The Second International Workshop on Logics of MultiAgent Systems (LMAS-2023) aims at the study of abilities, knowledge, norms and actions in logics of Multiagent systems. Please click here for the previous workshop.
Beijing Time: March 25th-26th, 2023
Offline Site: No.17 Teaching Building, Southwest University (西南大学国家治理学院（17教，原文社院）)
Workshop Chair:
Meiyun Guo (Southwest University ) & Thomas Ågotnes (University of Bergen // Southwest University)
Organization Committee:
Meiyun Guo (guomy007@swu.edu.cn): Chair
Zuojun Xiong (zuojunxiong@swu.edu.cn): registration, program, invitation letters
Junli Jiang (walk08@swu.edu.cn) : accommodations, tea breaks
Yan Sun (sunyang2082@swu.edu.cn): site maintanance, workshop posters, news
Sponsors:
西南大学国家治理学院（College of State Governance, SWU）
西南大学逻辑与智能研究中心（Institute of Logic and Intelligence, SWU）
国家社科基金重点项目“面向概率更新的动态认知逻辑研究” (Key Project of the National Social Science Fund of China: Towards to the Probabilities updates of DEL )
西南大学引进人才项目“推理与互动的逻辑研究”（Imported Experts Project of the Southwest University: On the Logical study of Reasoning and Interaction）
西南大学创新研究2035先导计划哲学百年中国化团队 (Innovation Research 2035 Pilot Plan of Southwest University)
ABSTRACT(in presentation order)
Thomas Ågotnes (University of Bergen, Norway & Southwest University, China): Weak Conjunctive Closure in Modal Logic
Normal modal logics are closed under conjunctive closure. There are, however, interesting non-normal logics that are not, but which nevertheless satisfy a weak form of conjunctive closure. An arch example is a notion of group knowledge in epistemic logic: somebody-knows. While something is general knowledge if it is known by *everyone*, this notion holds if it is known by *someone*. Somebody-knows is thus weaker than general knowledge but stronger than distributed knowledge. We introduce a modality for somebody-knows in the style of standard group knowledge modalities, and study its properties. Unlike most other group knowledge modalities, somebody-knows is not a normal modality; in particular it lacks the conjunctive closure property. We provide an equivalent neighbourhood semantics for the language with a single somebody-knows modality, together with a completeness result: the somebody-knows modalities are completely characterised by the modal logic EMN extended with a particular weak conjunctive closure axiom. The neighbourhood semantics and the completeness and complexity results also carry over other logics with weak conjunctive closure, including the logic of so-called local reasoning (Fagin et al., 1995) with bounded ``frames of mind'', correcting an existing completeness result in the literature (Allen 2005). The talk is based on joint work with Yi N. Wang (Sun Yat-sen University).
Chiaki Sakama (Wakayama University, Japan): A Formal Account of Hollis's Paradox
In his article in 1984, Martin Hollis introduces the following paradox: Two persons A and B each choose a positive integer and communicate their choices privately to another person C. C tells them that they have chosen different integers and C predicts that neither of them can work out whose number is greater. A reasons that B cannot have chosen 1, since, if he did, he would then be able to infer that A had chosen a greater number; for the same reason, B knows that A did not choose 1. Given this fact, no one can choose 2. This is because if B chose 2, he would know that A's number is greater; and if A chose it, A would know that B is greater. Repeating similar inference leads to the conclusion that neither of them can have chosen any number. This contradicts the fact that A and B have chosen integers. In this talk, we formulate the paradox using an epistemic (doxastic) logic and propose a solution.
Katsuhiko Sano (Hokkaido University, Japan): Axiomatizing a Simple Logic of the Hide and Seek Game
This talk provides a semantically complete axiomatization of a hybridization of a simple modal logic of the hide and seek game by Li et al. (2021). To describe the winning condition of the seeker of the game the logic has an equality constant, which is similar to the diagonal constant introduced in the product of modal logics by Kurucz (2009). While the original simple modal logic of the hide and seek game was shown to be undecidable in Li, et al. (2021), it is still open that the logic has a semantically complete axiomatization. In this talk, we show that a hybridization (a machinery of hybrid logic, in particular, that for the product of modal logic, cf. Sano (2010)) provides an answer to the open question, provided the syntax is expanded with nominals and satisfaction operators. This is joint work with Dazhu Li (Chinese Academy of Sciences) and Fenrong Liu (Tsinghua).
Yu Wei (East China Normal University, China): Understanding Understanding Why
Epistemic logic pays barely any attention to the notion of understanding, which stands in total contrast to the current situation in epistemology and in philosophy of science. This work aims to bridge the gap between logic and philosophy by developing a formal treatment of “understanding why”. It is generally acknowledged that understanding why moves beyond knowing why. Inspired by philosophical discussions, we consider whereas knowing why requires knowing horizontal explanations, understanding why additionally requires vertical explanations. Based on justification logic and existing logical work for knowing why, we build up an axiomatization introducing vertical explanations in the semantics, and show the logic could accommodate different philosophical viewpoints via adding different conditions to the models. Meanwhile, in light of this semantic idea, it becomes more interesting when other possible logical forms and variants of understanding why are considered and compared in a more general framework for explanatory knowledge and understanding.
Zhanglyu Li(Southwest University, China): Causation as a Tool or Causation as a Target
Judea Pearl's and David Lewis' theory of causation both hold important positions in the field of causation studies, but it seems their difference and applicability still need to be compared and discussed. Using Pearl's Logic of Structure-Based Counterfactuals, we analyze the three predicaments Lewis' theory of causation encountered, i.e., pre-emption, epiphenomena and cause-effect inversion, and show these problems can be answered easily with Pearl's theory. This "easy answer" reveals the logic preferences of the two theories: Pearl believes "the truth value doesn't influence the causal relation", but Lewis insists "the truth value do change the causal relation". Their logical preferences explain the major difference between Pearl's and Lewis' theory of causation: causation as a tool or causation as a target. Pearl's theory is more efficient to deal with the "tool-style" problems, which treats causation as a presupposed structure; while Lewis' theory is more suitable for "target-style" problems, which reject causation as an initial concept and try to find its metaphysical grounding.
Yanjun Li (Nankai University, China): Tableaux for Knowing-How Logic
We will present a tableau system for the knowing-how logic via simple plans, and we show that there is an algorithm that runs in polynomial space for deciding whether a formula of the knowing-how logic is satisfiable.
Valentin Goranko (Stockholm University, Sweden): A Logic for Conditional Strategic Reasoning
I will present the logic ConStR, featuring new modal operators for capturing conditional strategic reasoning. This is reasoning about strategic abilities, conditional on the agents' knowledge of the objectives and possible actions of the other (cooperating or competing) agents. I will present formal semantics for the logic ConStR and will discuss and illustrate its expressiveness. Then I will present an axiomatic system for it and will outline an ongoing work on proving completeness of some of its fragments. This talk is based on joint work with Fengkui Ju and YinFeng Li from Beijing Normal University.
Fengkui Ju (Beijing Normal University, China) Logic for strategic reasoning in dynamic strategy contexts
Strategic reasoning is usually performed in specific strategy contexts, concerning which strategies are in consideration. Strategic reasoning can involve the change of strategy contexts. In this talk, we present a logic for strategic reasoning involving the change of strategy contexts caused by commitments to strategies. The logic has two featured formulas: for every strategy of a coalition, if the coalition commits to it, a formula is guaranteed to be true; for some strategy of a coalition, if the coalition commits to it, a formula is guaranteed to be true. Commitments to strategies shrink strategy contexts, which are defined as sets of strategy profiles. This work is close to several previous works in the literature, and we make detailed comparisons to them.This is joint work with Valentin Goranko from Stockholm University.
Fernando Velázquez-Quesada (University of Bergen, Norway): Partial Communication and Arbitrary Partial Communication
Communication within groups of agents has been lately the focus of research in dynamic epistemic logic (DEL). This talk discusses a form of partial (more precisely, topic-based) communication. particularly useful in situations in which sharing all information is not feasible/advisable. We will present the framework, and then discuss some of its basics results (axiom system, structural invariance, expressivity, model checking). We will also compare partial communication with public announcements, probably the most well-known type of communication in DEL. It will be shown that, update-wise, the settings are incomparable. Then, time permitting, we will shift our attention to arbitrary topic-based communication, doing so by extending the language with a modality that quantifies over the topics the agents can ‘talk about’. We will discuss again some basic results, closing with a comparison between arbitrary partial communication and arbitrary public announcements.
Zoé Christoff (University of Groningen, Netherlands): The illusion of majority in social networks
The popularity of an opinion in one's direct circles is not necessarily a good indicator of its popularity in one's entire community. For instance, when confronted with a majority of opposing opinions in one's circles, one might get the impression that they belong to a minority. From this perspective, network structures act as a sort of "sample distortion filter". However, the way a social network is wired also constrains how much distortion can actually occur. In this paper, we discuss which classes of networks allow for a majority of agents to be given the impression that they belong to a minority, that is, to be in a ``majority illusion''. This is joint work with Maaike Los and Davide Grossi.
Kai Li (China University of Political Science and Law, China): Norms as Speech Acts in DEL
Jörgensen’s dilemma has been haunting the logic concerning norms for almost a century. In response to its challenge, some logicians take norms as imperatives, and proposed imperative semantic for deontic propositions. In social science, however, a norm is interpreted as an equilibrium selection device. Following the speech-act theory, I treat a norm as a dynamic update on an epistemic game tree that affects population's equilibrium. This hybrid approach provides a more explicit explanation for how a norm generates an ought than the imperative semantic. Although this framework is not designed for non-monotonic normative reasoning, it gives some insights for both logical approach and game theoretical approach. On one hand norms are not all imperatives as many logicians would believe. On the other hand, values sometimes are the results of normative reasoning, but can only be seen as exterior parameters for utility function in game theoretical approach.
Dazhu Li (University of Chinese Academy of Sciences, China): Action-information interplay in the cops and robber game
The cops and robber game can be considered as a testbed for many computational issues in the pursuit-evasion environments. We propose a formal framework, namely, LCR, to make the core notions of the game precise, for instance, players’ positions, observational power, knowledge, and also the winning conditions of the game. Applying LCR to analyze the game, we obtain a novel and automated way of tracking interactions between players and characterizing their information-updates during the play. The update mechanism is defined by a new dynamic operator and its formal properties are studied and illustrated. We show that the ‘static fragment’ of LCR without the dynamic operators has a decidable satisfiability problem and a P-complete model-checking problem. An equivalent modal framework for LCR is proposed, which provides us with a better insight in understanding the interplay between players’ information and actions in these games.
Jie Fan (University of Chinese Academy of Sciences, China): Unknown Truths and Unknowable Truths
Notions of unknown truths and unknowable truths are important in formal epistemology, which are related to each other in e.g., Fitch's paradox of knowability. Although there has been some research on the two notions, there seems to be no logical research on unknowable truths. In this paper, we propose a logic of unknowable truths, investigate the logical properties of unknown truths and unknowable truths, which includes the similarities of the two notions and the relationship between the two notions, and axiomatize this logic. We also propose an alternative semantics of unknowable truths, which relates the notion of unknowable truths to the notion of unknown truths more suitably.
Jeremy Seligman (The University of Auckland, New Zealand) Sudoku for Logicians
Logic puzzles present what Johan van Benthem has called "the Sudoku challenge", namely, to account for the epistemology of inference when the solution is entailed by the puzzle itself. At an abstract level, this is the problem of epistemic closure. Most systems of epistemic logic validate the principle that a rational agent knows whatever is entailed by their knowledge, and yet this renders it difficult to account for the epistemic gain afforded by inference. In this talk I will try to show how a closer look at Sudoku solving presents a much richer view of the epistemology of deductive reasoning, especially when one considers Sudoku variants in which additional rules create the possibility of many new strategies. I'll also attempt to make this relevant to multi-agent systems!
