AbstractArgumentation is an approach to reasoning that can be implemented in machines because of its logical foundations, and which is easily understood by humans because of its dialectical nature. By enabling humans to reason with machines through dialogues, argumentation accommodates even non-experts to scrutinise and communicate with computational agents. Such technology is valuable at a time in which the need for accountability in intelligent machines is ever increasing, and as human-machine interaction becomes ever more commonplace.
However, if practical argument-based applications are to be realised, the technologies and systems that underpin them should be effective and efficient. Furthermore, applications should be optimised to run on the kinds of structures of argumentation which exist in the domain that they operate in.
In this thesis, therefore, we seek to understand the performance of computational argument based dialogue systems. We begin by investigating the effect of domain on the performance of these systems, and then develop two approaches to strategic reasoning in argument-based dialogues that are computationally efficient and effective.
|Date of Award||1 Jun 2019|
|Supervisor||Elizabeth Black (Supervisor) & Michael Luck (Supervisor)|