Hajime Sawamura

JVenn: A Visual Reasoning System with Diagrams and Sentences

Deduction by a computer studied so far has been centered around symbolic reasoning with formulas. Recently, attention has been directed to reasoning with diagrams as well, in order to augment the deficiency of reasoning with symbols only. In this paper, we propose a visual reasoning system called JVenn which attains a unique amalgamation of the diagrammatic reasoning and the symbolic reasoning, having perspicuity of diagrams and strictness of symbols complementarily. JVenn is unique particularly in the points that it has the strategy for proving a chain of syllogisms, allows for an interplay between diagrams and symbols, and guides reasoning with the beauty measure for diagrams.

Argumentation-Based Learning

Computational argumentation has been accepted as a social computing mechanism or paradigm in the multi-agent systems community. In this paper, we are further concerned with what agents believe after argumentation, such as how agents should accommodate justified arguments into their knowledge bases after argumentation, what and how agents acquire or learn, based on the results of argumentation. This is an attempt to create a new learning method induced by argumentation that we call Argument-Based Learning (ABL). To this end, we use our logic of multiple-valued argumentation LMA built on top of Extended Annotated Logic Programming EALP, and propose three basic definitions to capture agents’ beliefs that should be rationally acquired after argumentation: knowledge acquisition induced by the undercut of assumptions, knowledge acquisition induced by difference of recognition, and knowledge acquisition induced by rebut. They are derived from two distinctive and advantageous apparatuses of our approach to multi-valued argumentation under : Paraconsistency and multiple-valuedness that EALP and LMA feature. We describe an overall argument example to show the effectiveness and usefulness of the agent learning methods based on argumentation.

Towards an argument-based agent system

In network computing and agent-oriented computing, it is desired that several computers on network can resolve conflicting problems or make better solutions through argumentation. In this paper, we propose such a system in which several agents communicate, argue with each other and finally make a decision through argumentation from distributed non-sharing knowledge bases with two sorts of negation: negation as failure (or weak negation) and classical (or strong) negation. We show the potential and practical usefulness of the system by applying it to a realistic problem.

How agents should exploit tetralemma with an eastern mind in argumentation

Argumentation is a ubiquitous but effective mode of interaction and dialogue in the human society. It has come to be known that argumentation has many implications to interaction among computational agents as well. After observing and discussing the tetralemma, which is said to characterize the Eastern thought, in this paper we propose an argumentation framework with the paraconsistent logic programming based on the tetralemma. It allows us to represent typical eastern modes of truth:

Argumentation as a Social Computing Paradigm

\top, \bot

Mechanizing Relevant Logics with HOL

which are considered epistemic states of propositions. We introduce various notions for our argumentation framework, such as attack relations in terms of differences as a momentum of argumentation, argument justification, preferential criteria of arguments based on social norms, and so on, in a way proper to the four-valued paraconsistent logic programming. Finally, we provide the fixpoint semantics and dialectical proof theory for the argumentation framework. We illustrate our ideas with various argument examples.

Computing abductive argumentation in answer set programming

In this paper, we claim that argumentation is a novel and prominent computing principle and provides a unified approach to technologies needed in agent-oriented computing, where social concepts play important roles in computation. Viewed as the reasoning methods of attaining a consensus, they can be roughly classified into three categories: (i) conflict-resolving reasoning, (ii) dialectical reasoning, and (iii) cooperative reasoning. We describe these formally in a unified manner, and build an argument-based agent system with those argument-based reasoning capabilities. Finally, we show its potential usefulness and feasibility in a convincing manner by applying it to a wide variety of the contemporary application domains.

A hybrid argumentation of symbolic and neural net argumentation (Part II)

Relevant logics are non-classical logics, whose motivation is to remove logical fallacies caused by the classical “implication≓. In this paper, we propose a method to build an interactive theorem prover for relevant logics. This is done first by translating the possible world semantics for relevant logics to the higher-order representation of HOL, and then under the HOL theory obtained by this translation, relevant formulas are shown to be valid using the powerful HOL proof capabilities such as backward reasoning with tactics and tacticals. Relevant logics we have dealt with so far includes Routley and Meyers R system (originally Hilbert-type axiomatization) and Reads R system (basically Gentzentype axiomatization). Our various proof experiences of relevant formulas by HOL and their analyses yielded a powerful proof heuristics for relevant logics. It actually allowed us to prove a formula which has been known to be difficult for traditional theorem provers and even relevant logicians.

An Integrated Argumentation Environment for Arguing Agents

In our daily life, humans often argue with each other using abductive knowledge which includes not only facts known to be true but also hypotheses that may be expected to be true. This paper presents a novel approach to find out every skeptical (resp. credulous) explanation which is the set of hypotheses needed to skeptically (resp. credulously) justify the argument supporting a disputers claim based on abductive knowledge base under the specified argumentation semantics. The main subject of this paper is the definition of the Abductive Argumentation Framework which is equivalent to the widely adopted Dungs framework except handling hypotheses, and from which skeptical (resp. credulous) explanations in argumentation can be defined. In general, there are multiple explanations under the specified argumentation semantics. Our approach is capable of finding out all of them by means of applying traditional abductive logic programming to our previous work of computing argumentation semantics in answer set programming (ASP). Thus this study eventually reveals the greatest advantage of applying ASP to the crucial decision problems in the research field of argumentation.

An integrated argumentation environment for arguing agents

A novel approach to argumentation has been started by A. Garcez et al. Inspired by their work, we further go on investigating it, but turn to a more syncretic direction such as the interplay between neural net argumentation and symbolic argumentation. This paper is a sequel to our former one (Part I) [1]. In this paper we address ourselves to the following basic questions that can not be overlooked. 1. Can the neural argumentation algorithm compute the fixpoint semantics for formal argumentation? 2. Can argumentative dialogues be extracted from the neural net argumentation? Consequently, we give the positive answers to them. They are beneficial for understanding or characterizing the computation power and outcome of the neural net argumentation from the perspective of the symbolic argumentation. We also exemplify these results.