What should we do? : computational representation of persuasive argument in practical reasoning

What Should We Do?” by Katie Atkinson The design and development of autonomous software agents requires a multitude of elements to be considered and accounted for. In order for a software agent to be considered ‘intelligent’ it must be able to perform effective reasoning about its beliefs and the environment in which it is situated, and also act in this environment. It must therefore reason both about what is the case, and what should be done: the latter is known as practical reasoning. Additionally, it must also be able to interact and reason with other such agents in its environment, as it may rely on them for information and help to enable it to accomplish its tasks. This thesis is concerned with one particular aspect of such agency: modelling the process of argument in practical reasoning to equip autonomous agents with the capability to determine the best action to take, in a given situation. The background setting for this work deals with the topic of practical reasoning and attempts to address some issues regarding its treatment in philosophy, as well as the problems inherent in the computational modelling of such reasoning. The main output of the study is a theory of persuasion in practical reasoning which makes use of techniques from the field of argumentation theory, to enable autonomous software agents to construct and reason about arguments in support of and against proposals for action. The theory is embodied in a model describing how agents based on the BeliefDesire-Intention (BDI) architecture can put forward a proposal for action and how this proposal can be systematically attacked in a variety of ways. This enables them to consider all available options and come to a conclusion about the best action to take, in the given context. The underlying theory extends a well established account from the field of philosophy, based on the use of argument schemes and critical questions. The account given is then formalised in terms to enable its representation in agent systems. The underlying theory has formed the basis for a number of applications: an implementation of a dialogue game protocol to provide a proof of concept; an implementation to provide computer mediated support for human decision making in a particular context; and finally, a formalism to enable autonomous agents to reason about decisions regarding actions. The account for use in BDI agents is applied to three example domains – law, medicine and politics – to show how BDI agents can reason and argue about matters of practical action, in accordance with the theory.