In the current work, we provide a formal Mackenzie-style persuasion dialogue for grounded semantics. We show that an argument is in the grounded extension iff the proponent is able to persuade a maximally sceptical opponent in the dialogue.
Dionysios Kontarinis, Elise Bonzon, Nicolas Maudet, Pavlos Moraitis
486 - 497
Agents contributing to (online) debate systems often have different areas of expertise. This must be considered if we want to define a decision making process based on the output of such a system. Distinguishing agents on the basis of their areas of expertise also opens an interesting perspective: when a debate is deemed “controversial”, calling an additional expert may be a natural way to make the decision easier. We introduce possible definitions that capture these notions and we provide a preliminary analysis with the objective to help a designer find the “right” expert.
“The Synergy” is an on-line collaborative argument-based decision making platform. Our goal is to create a system allowing for both user-driven (the users themselves can vote “for” and “against” any particular option) and machine-driven (the system can propose an order of options based on the arguments provided by users) decision making. For the second option, we implemented existing and newly developed decision-making criteria. The basic concepts of our system are an option, a goal and an argument. An argument links an option with a goal. It can be in favour of or against an option and it can be attached a probability measure, which we believe is necessary for representing numerous scenarios in decision making under uncertainty. Our long term goal is to have pre-made answers for some general decisions: like Wikipedia collects data, we will collect PROS and CONS of possible decisions.
Most reasoning tasks in abstract argumentation are in general computationally hard. One approach of dealing with such problems stems from the field of parameterized complexity theory. For so-called fixed-parameter tractable algorithms, one identifies problem parameters, e.g. the graph parameter tree width, such that the run-time of algorithms heavily scales with the parameter but only polynomially with the input size. The dynPARTIX system turns these fixed-parameter tractability results into practice by implementing dynamic programming algorithms for the graph parameter tree width.
In this paper, we present TOAST, a system that implements the ASPIC+ framework. TOAST accepts a knowledge base and rule set with associated preference and contrariness information, and returns both textual and visual commentaries on the acceptability of arguments in the derived abstract framework.
Mark Snaith, Floris Bex, John Lawrence, Chris Reed
511 - 512
In this paper, we present ArguBlogging, a simple tool that allows blog users to directly respond to text on a web page, publishing the response to their blog while simultaneously capturing the argumentative structure in the Argument Web via the Argument Interchange Format.
In this paper, we present Arvina, an online discussion tool supporting mixed initiative argumentation. Arvina allows stored arguments in the Argument Web to be introduced by software agents which human participants can then interact with.
Reinforcement Learning (RL) suffers from several difficulties when applied to domains with no obvious goal-state defined; this leads to inefficiency in RL algorithms. We consider a solution within the context of a widely-used testbed for RL: RoboCup Keepaway. We introduce Argumentation-Based RL (ABRL), using methods from argumentation theory to integrate domain knowledge, represented by arguments, into the SMDP algorithm for RL by using potential-based reward shaping. Empirical results show that ABRL outperforms the original SMDP algorithm, for this game, by improving convergence speed and optimality.