Nguyen Duy Hung

A Generalization of Probabilistic Argumentation with Dempster-Shafer Theory

One of the first (and key) steps in analyzing an argumentative exchange is to reconstruct complete arguments from utterances which may carry just enthymemes. In this paper, using legal argument from analogy, we argue that in this reconstruction process interpreters may have to deal with a kind of uncertainty that can be appropriately represented in Dempster-Shafer (DS) theory rather than classical probability theory. Hence we generalize and relax existing frameworks of Probabilistic Argumentation (PAF), which are currently based on classical probability theory, by what we refer to as DS-based Argumentation framework (DSAF). Concretely, we first define a DSAF form and semantics by generalizing existing PAF form and semantics. We then present a method to translate existing proof procedures for standard Abstract Argumentation into DSAF inference procedures. Finally we provide a Prolog-based implementation of the resulted DSAF inference procedures.

Inference procedures and engine for probabilistic argumentation

Abstract Probabilistic Argumentation (PA) is a recent line of research in AI aiming to combine the strengths of argumentation and probabilistic reasoning. Though several models of PA have been proposed, the development of practical applications is still hindered by the lack of inference procedures and reasoning engines. In this paper, we present a reduction method to compute a recently proposed model of PA called PABA. Using the method we design inference procedures to compute the credulous semantics, the ideal semantics and the grounded semantics for a general class of PABA frameworks, that we refer to as Bayesian PABA frameworks. We also show that, though restricting to Bayesian PABA frameworks, the inference procedures can be used to compute other PA models thanks to simple translations. Finally, we implement the inference procedures to obtain a multi-semantics engine for probabilistic argumentation and demonstrate its usage.

Computing probabilistic assumption-based argumentation

We develop inference procedures for a recently proposed model of probabilistic argumentation called PABA, taking advantages of well-established dialectical proof procedures for Assumption-based Argumentation and Bayesian Network algorithms. We establish the soundness and termination of our inference procedures for a general class of PABA frameworks. We also discuss how to translate other models of probabilistic argumentation into this class of PABA frameworks so that our inference procedures can be used for these models as well.

Argument-Based Logic Programming for Analogical Reasoning

Analogical reasoning can be understood as a kind of resemblance of one thing to another, thus assigning properties from one context to another. The key idea is to use similarity information to support an inference which cannot be deductively inferred. In this paper, we present a formal and intuitive framework of this phenomena using an argument-based logic-programming-like language. A proof theory of our system is stated in the dialectical style, where a proof takes the form of dialogue between a proponent and an opponent of an argument. We also discuss how the proposed framework can be fine tuned for optimistic analogical reasoning and pessimistic analogical reasoning. Finally, we discuss a design sketch of our proposed analogical reasoner called Analogist.

Inference and Learning in Probabilistic Argumentation

Inference for Probabilistic Argumentation has been focusing on computing the probability that a given argument or proposition is acceptable. In this paper, we formalize such tasks as computing marginal acceptability probabilities given some evidence and learning probabilistic parameters from a dataset. We then show that algorithms for them can be composed by finely joining a basic PA inference algorithm and existing algorithms for the corresponding tasks in Probabilistic Logic Programming or even Bayesian networks.

The Distribution Semantics of Extended Argumentation

The distribution semantics is a de facto approach for integrating logic programming with probability theory, and recently has been applied for the standard abstract argumentation framework. In this paper, we define the distribution semantics for extended argumentation frameworks, and moreover derive inference procedures from existing proof procedures of such extended argumentation frameworks. While doing so we focus on extended argumentation frameworks with attacks on attacks and the inductive defense semantics thereof. However our results can be easily obtained for other extended frameworks and semantics.

Combining Answer Set Programming with Description Logics for Analogical Reasoning Under an Agent’s Preferences

Analogical reasoning makes use of a kind of resemblance of one thing to another for assigning properties from one context to another. This kind of reasoning is used quite often by human beings, especially in unseen situations. The key idea of analogy is to identify a good similarity; however, similarity may be varied on subjective factors (i.e. an agent’s preferences). This paper studies an implementation of this phenomena using an answer set programming with Description Logics. The main idea underlying the proposed approach lies in the so-called Argument from Analogy developed by Walton [1]. Finally, the paper relates the approach to others and discusses future directions.

A Probabilistic Argumentation Engine

We develop an inference procedure for the grounded semantics of Probabilistic Assumption-based Argumentation (PABA) using the approach we used to develop inference procedures for PABAs credulous and ideal semantics. We establish the soundness and termination of the new procedure for a general class of PABA frameworks. We also implement the new procedure, together with the previous procedures, obtaining a multi-semantics Probabilistic Argumentation Engine.

A Simulation-Based Framework for Distribution Service Restoration

Service restoration in a distribution power system aims at finding a series of switching operations - a restoration plan, that restores power given an occurrence of a fault, considering multiple objectives and constraints. In this research, we first elaborate that many of common objectives and constraints depend on plan’s physical properties that can be obtained efficiently by simulating the plan, using simulation software like DIgSILENT PowerFactory. However other objectives/constraints cannot be computed by physical simulation because they depend on non-physical and/or non-functional properties not intrinsic to the plan, such as the power companys policies and the market information. From this observation, we propose a framework for distribution service power restoration integrating multiple knowledge sources: heuristics representing the expertise of experienced operators are used to construct possible plans, DIgSILENT PowerFactory computes plan’s physical properties, while non-physical properties are provided by relevant knowledge bases. To demonstrate the appropriateness of the framework, we developed a proof-of-concept implementation.

Probabilistic assumption-based argumentation with DST evidence

We study the relationships between two prominent approaches to, respectively, non-additive degrees of belief and probabilistic argumentation: Demspter-Shafer Theory (DST) and Probabilistic Assumption-based Argumentation (PABA). In particular we show that each DST body of evidence can be represented by a PABA framework, and DST combination rules can be simulated by a canonical rule for combining PABA frameworks. We then develop a PABA framework capable of taking DST evidence directly besides logical knowledge. We illustrate how this framework naturally models a decision making situation that DST alone seems ill-suited.