Date of Degree

5-2019

Document Type

Dissertation

Degree Name

Ph.D.

Program

Computer Science

Advisor

Simon Parsons

Committee Members

Changhe Yuan

Sergei Artemov

Nir Oren

Subject Categories

Other Computer Engineering

Keywords

Argument, Non-monotonic reasoning, Argumentation framework, defeasible reasoning

Abstract

Argumentation theory is concerned with the way that intelligent agents discuss whether some statement holds. It is a claim-based theory that is widely used in many areas, such as law, linguistics and computer science. In the past few years, formal argumentation frameworks have been heavily studied and applications have been proposed in fields such as natural language processing, the semantic web and multi-agent systems. Studying argumentation provides results which help in developing tools and applications in these areas. Argumentation is interesting as a logic-based approach to deal with inconsistent information. Arguments are constructed using a process like logical inference, with inconsistencies giving rise to conflicts between arguments. These conflicts can then be handled by well-founded means, giving a consistent set of well-justified arguments and conclusions. Dung's seminal work tells us how to handle the conflicts between arguments. However, it says nothing about the structure of arguments, or how to construct arguments and attack relationships from a knowledge base. ASPIC+ is one of the most widely used systems for structured arguments. However, there are some limitations on ASPIC+ if it is to satisfy widely accepted standards of rationality. Since most of these limitations are due to the use of strict rules, it is worth considering using a purely defeasible subset of ASPIC+. The main contribution of this dissertation is the purely defeasible argumentation framework ASPIC+D. There are three research questions related to this topic which are investigated here: (1) Do we lose anything in removing the strict elements? (2) Do purely defeasible version of theories generate the same results as the original theories? (3) What do we gain by removing the strict elements? I show that using ASPIC+D, it is possible, in a well-defined sense, to capture the same information as using ASPIC+ with strict rules. In particular, I prove that under some reasonable assumptions, it is possible to take a well-defined theory in ASPIC+, that is one with a consistent set of conclusions, and translate it into ASPIC+D such that, under the grounded semantics, we obtain the same set of justified conclusions. I also show that, under some additional assumptions, the same is true under any complete-based semantics. Furthermore, I formally characterize the situations in which translating an ASPIC+ theory that is ill-defined into ASPIC+D will lead to the same sets of justified conclusions. In doing this I deal both with ASPIC+ theories that are not closed under transposition and theories that are axiom inconsistent. At last, I analyze the two systems in the context of the non-monotonic axioms. I show that ASPIC+ and ASPIC+D satisfy exactly same axioms under what I call the “argument construction” interpretation and the “justified conclusions” interpretation under the grounded semantics. Furthermore, because of the lack of strict elements, ASPIC+ satisfies more of the non-monotonic axioms than ASPIC+ in the ``justified conclusions'' interpretation under the preferred semantic. This means that ASPIC+ and ASPIC+D may not have the same justified conclusions under the preferred semantics.

Share

COinS