In this talk I will describe recent work on developing classical logic based models of argumentation for use by distributed, resource-bounded agents. The talk will be divided into three parts. I will first describe how argumentation based characterisations of non-monotonic inference, enable computational (and human) agents to interactively engage in epistemic and practical reasoning. I will then review existing approaches to argumentation based non-monotonic inference over classical logic knowledge bases, and show that these approaches have a number of limitations. Specifically, they: 1) fail to accommodate real world modes of dialectical reasoning (such as the Socratic move whereby an interlocutor is shown to contradict themselves), and; 2) only yield rational conclusions under the assumption that agents have unlimited resources. The third part of the talk will then review recent work with M. D'Agostino (University of Milan), in which we upgrade the ontology of arguments(essentially proofs in classical logic) so as to explicitly distinguish premises assumed true, and those supposed true for the sake of argument. I will show how our approach addresses the aforementioned limitations, in that it enables real world dialectical reasoning amongst resource bounded agents, while still ensuring that the conclusions reached are rational.