KR 2018 tutorial: An overview of ranking-based argumentation semantics (T3)

Welcome to the page of this KR 2018 tutorial. Tutorials and workshops will be held from 27 to 29 October 2018, prior to the KR main technical program, which will run from 30 October to 1 November 2018. The attendance of tutorials is complimentary to all KR registered participants. Workshop attendance will be subject to payment of a workshop fee, which is separate from that of the main conference.

Brief description of the tutorial

Argumentation is a process of constructing arguments and attacks between them. Traditionally, an argument can have three different statuses: sceptically accepted, credulously accepted and rejected. Recently, in order to allow for a more fine graded evaluation of arguments, ranking-based semantics were introduced. They allow to order the arguments from the most to the least acceptable one. Most of them also attach a numerical score to each argument. We start by a general introduction to argumentation theory. Then, we study the main motivation and ideas behind ranking-based semantics. We continue by reviewing the principles for ranking by semantics from the literature. Those are desirable properties that can be satisfied by a ranking-based semantics. We study the links between the principles (e.g. one set of principles implies another principle; some sets of principles are incompatible). In the last part of the tutorial, we introduce the existing ranking-based semantics from the literature. We illustrate how they work on examples and show which principles are satisfied by which semantics.

When?

Duration: half-day
When: Monday, October 29, 2018 (morning)
Schedule:
09:00 - 10:30 session 1
10:30 - 11:00 coffee break
11:00 - 12:30 session 2

Target audience

This tutorial is suitable for both those who A gentle introduction to argumentation theory will be provided in the first part of the tutorial.

Why is the tutorial of interest to the KR audience?

Argumentation is the key word of the KR conference. Also, many people who have background in decision making, explanation finding, non monotonic logic, reasoning under inconsistency or uncertainty might be interested in recent developments in the area of argumentation theory.

Brief resume of the presenter

Srdjan Vesic is a CNRS scientific researcher at CRIL laboratory in Lens (France) since 2013. From 2011 to 2013, he was a post-doctoral researcher at University of Luxembourg, where he held an ERCIM postdoctoral fellowship and an FNR AFR fellowship, both co-funded by Marie-Curie Actions. Before that, he received his PhD from IRIT - University of Toulouse in 2011. His research interests include argumentation theory, reasoning under uncertainty and/or inconsistency and computational social choice. Srdjan has published more than thirty peer-reviewed papers. His publication record includes Journal of Artificial Intelligence Research, Social Choice and Welfare, International Journal of Approximate Reasoning, as well as conferences like IJCAI, AAAI, ECAI and KR. He was a keynote speaker at GKR 2017@IJCAI. He has served on the program committee in several AI conferences (IJCAI, AAMAS, ECAI...). He is also a reviewer for Artificial Intelligence Journal, Journal of Artificial Intelligence Research and Artificial Intelligence Review. He likes water sports, mountaineering, skiing, classical music and travelling.

To find out more about Srdjan, click here.

Outline of the tutorial

Argumentation is a process of constructing arguments and attacks between them. The most abstract and the most general framework for formal argumentation was introduced by Dung in 1995. The input is represented in form of a graph where arguments correspond to nodes and attacks correspond to edges between the nodes. An extension is a set of arguments that can be accepted together. The output of this framework is a set of extensions. An argument is said to be sceptically accepted if it belongs to all extensions, credulously accepted if it belongs to some but not to all extensions, and rejected if it does not belong to any extension. This means that each argument can have three different statuses.

Recently, in order to allow for more fine graded evaluation of arguments, ranking-based semantics were introduced. They allow to order the arguments, from the most acceptable, or the best, to the worst argument. Most of the existing ranking-based semantics also attach a numerical score to each argument.

The tutorial is composed of five parts. We conclude the tutorial by mentioning some open problems, challenges and possible research directions in this area.

References


Slides

Download the slides and the cheat sheet.

Contact

If you have any questions about the tutorial, do not hesitate to contact me at vesic@cril.fr

  Last update: May 2018.