ľ�ϸ���Ӱ��

14:00 - 14:30: Henry Prakken: Expansion games as a component of a model of dialectical argument strength

14:30 - 15:00: Giuseppe Greco: Integrating the logical and the data-analytic perspectives in natural language

15:00 - 15:30: Coffee break

15:30 - 16:00: Beishui Liao: Logic for New Generation AI: An Argumentation-Based Methodology

16:00 - 16:30: Fei Liang: Modelling indicative conditionals as epistemic filter conditionals

16:30 - 17:00: Apostolos Tzimoulis: Categorization and deliberation

19:00  Workshop dinner

Abstracts and short biographies:

Henry Prakken: Expansion games as a component of a model of dialectical argument strength

Abstract: TBA In this talk I will discuss how I have used the theory of expansions of argumentation frameworks in my model of dialectical argument strength, recently published in the AI Journal. The model captures the degree to which an argument can be successfully attacked in expansions of an argumentation framework. This is formalised, among other things, with a game in which an opponent and a proponent of an argument in turn try to expand the 'current' argumentation framework in a way that decreases, respectively, increases the semantical status of an argument. Successful attack is defined as having a winning strategy in this game.

Reference: H. Prakken, An abstract and structured account of dialectical argument strength. Artificial Intelligence 335:104193, 2024.

Bio: Henry Prakken is a professor of artificial intelligence and law at Utrecht ľ�ϸ���Ӱ��, the Netherlands. Prakken has master degrees in law (1985) and philosophy (1988) from the ľ�ϸ���Ӱ�� of Groningen, the Netherlands. In 1993 he obtained his PhD degree (cum laude) at the Free ľ�ϸ���Ӱ�� Amsterdam with a thesis titled Logical Tools for Modelling Legal Argument.
Prakken's main research interests concern artificial intelligence & law and computational models of argumentation. Prakken is a past president of the International Association for AI & Law (IAAIL), of the Foundation for Legal Knowledge-Based Systems (JURIX) and of the steering committee of the COMMA conferences on Computational Models of Argument. He is on the editorial board of several journals, including Artificial Intelligence and Law. During 2017-2022 I was an associate editor of the Artificial Intelligence journal.

Giuseppe Greco: Integrating the logical and the data-analytic perspectives in natural language

Abstract: The slogan “You shall know a word by the company it keeps!” perfectly summarizes the so-called "distributional hypothesis": The meaning of a word is related to the distribution of the words around it, so words are represented as a vector in a multi-dimensional vector space. Applications span from computing semantic similarity between words, sentences, and documents to paraphrasing, question-answering, and summarization.
A seemingly very different approach is Lambek's program on type logical grammar, which uses logic as a grammaticality check: A grammatically correct sentence is matched with a derivable sequent of a given logic. Applications span from parsing to reasoning tasks.
Among recent attempts to integrate the two approaches, the "compositional distributional semantics" is the most successful, since it manages to integrate the very successful data analytic account of lexical meaning with the logical and grammatical approach to parsing. Unfortunately, the matching is only partial, and hence the proposed calculi are sound but not complete w.r.t. the intended semantics.
In a recent work, pivoting on the idea that vector spaces can be regarded as a type of Kripke frames, we introduce a complete vector space semantics for the Lambek calculus and some of its very well-known axiomatic extensions. This perspective makes it possible to establish a systematic connection between vector space semantics and the standard Routley-Meyer semantics of (modal) substructural logics and prove standard results as the completeness of the calculus.
In ongoing work,  we consider a "multi-type" version of the Lambek calculus for capturing grammaticality, and we introduce a "multi-type" vector space semantics for capturing lexical meaning. The multiplicity of types enables us to assign different semantic entities to the different words of language, e.g. noun vs verb, vs adjective or determiner and establishes a one-to-one correspondence between syntactic and semantic types.

Bio: Giuseppe Greco is an assistant professor at the VU, Amsterdam. He is an expert in algorithmic proof theory. His current research focuses on formal models and theories for understanding information flow and social interaction. He is interested in logical approaches for explainable AI.

Beishui Liao: Logic for New Generation AI: An Argumentation-Based Methodology

Abstract: A new generation of artificial intelligence (NGAI), currently based on big data and machine learning, follows a path of connectionism. Although this path achieves huge success in data-intensive applications under closed environments, there are some bottleneck problems, including a lack of explainability, the difficulty of ethical alignment, the weakness of cognitive reasoning, etc. To address these problems inevitably involves the depiction of information from an open, dynamic and real environment and the modeling of human reasoning and explanation mechanisms. In this talk, I will introduce a methodology based on formal argumentation, which is a general formalism for modeling various types of knowledge representation and reasoning in a context of disagreement, and is flexible enough to incorporate other types of knowledge for decision-making, such as preferences, weights, and probabilities. Meanwhile, there are various approaches for efficient computation of argumentation semantics by exploiting the locality and modularity of argumentation, and for providing explanations based on arguments and dialogues.

Bio: Beishui Liao is Full Professor of Logic and Computer Science at Zhejiang ľ�ϸ���Ӱ�� (since 2013) and Qiushi Distinguished Professor (since 2019). He obtained his PhD degree from the College of Computer Science and Technology, Zhejiang ľ�ϸ���Ӱ��, in 2006. His main research interests are AI logic, formal argumentation, and their application in agents and multi-agent systems, explainable AI and ethical AI. He is now leading a major project on Logics for a New Generation Artificial Intelligence granted by the National Social Science Fund of China. He is the director of the Institute of Logic and Cognition of Zhejiang ľ�ϸ���Ӱ��, co-director of Zhejiang ľ�ϸ���Ӱ�� - ľ�ϸ���Ӱ�� of Luxembourg Joint Lab on Advanced Intelligent Systems and REasoning (ZLAIRE), and a guest professor of Luxembourg ľ�ϸ���Ӱ�� (since 2014). He visited the ľ�ϸ���Ӱ�� of Texas at Austin from Jul 2009 to Jul 2011, the ľ�ϸ���Ӱ�� of Brescia from Jul to Oct 2014, and the ľ�ϸ���Ӱ�� of Oxford and the ľ�ϸ���Ӱ�� of Cambridge (2018, 2019, 2020).

Fei Liang: Modeling indicative conditionals as epistemic filter conditionals

Abstract:  Deng & Lee (2021)  give an “extrapolationist” causal­ modeling account of the semantics of indicative conditionals. It requires us to have a priority relation on variables and to calculate the induced ordering for submodels, in terms of which the set of minimally extrapolated submodels is defined to account for the semantics of indicative conditionals. In this talk, we  argue that this extrapolationist proposal has unnecessary complexity and may yield incorrect results first, and then offer an “epistemic filter” account of indicative conditionals that can successfully avoid these problems.  This work is joint with Wenfang Wang from Shandong university.

Bio: Fei Liang is an associate professor at Shandong ľ�ϸ���Ӱ��. He is an expert in structural proof theory, algebraic logic and philosophical logic.

Apostolos Tzimoulis: Categorization and deliberation

Abstract: In this talk I will explore the role of the epistemic stances of different agents (decision-makers) played in the decision-making processes. Epistemic stances can be modelled as interrogative agendas, a notion introduced in epistemology and formal philosophy indicating the set of questions individual agents (or groups of agents) are interested in, or what they consider relevant for deciding, relative to a certain circumstance (independently of whether they utter the questions explicitly). Interrogative agendas might differ for the same agent in different moments or in different contexts. Deliberation and negotiation processes can be understood in terms of whether and how decision-makers/negotiators succeed in modifying their own interrogative agendas or those of their counterparts, and the outcomes of these processes can be described in terms of the “common ground” agenda reached.

Bio: Apostolos Tzimoulis is an assistant professor at the Vrije Universiteit Amsterdam. He is an expert in algebraic logic and its applications. His research focuses on formal models and theories for understanding multi-agent interaction and social behaviour.