Main content

Top content

Modeling and Solving Weighted Bipolar Argumentation Problems

Tutorial at the 42nd German Conference on Artificial Intelligence, University of Kassel

Abstract

Weighted bipolar argumentation frameworks are a computationally efficient tool to solve problems in areas like decision support and social media analysis. This tutorial will give a high-level introduction to some recent frameworks and the corresponding computational problems and algorithms.

Overview

Weighted bipolar argumentation frameworks can be applied to problems like decision support and social media analysis. Problems are modeled as weighted directed graphs, where nodes correspond to arguments and edges to attack or support relations. The weights of arguments correspond to an initial weight that can be based on the plausibility of an argument or the reputation of the argument's source. The computational problem is then to assign a final strength value to each argument based on its initial weight and the strength of its attackers and supporters. Afterwards, the final strength values can be used to decide which arguments can be accepted.

Weighted bipolar argumentation is potentially interesting for many AI researchers. For example, acceptance can correspond to making a particular choice in decision support. In social media analysis, the final strength value can be associated with the public opinion about a claim. Weighted bipolar argumentation is also computationally attractive. Problems with thousands of nodes and ten thousands of edges can be solved within seconds on a regular computer. This can make them a valuable alternative to computationally expensive probabilistic reasoning approaches, when no rigorous probabilistic interpretation is required.

Target Audience

We will focus on the high-level ideas and the tutorial will be mostly self-contained. No prior knowledge is required. Familiarity with Java may be helpful for the last part about implementing and solving weighted argumentation problems with Attractor, but the code snippets will be comprehensible for everyone with a basic understanding of any programming language.