Course: An introduction to abstract argumentation

Video: pt1 and pt2

Tutor: Pietro Baroni

Abstract: The course will provide an introduction to the theory of abstract argumentation frameworks, discussing its basic concepts and analyzing in detail the notion of abstract argumentation semantics, with a coverage of the general properties characterizing different semantics and of the main proposals available in the literature. After covering these foundational aspects, the final part of the course will give an overview of various research trends devoted to developing and extending the basic abstract argumentation model.

Course: Dialectical Formalisations of Non-monotonic Reasoning via Structured Argumentation

Video: pt1 and pt2

Tutor: Sanjay Modgil

Abstract: The aim of the tutorial is to give an overview of the ASPIC+ framework for structured argumentation, and developments and applications of ASPIC+. I will describe how the framework allows one to specify construction of arguments from a knowledge base (KB) of premises in a logical language of one's choosing, and inference rules of one's choosing, which may be the inference rules of a deductive logic and/or defeasible (default) inference rules. The constructed arguments can then be related to each other, according to whether one argument is a counter-argument to (i.e., 'attacks') another. Preferences over arguments can then determine which attacks are successful. The arguments, related by successful attacks, are then evaluated using Dung's argumentation theory (reviewed in the preceding tutorial by Pietro Baroni), so that the claims of the winning (justified) arguments identify the non-monotonic inferences from the underlying KB. ASPIC+ thus enables dialectical formalisations of non-monotonic logics, and so paves the way for distributed non-monotonic reasoning amongst multiple agents, in the form of dialogues (I will point to the importance of such dialogues in addressing ethical concerns relating to AI). Moreover, I will review ASPIC+ guidelines constraining the above choices, as well as how preferences are defined, in order to ensure that the outcome of argument evaluation yields rational outcomes. The tutorial will also cover some key developments of the ASPIC+ framework; notably: 1) A modification of ASPIC+ that studies preservation of rationality under resource bounds; 2) an extension of ASPIC+ that enables argumentation based reasoning about possibly conflicting preferences/values that are used to decide the success of attacks; 3) Evaluation of ASPIC+ arguments under generalisations of Dung's theory that yield more fine grained evaluation of the status of arguments (i.e., not just as winning, losing or undecided.

Course: Defeasible Logic

Video: pt1

Tutor: Guido Governatori

Abstract: Defeasible logic is a rule based proof theoretic computationally oriented non-monotonic logic whose reasoning mechanism is closely related to argumentation. In addition, the logic proved to be flexible enough to capture different and somehow incompatible facets of non-monotonic reasoning. In the lecture, we are going to present the basics of the logic, and how to generate variants modelling different aspects of non-monotonic reasoning. Moreover, we show how to provide argumentation semantics for the variants of the logic.

Course: Argumentation Mining

Video: pt1 and pt2

Tutor: Paolo Torroni

Abstract: Argumentation mining aims to automatically extract argumentative content from textual sources by exploiting natural language processing methods. The course is directed to a general audience and intends to offer a smooth introduction to argumentation mining by discussing models, methods, tasks, datasets and possible applications.

Course: Argumentation-based Dialogue

Video: pt1 and pt2

Tutor: Simon Parsons

Abstract: The notion of dialogue is closely related to the basic concepts of argumentation, and there is well-known dialectical basis for establishing which arguments are acceptable. Argumentation can also be used to underpin a formal model of dialogue. In such ``argumentation-based'' dialogue, participants exchange arguments, and the mechanisms of argumentation are used to establish what elements participants take to be acceptable at the end of the exchange. This tutorial covers the basics of such dialogues, discussing the elements that are required in order to carry out argumentation-based dialogues.

Course: Probabilistic Abstract Argumentation

Video: pt1

Tutor: Sergio Flesca

Abstract: Several proposals have been made to model uncertainty in Abstract Argumentation Frameworks (AAFs), by considering weights, preferences, or probabilities associated with arguments and/or defeats. One of the most popular approaches based on probability theory for modeling the uncertainty is the so called constellations approach: the dispute is represented by means of a Probabilistic Argumentation Framework (prAAF), that consists in a set of alternative scenarios, each of which is represented by a (deterministic) AAF (or, equivalently, an argumentation graph) associated with a probability. The various works in the literature investigating prAAFs make different assumptions on how the probability distribution function (pdf) over the scenarios is specified, for instance assuming that the pdf is defined "extensively", by enumerating all the possible scenarios and, for each of them, the value of its probability, or adopting the restriction that arguments and defeats are independent and specifying only the marginal probabilities associated with arguments and defeats. In this talk we give an overview of Probabilistic Argumentation Frameworks and the reasoning problems defined over them and briefly describe the various research trends related to prAAFs.

Course: Argument-based Normative and Legal Reasoning

Video: pt1 and pt2

Tutor: Guido Governatori, Giovanni Sartor

Abstract: The course is organized into two lectures.
Defeasible Logic for Deontic Reasoning.
Normative reasoning incorporates two characteristic features: norms are defeasible (in the sense that typically they admit exceptions) and norms provide conditions under which obligations/permissions/prohibitions are in force in a normative system. Also, norms can be violated, and normative systems provide mechanisms to compensate for violations. In the lecture, we are going to show how to integrate in a computationally oriented way Defeasible Logic (for the efficient treatment of exceptions) and Deontic Logic of Violations. Examples from real-life norms illustrate how the resulting logic works.
Argumentation in case based reasoning and interpretation in the legal domain.
The session will consider how lawyers develop arguments when reasoning with cases and when engaging with the interpretation of statutory provisions. We shall also focus on the way in which arguments based on rules and values are developed and interact. We shall then consider how this arguments can be captured by using abstract and structured argumentation. Finally we shall consider how these kinds of arguments and model can be relevant not only in the legal domains, but also in other domains where normative reasoning is needed, to direct and constrain choices or human and artificial agents.

Course: Formal Argumentation Approaches in TweetyProject

Video: pt1 and pt2

Tutor: Matthias Thimm

Abstract: The TweetyProject (http://tweetyproject.org) is a collection of various Java libraries that implement approaches to different areas of artificial intelligence. In this tutorial, I provide an overview on the argumentation libraries of the TweetyProject. These libraries comprise of implementations to abstract argumentation frameworks, as well as the most popular approaches to structured argumentation (ASPIC+, ABA, Deductive Argumentation, DeLP), and various further aspects such as ranking and probabilistic semantics. I briefly sketch the functionalities of these libraries and give some pointers to how they can be used.

Course: Argument technologies and online disinformation: pitfalls and prospects

Video: pt1

Tutor: Fabio Paglieri

Abstract: The COVID-19 pandemic has exacerbated pre-existing concerns about online disinformation, although extant evidence on the quality of online information in relation to this global event is still confused and controversial. On the one hand, the pandemic gave ample opportunities for the spread of disinformation online, especially via social media, including massive “information operations” (either state-sponsored or privately funded) to leverage the virus in fostering various agendas, both at the national and international level. On the other hand, increased attention and alertness of the population caused several positive instances of self-corrective behaviours, in which fake news were promptly detected, analyzed, and rectified using the same technological platforms responsible for their spread to begin with. In this rapidly shifting scenario, it becomes more urgent than ever to understand whether, how, and to what extent argument technologies might help in the collective fight against disinformation. In this tutorial I will emphasize the cognitive and socio-technical roots of online disinformation, to discuss how some current argument technologies relate to them and to outline a set of desiderata on how to design future argument-based tools to combat disinformation.

Invited Talk: Bridging the gap from linguistic to computational models of argument

Video: pt1

Speaker: Chris Reed

Abstract: One of the ways in which the COMMA community is composed of different threads of research is the varying extent to which use is made, on the one hand, of linguistic models, and, on the other, of formal techniques. Linguistic models have much to say about what an argument is, about how it is composed, and how it is situated in the world. Formal (both structured and abstract) models focus instead on a much narrower conception of what an argument is, and much more on one argument’s interactions with others. We have learnt that the Argument Interchange Format is an effective way of delivering data from the linguistic world and delivering computation back from the formal world, yet AIF on its own cannot handle real world data directly. This talk will summarise Inference Anchoring Theory, a technique that ties together discursive activity with formal structure, all of which can be represented in AIF and transported across the bridge to the rich formal and computational techniques.

Invited Talk: The Dynamics of Knowledge: Argumentation and Belief Revision

Video: pt1

Speaker: Guillermo R. Simari

Abstract: The exploration of the relationships between belief revision and computational argumentation has led to significant contributions for both areas; several techniques employed in belief revision are being studied to formalize the dynamics of argumentation frameworks and the capabilities of the argumentation-based defeasible reasoning are being used to define belief change operators. By briefly considering the fundamental ideas of both areas it is possible to examine some of the mutually beneficial cross-application in different proposals that model reasoning mechanisms that combine contributions from the two domains.