultaneous presence of supports (borrowed from BAFs – Bipolar AAFs) and of uncertain elements of the argumentation graph (borrowed from iAAFs – incomplete AAFs). We investigate the computational complexity of verification problem (under the possible perspective) and the acceptance problem, by studying its sensitivity to the semantics of supports and the semantics of extensions. On the one hand, we show that adding supports on top of incompleteness does not affect the complexity of the acceptance. On the other hand, surprisingly, we show that the joint use of bipolarity and incompleteness has a deep impact on the complexity of the verification: for the semantics under which the verification over AAFs is polynomial-time solvable, although moving from AAFs to BAFs or to iAAFs does not change the complexity, the complexity of the verification over iBAFs may increase up to NP-complete.

We introduce a novel method to aggregate Bipolar Argumentation (BA) Frameworks expressing opinions by different parties in debates. We use Bipolar Assumption-based Argumentation (ABA) as an all-encompassing formalism for BA under different semantics. By leveraging on recent results on judgement aggregation in Social Choice Theory, we prove several preservation results, both positive and negative, for relevant properties of Bipolar ABA.

We develop a method of reasoning using an incrementally constructed bipolar argumentation framework (BAF) aiming to apply computational argumentation to legal reasoning. A BAF that explains the judgment of a certain case is constructed based on the user’s knowledge and recognition. More specifically, a set of effective laws are derived as the conclusions from evidential facts recognized by the user, in a bottom-up manner; conversely, the evidences required to derive a new conclusion are identified if certain conditions are added, in a top-down manner. The BAF is incrementally constructed by repeated exercise of this bidirectional reasoning. The method provides support for those who are not familiar with the law, so that they can understand the judgment process and identify strategies that might allow them to win their case.

As algorithmic decision-making systems become more prevalent in society, ensuring the fairness of these systems is becoming increasingly important. Whilst there has been substantial research in building fair algorithmic decision-making systems, the majority of these methods require access to the training data, including personal characteristics, and are not transparent regarding which individuals are classified unfairly. In this paper, we propose a novel model-agnostic argumentation-based method to determine why an individual is classified differently in comparison to similar individuals. Our method uses a quantitative argumentation framework to represent attribute-value pairs of an individual and of those similar to them, and uses a well-known semantics to identify the attribute-value pairs in the individual contributing most to their different classification. We evaluate our method on two datasets commonly used in the fairness literature and illustrate its effectiveness in the identification of bias.

The bipolar argumentation framework (BAF) setting is an extension of Dung’s setting for abstract argumentation, that considers an additional relation, called support relation. Several interpretations of such a support relation have been pointed out so far, including deductive, necessity, general and backing supports. These notions of support capture different kinds of interactions between arguments, that do not primarily correspond to attacks. In this paper, we propose a new notion of support, called monotonic support. Our approach is axiomatic: two postulates are introduced for capturing the intuition that underlies this notion of support in formal terms. The first postulate, monotony, prevents the support relation from downgrading the acceptance status of the supported argument. The second postulate, non-triviality, requires the existence of BAFs for which supporting an argument leads to increase its acceptance status. We present a general family of extension-based semantics for BAFs, called support score-based (SSB) semantics, that satisfy the two postulates and are parameterized by some aggregation functions. We prove a characterisation result linking the postulates that a SBB semantics satisfies with the properties of the aggregation functions used to define it. We also show that none of the previously introduced semantics for BAFs satisfies the monotony postulate.

In the research on dynamics of argumentation frameworks (AFs), the enforcement problem deals with changing an AF for the purpose of ensuring that a certain set of desired arguments becomes (part of) an extension. In this paper we focus on expansions of an AF where solely the addition of new arguments and attacks is allowed and the original framework remains unchanged. Existing results about the enforcement problem are all sufficient conditions, and we argue that necessary and sufficient conditions are essential concerning the solvability of the enforcement problem. A suite of jointly necessary and sufficient conditions are identified for the enforcement problem under normal expansion. These conditions are presented from multiple perspectives including the sub-strict versus non-strict problem, constructive versus eliminative conditions and arbitrary AFs versus even(odd)-length cycle free AFs, determined by the nature of various argumentation semantics. Such necessary and sufficient conditions are of the essence to the enforcement problem by defining requirements that are both minimum and absolute, which can be used to determine when the enforcement problem is (un)solvable. We further infer solely sufficient and solely necessary conditions based on the partial order relation (w.r.t. set-inclusion) among various types of AF extensions.

Counterfactual explanations (CXs) are widely recognised as an essential technique for providing recourse recommendations for AI models. However, it is not obvious how to determine CXs in model multiplicity scenarios, where equally performing but different models can be obtained for the same task. In this paper, we propose novel qualitative and quantitative definitions of CXs based on explicit, nested quantification over (groups) of model decisions. We also study properties of these notions and identify decision problems of interest therefor. While our CXs are broadly applicable, in this paper we instantiate them within computational argumentation where model multiplicity naturally emerges, e.g. with incomplete and case-based argumentation frameworks. We then illustrate the suitability of our CXs for model multiplicity in legal and healthcare contexts, before analysing the complexity of the associated decision problems.

Argumentation frameworks have been widely studied both in terms of formal properties they exhibit under different semantics and in terms of applications they can support. But where are argumentation frameworks coming from, and how can argumentation, a model-based approach to AI, beneficially integrate with the nowadays-much-widespread data-centric AI perspective? In this talk I will overview applications empowered by a variety of (extension-based and gradual) semantics for abstract and bipolar argumentation frameworks automatically obtained from data (including but not limited to text) and from logical rules. Some of these applications require the integration of argumentation and machine learning, and result in a mixed model-based and data-centric pipeline. For some applications, the semantics informs the definition of the frameworks rather than, as is conventionally the case, being enforced on frameworks a posteriori.

In this paper, we extend the abstract argumentation framework proposed by [1] in order to take into account two kinds of interaction between arguments: a positive interaction (an argument can help, support another argument) and a negative interaction (an argument can attack another argument). In this new abstract argumentation framework, called a bipolar argumentation framework, we propose a gradual interaction-based valuation process. With this process, the value of each argument A only depends on the value of the arguments which are directly interacting with A in the argumentation system.

We introduce Forecasting Argumentation Frameworks (FAFs), a novel argumentation-based methodology for forecasting informed by recent judgmental forecasting research. FAFs comprise update frameworks which empower (human or artificial) agents to argue over time about the probability of outcomes, e.g. the winner of an election or a fluctuation in inflation rates, whilst flagging perceived irrationality in the agents' behaviour with a view to improving their forecasting accuracy. FAFs include five argument types, amounting to standard pro/con arguments, as in bipolar argumentation, as well as novel proposal arguments and increase/decrease amendment arguments. We adapt an existing gradual semantics for bipolar argumentation to determine the aggregated dialectical strength of proposal arguments and define irrational behaviour. We then give a simple aggregation function which produces a final group forecast from rational agents' individual forecasts. We identify and study properties of FAFs, and conduct an empirical evaluation which signals FAFs' potential to increase the forecasting accuracy of participants.

This paper deals with strategical issues of arguing agents in a multi-agent setting. We investigate different scenarios of such argumentation games that differ in the protocol used for argumentation, i.e. direct, synchronous, and dialectical argumentation protocols, the awareness that agents have on other agents beliefs, and different settings for the preferences of agents. We give a thorough investigation and classification of these scenarios employing structured argumentation frameworks which are an extension to Dung’s abstract argumentation frameworks that give a simple inner structure to arguments. We also provide some game theoretical results that characterize a specific argumentation game as strategy-proof and develop some argumentation selection strategies that turn out to be the dominant strategies for other specific argumentation games.

Shoham's logic of chronological ignorance (CI) provides a formal approach for representing and reasoning about causal relationships. The utility of the formalism comes from imposing a partial order on Kripke interpretations for theories in a modal logic of temporal knowledge. In this paper we show that the use of modal logic in this context is unnecessary and that the same result can be achieved using classical logic. While semantically simpler, the classical approach is somewhat unwieldy. To overcome this problem we suggest an alternative underlying language, called asserted logic. The logic allows the convenient representation found in Shoham's approach without sacrificing the truth functional semantics and proof theory of classical logic. We demonstrate the utility of the logic by providing an equivalent nonmodal CI framework for causal theories.